representation of conceptual knowledge

• Table of Contents
• Random Entry
• Chronological
• Editorial Information
• About the SEP
• Editorial Board
• How to Cite the SEP
• Special Characters
• Advanced Tools
• Support the SEP
• PDFs for SEP Friends
• Make a Donation
• SEPIA for Libraries
• Entry Contents

Bibliography

Academic tools.

• Friends PDF Preview
• Author and Citation Info
• Back to Top

The Analysis of Knowledge

For any person, there are some things they know, and some things they don’t. What exactly is the difference? What does it take to know something? It’s not enough just to believe it—we don’t know the things we’re wrong about. Knowledge seems to be more like a way of getting at the truth. The analysis of knowledge concerns the attempt to articulate in what exactly this kind of “getting at the truth” consists.

More particularly, the project of analysing knowledge is to state conditions that are individually necessary and jointly sufficient for propositional knowledge, thoroughly answering the question, what does it take to know something? By “propositional knowledge”, we mean knowledge of a proposition—for example, if Susan knows that Alyssa is a musician, she has knowledge of the proposition that Alyssa is a musician. Propositional knowledge should be distinguished from knowledge of “acquaintance”, as obtains when Susan knows Alyssa. The relation between propositional knowledge and the knowledge at issue in other “knowledge” locutions in English, such as knowledge-where (“Susan knows where she is”) and especially knowledge-how (“Susan knows how to ride a bicycle”) is subject to some debate (see Stanley 2011 and his opponents discussed therein).

The propositional knowledge that is the analysandum of the analysis of knowledge literature is paradigmatically expressed in English by sentences of the form “ S knows that p ”, where “ S ” refers to the knowing subject, and “ p ” to the proposition that is known. A proposed analysis consists of a statement of the following form: S knows that p if and only if j , where j indicates the analysans: paradigmatically, a list of conditions that are individually necessary and jointly sufficient for S to have knowledge that p .

It is not enough merely to pick out the actual extension of knowledge. Even if, in actual fact, all cases of S knowing that p are cases of j , and all cases of the latter are cases of the former, j might fail as an analysis of knowledge. For example, it might be that there are possible cases of knowledge without j , or vice versa. A proper analysis of knowledge should at least be a necessary truth. Consequently, hypothetical thought experiments provide appropriate test cases for various analyses, as we shall see below.

Even a necessary biconditional linking knowledge to some state j would probably not be sufficient for an analysis of knowledge, although just what more is required is a matter of some controversy. According to some theorists, to analyze knowledge is literally to identify the components that make up knowledge—compare a chemist who analyzes a sample to learn its chemical composition. On this interpretation of the project of analyzing knowledge, the defender of a successful analysis of knowledge will be committed to something like the metaphysical claim that what it is for S to know p is for some list of conditions involving S and p to obtain. Other theorists think of the analysis of knowledge as distinctively conceptual —to analyse knowledge is to limn the structure of the concept of knowledge. On one version of this approach, the concept knowledge is literally composed of more basic concepts, linked together by something like Boolean operators. Consequently, an analysis is subject not only to extensional accuracy, but to facts about the cognitive representation of knowledge and other epistemic notions. In practice, many epistemologists engaging in the project of analyzing knowledge leave these metaphilosophical interpretive questions unresolved; attempted analyses, and counterexamples thereto, are often proposed without its being made explicit whether the claims are intended as metaphysical or conceptual ones. In many cases, this lack of specificity may be legitimate, since all parties tend to agree that an analysis of knowledge ought at least to be extensionally correct in all metaphysically possible worlds. As we shall see, many theories have been defended and, especially, refuted, on those terms.

The attempt to analyze knowledge has received a considerable amount of attention from epistemologists, particularly in the late 20 th Century, but no analysis has been widely accepted. Some contemporary epistemologists reject the assumption that knowledge is susceptible to analysis.

1.1 The Truth Condition

1.2 the belief condition, 1.3 the justification condition, 2. lightweight knowledge, 3. the gettier problem, 4. no false lemmas, 5.1 sensitivity, 5.3 relevant alternatives, 6.1 reliabilist theories of knowledge, 6.2 causal theories of knowledge, 7. is knowledge analyzable, 8. epistemic luck, 9. methodological options, 10.1 the “aaa” evaluations, 10.2 fake barn cases, 11. knowledge first, 12. pragmatic encroachment, 13. contextualism, other internet resources, related entries, 1. knowledge as justified true belief.

There are three components to the traditional (“tripartite”) analysis of knowledge. According to this analysis, justified, true belief is necessary and sufficient for knowledge.

• S believes that p ;
• S is justified in believing that p .

The tripartite analysis of knowledge is often abbreviated as the “JTB” analysis, for “justified true belief”.

Much of the twentieth-century literature on the analysis of knowledge took the JTB analysis as its starting-point. It became something of a convenient fiction to suppose that this analysis was widely accepted throughout much of the history of philosophy. In fact, however, the JTB analysis was first articulated in the twentieth century by its attackers. [ 1 ] Before turning to influential twentieth-century arguments against the JTB theory, let us briefly consider the three traditional components of knowledge in turn.

Most epistemologists have found it overwhelmingly plausible that what is false cannot be known. For example, Hillary Clinton did not win the 2016 US Presidential election. Consequently, nobody knows that Hillary Clinton won the election. One can only know things that are true.

Sometimes when people are very confident of something that turns out to be wrong, we use the word “knows” to describe their situation. Many people expected Clinton to win the election. Speaking loosely, one might even say that many people “knew” that Clinton would win the election—until she lost. Hazlett (2010) argues on the basis of data like this that “knows” is not a factive verb. [ 2 ] Hazlett’s diagnosis is deeply controversial; most epistemologists will treat sentences like “I knew that Clinton was going to win” as a kind of exaggeration—as not literally true.

Something’s truth does not require that anyone can know or prove that it is true. Not all truths are established truths. If you flip a coin and never check how it landed, it may be true that it landed heads, even if nobody has any way to tell. Truth is a metaphysical , as opposed to epistemological , notion: truth is a matter of how things are , not how they can be shown to be. So when we say that only true things can be known, we’re not (yet) saying anything about how anyone can access the truth. As we’ll see, the other conditions have important roles to play here. Knowledge is a kind of relationship with the truth—to know something is to have a certain kind of access to a fact. [ 3 ]

The belief condition is only slightly more controversial than the truth condition. The general idea behind the belief condition is that you can only know what you believe. Failing to believe something precludes knowing it. “Belief” in the context of the JTB theory means full belief, or outright belief. In a weak sense, one might “believe” something by virtue of being pretty confident that it’s probably true—in this weak sense, someone who considered Clinton the favourite to win the election, even while recognizing a nontrivial possibility of her losing, might be said to have “believed” that Clinton would win. Outright belief is stronger (see, e.g., Fantl & McGrath 2009: 141; Nagel 2010: 413–4; Williamson 2005: 108; or Gibbons 2013: 201.). To believe outright that p , it isn’t enough to have a pretty high confidence in p ; it is something closer to a commitment or a being sure. [ 4 ]

Although initially it might seem obvious that knowing that p requires believing that p , a few philosophers have argued that knowledge without belief is indeed possible. Suppose Walter comes home after work to find out that his house has burned down. He says: “I don’t believe it”. Critics of the belief condition might argue that Walter knows that his house has burned down (he sees that it has), but, as his words indicate, he does not believe it. The standard response is that Walter’s avowal of disbelief is not literally true; what Walter wishes to convey by saying “I don’t believe it” is not that he really does not believe that his house has burned down, but rather that he finds it hard to come to terms with what he sees. If he genuinely didn’t believe it, some of his subsequent actions, such as phoning his insurance company, would be rather mysterious.

A more serious counterexample has been suggested by Colin Radford (1966). Suppose Albert is quizzed on English history. One of the questions is: “When did Queen Elizabeth die?” Albert doesn’t think he knows, but answers the question correctly. Moreover, he gives correct answers to many other questions to which he didn’t think he knew the answer. Let us focus on Albert’s answer to the question about Elizabeth:

• (E) Elizabeth died in 1603.

Radford makes the following two claims about this example:

• Albert does not believe (E).
• Albert knows (E).

Radford’s intuitions about cases like these do not seem to be idiosyncratic; Myers-Schutz & Schwitzgebel (2013) find evidence suggesting that many ordinary speakers tend to react in the way Radford suggests. In support of (a), Radford emphasizes that Albert thinks he doesn’t know the answer to the question. He doesn’t trust his answer because he takes it to be a mere guess. In support of (b), Radford argues that Albert’s answer is not at all just a lucky guess. The fact that he answers most of the questions correctly indicates that he has actually learned, and never forgotten, such historical facts.

Since he takes (a) and (b) to be true, Radford holds that belief is not necessary for knowledge. But either of (a) and (b) might be resisted. One might deny (a), arguing that Albert does have a tacit belief that (E), even though it’s not one that he thinks amounts to knowledge. David Rose and Jonathan Schaffer (2013) take this route. Alternatively, one might deny (b), arguing that Albert’s correct answer is not an expression of knowledge, perhaps because, given his subjective position, he does not have justification for believing (E). The justification condition is the topic of the next section.

Why is condition (iii) necessary? Why not say that knowledge is true belief? The standard answer is that to identify knowledge with true belief would be implausible because a belief might be true even though it is formed improperly. Suppose that William flips a coin, and confidently believes—on no particular basis—that it will land tails. If by chance the coin does land tails, then William’s belief was true; but a lucky guess such as this one is no knowledge. For William to know, his belief must in some epistemic sense be proper or appropriate: it must be justified . [ 5 ]

Socrates articulates the need for something like a justification condition in Plato’s Theaetetus , when he points out that “true opinion” is in general insufficient for knowledge. For example, if a lawyer employs sophistry to induce a jury into a belief that happens to be true, this belief is insufficiently well-grounded to constitute knowledge.

1.3.1 Approaches to Justification

There is considerable disagreement among epistemologists concerning what the relevant sort of justification here consists in. Internalists about justification think that whether a belief is justified depends wholly on states in some sense internal to the subject. According to one common such sense of “internal”, only those features of a subject’s experience which are directly or introspectively available count as “internal”—call this “access internalism”. According to another, only intrinsic states of the subject are “internal”—call this “state internalism”. See Feldman & Conee 2001 for the distinction.

Conee and Feldman present an example of an internalist view. They have it that S ’s belief that p is justified if and only if believing that p is the attitude towards p that best fits S ’s evidence, where the latter is understood to depend only on S ’s internal mental states. Conee and Feldman call their view “evidentialism”, and characterize this as the thesis that justification is wholly a matter of the subject’s evidence. Given their (not unsubstantial) assumption that what evidence a subject has is an internal matter, evidentialism implies internalism. [ 6 ] Externalists about justification think that factors external to the subject can be relevant for justification; for example, process reliabilists think that justified beliefs are those which are formed by a cognitive process which tends to produce a high proportion of true beliefs relative to false ones. [ 7 ] We shall return to the question of how reliabilist approaches bear on the analysis of knowledge in §6.1 .

1.3.2 Kinds of Justification

It is worth noting that one might distinguish between two importantly different notions of justification, standardly referred to as “propositional justification” and “doxastic justification”. (Sometimes “ ex ante ” justification and “ ex post ” justification, respectively.) [ 8 ] Unlike that between internalist and externalist approaches to justification, the distinction between propositional and doxastic justification does not represent a conflict to be resolved; it is a distinction between two distinct properties that are called “justification”. Propositional justification concerns whether a subject has sufficient reason to believe a given proposition; [ 9 ] doxastic justification concerns whether a given belief is held appropriately. [ 10 ] One common way of relating the two is to suggest that propositional justification is the more fundamental, and that doxastic justification is a matter of a subject’s having a belief that is appropriately responsive to or based on their propositional justification.

The precise relation between propositional and doxastic justification is subject to controversy, but it is uncontroversial that the two notions can come apart. Suppose that Ingrid ignores a great deal of excellent evidence indicating that a given neighborhood is dangerous, but superstitiously comes to believe that the neighborhood is dangerous when she sees a black cat crossing the street. Since forming beliefs on the basis of superstition is not an epistemically appropriate way of forming beliefs, Ingrid’s belief is not doxastically justified; nevertheless, she does have good reason to believe as she does, so she does have propositional justification for the proposition that the neighborhood is dangerous.

Since knowledge is a particularly successful kind of belief, doxastic justification is a stronger candidate for being closely related to knowledge; the JTB theory is typically thought to invoke doxastic justification (but see Lowy 1978).

Some epistemologists have suggested that there may be multiple senses of the term “knowledge”, and that not all of them require all three elements of the tripartite theory of knowledge. For example, some have argued that there is, in addition to the sense of “knowledge” gestured at above, another, weak sense of “knowledge”, that requires only true belief (see for example Hawthorne 2002 and Goldman & Olsson 2009; the latter contains additional relevant references). This view is sometimes motivated by the thought that, when we consider whether someone knows that p , or wonder which of a group of people know that p , often, we are not at all interested in whether the relevant subjects have beliefs that are justified; we just want to know whether they have the true belief. For example, as Hawthorne (2002: 253–54) points out, one might ask how many students know that Vienna is the capital of Austria; the correct answer, one might think, just is the number of students who offer “Vienna” as the answer to the corresponding question, irrespective of whether their beliefs are justified. Similarly, if you are planning a surprise party for Eugene and ask whether he knows about it, “yes” may be an appropriate answer merely on the grounds that Eugene believes that you are planning a party.

One could allow that there is a lightweight sense of knowledge that requires only true belief; another option is to decline to accept the intuitive sentences as true at face value. A theorist might, for instance, deny that sentences like “Eugene knows that you are planning a party”, or “eighteen students know that Vienna is the capital of Austria” are literally true in the envisaged situations, explaining away their apparent felicity as loose talk or hyperbole.

Even among those epistemologists who think that there is a lightweight sense of “knows” that does not require justification, most typically admit that there is also a stronger sense which does, and that it is this stronger state that is the main target of epistemological theorizing about knowledge. In what follows, we will set aside the lightweight sense, if indeed there be one, and focus on the stronger one.

Few contemporary epistemologists accept the adequacy of the JTB analysis. Although most agree that each element of the tripartite theory is necessary for knowledge, they do not seem collectively to be sufficient . There seem to be cases of justified true belief that still fall short of knowledge. Here is one kind of example:

Imagine that we are seeking water on a hot day. We suddenly see water, or so we think. In fact, we are not seeing water but a mirage, but when we reach the spot, we are lucky and find water right there under a rock. Can we say that we had genuine knowledge of water? The answer seems to be negative, for we were just lucky. (quoted from Dreyfus 1997: 292)

This example comes from the Indian philosopher Dharmottara, c. 770 CE. The 14 th -century Italian philosopher Peter of Mantua presented a similar case:

Let it be assumed that Plato is next to you and you know him to be running, but you mistakenly believe that he is Socrates, so that you firmly believe that Socrates is running. However, let it be so that Socrates is in fact running in Rome; however, you do not know this. (from Peter of Mantua’s De scire et dubitare , given in Boh 1985: 95)

Cases like these, in which justified true belief seems in some important sense disconnected from the fact, were made famous in Edmund Gettier’s 1963 paper, “Is Justified True Belief Knowledge?”. Gettier presented two cases in which a true belief is inferred from a justified false belief. He observed that, intuitively, such beliefs cannot be knowledge; it is merely lucky that they are true.

In honour of his contribution to the literature, cases like these have come to be known as “Gettier cases”. Since they appear to refute the JTB analysis, many epistemologists have undertaken to repair it: how must the analysis of knowledge be modified to accommodate Gettier cases? This is what is commonly referred to as the “Gettier problem”.

Above, we noted that one role of the justification is to rule out lucky guesses as cases of knowledge. A lesson of the Gettier problem is that it appears that even true beliefs that are justified can nevertheless be epistemically lucky in a way inconsistent with knowledge.

Epistemologists who think that the JTB approach is basically on the right track must choose between two different strategies for solving the Gettier problem. The first is to strengthen the justification condition to rule out Gettier cases as cases of justified belief. This was attempted by Roderick Chisholm; [ 11 ] we will refer to this strategy again in §7 below. The other is to amend the JTB analysis with a suitable fourth condition, a condition that succeeds in preventing justified true belief from being “gettiered”. Thus amended, the JTB analysis becomes a JTB+ X account of knowledge, where the “ X ” stands for the needed fourth condition.

Let us consider an instance of this attempt to articulate a “degettiering” condition.

According to one suggestion, the following fourth condition would do the trick:

• S ’s belief that p is not inferred from any falsehood. [ 12 ]

In Gettier’s cases, the justified true belief is inferred from a justified false belief. So condition (iv) explains why it isn’t knowledge. However, this “no false lemmas” proposal is not successful in general. There are examples of Gettier cases that need involve no inference; therefore, there are possible cases of justified true belief without knowledge, even though condition (iv) is met. Suppose, for example, that James, who is relaxing on a bench in a park, observes an apparent dog in a nearby field. So he believes

• There is a dog in the field.

Suppose further that the putative dog is actually a robot dog so perfect that it could not be distinguished from an actual dog by vision alone. James does not know that such robot dogs exist; a Japanese toy manufacturer has only recently developed them, and what James sees is a prototype that is used for testing the public’s response. Given these assumptions, (d) is of course false. But suppose further that just a few feet away from the robot dog, there is a real dog, concealed from James’s view. Given this further assumption, James’s belief in (d) is true. And since this belief is based on ordinary perceptual processes, most epistemologists will agree that it is justified. But as in Gettier’s cases, James’s belief appears to be true only as a matter of luck, in a way inconsistent with knowledge. So once again, what we have before us is a justified true belief that isn’t knowledge. [ 13 ] Arguably, this belief is directly justified by a visual experience; it is not inferred from any falsehood. If so, then the JTB account, even if supplemented with (iv) , gives us the wrong result that James knows (d).

Another case illustrating that clause (iv) won’t do the job is the well-known Barn County case (Goldman 1976). Suppose there is a county in the Midwest with the following peculiar feature. The landscape next to the road leading through that county is peppered with barn-facades: structures that from the road look exactly like barns. Observation from any other viewpoint would immediately reveal these structures to be fakes: devices erected for the purpose of fooling unsuspecting motorists into believing in the presence of barns. Suppose Henry is driving along the road that leads through Barn County. Naturally, he will on numerous occasions form false beliefs in the presence of barns. Since Henry has no reason to suspect that he is the victim of organized deception, these beliefs are justified. Now suppose further that, on one of those occasions when he believes there is a barn over there, he happens to be looking at the one and only real barn in the county. This time, his belief is justified and true. But since its truth is the result of luck, it is exceedingly plausible to judge that Henry’s belief is not an instance of knowledge. Yet condition (iv) is met in this case. His belief is not the result of any inference from a falsehood. Once again, we see that (iv) does not succeed as a general solution to the Gettier problem.

5. Modal Conditions

Another candidate fourth condition on knowledge is sensitivity . Sensitivity, to a first approximation, is this counterfactual relation:

S ’s belief that p is sensitive if and only if, if p were false, S would not believe that p . [ 14 ]

A sensitivity condition on knowledge was defended by Robert Nozick (1981). Given a Lewisian (Lewis 1973) semantics for counterfactual conditionals, the sensitivity condition is equivalent to the requirement that, in the nearest possible worlds in which not- p , the subject does not believe that p .

One motivation for including a sensitivity condition in an analysis of knowledge is that there seems to be an intuitive sense in which knowledge requires not merely being correct, but tracking the truth in other possible circumstances. This approach seems to be a plausible diagnosis of what goes wrong in at least some Gettier cases. For example, in Dharmottara’s desert water case, your belief that there is water in a certain location appears to be insensitive to the fact of the water. For if there were no water there, you would have held the same belief on the same grounds— viz. , the mirage.

However, it is doubtful that a sensitivity condition can account for the phenomenon of Gettier cases in general. It does so only in cases in which, had the proposition in question been false, it would have been believed anyway. But, as Saul Kripke (2011: 167–68) has pointed out, not all Gettier cases are like this. Consider for instance the Barn County case mentioned above. Henry looks at a particular location where there happens to be a barn and believes there to be a barn there. The sensitivity condition rules out this belief as knowledge only if, were there no barn there, Henry would still have believed there was. But this counterfactual may be false, depending on how the Barn County case is set up. For instance, it is false if the particular location Henry is examining is not one that would have been suitable for the erecting of a barn façade. Relatedly, as Kripke has also indicated (2011: 186), if we suppose that barn facades are always green, but genuine barns are always red, Henry’s belief that he sees a red barn will be sensitive, even though his belief that he sees a barn will not. (We assume Henry is unaware that colour signifies anything relevant.) Since intuitively, the former belief looks to fall short of knowledge in just the same way as the latter, a sensitivity condition will only handle some of the intuitive problems deriving from Gettier cases.

Most epistemologists today reject sensitivity requirements on knowledge. The chief motivation against a sensitivity condition is that, given plausible assumptions, it leads to unacceptable implications called “abominable conjunctions”. [ 15 ] To see this, suppose first that skepticism about ordinary knowledge is false—ordinary subjects know at least many of the things we ordinarily take them to know. For example, George, who can see and use his hands perfectly well, knows that he has hands. This is of course perfectly consistent with a sensitivity condition on knowledge, since if George did not have hands—if they’d been recently chopped off, for instance—he would not believe that he had hands.

Now imagine a skeptical scenario in which George does not have hands. Suppose that George is the victim of a Cartesian demon, deceiving him into believing that he has hands. If George were in such a scenario, of course, he would falsely believe himself not to be in such a scenario. So given the sensitivity condition, George cannot know that he is not in such a scenario.

Although these two verdicts—the knowledge-attributing one about ordinary knowledge, and the knowledge-denying one about the skeptical scenario—are arguably each intuitive, it is intuitively problematic to hold them together. Their conjunction is, in DeRose’s term, abominable: “George knows that he has hands, but he doesn’t know that he’s not the handless victim of a Cartesian demon”. A sensitivity condition on knowledge, combined with the nonskeptical claim that there is ordinary knowledge, seems to imply such abominable conjunctions. [ 16 ]

Most contemporary epistemologists have taken considerations like these to be sufficient reason to reject sensitivity conditions. [ 17 ] However, see Ichikawa (2011a) for an interpretation and endorsement of the sensitivity condition according to which it may avoid commitment to abominable conjunctions.

Although few epistemologists today endorse a sensitivity condition on knowledge, the idea that knowledge requires a subject to stand in a particular modal relation to the proposition known remains a popular one. In his 1999 paper, “How to Defeat Opposition to Moore”, Ernest Sosa proposed that a safety condition ought to take the role that sensitivity was intended to play. Sosa characterized safety as the counterfactual contrapositive of sensitivity.

Sensitivity: If p were false, S would not believe that p .

Safety: If S were to believe that p , p would not be false. [ 18 ]

Although contraposition is valid for the material conditional \((A \supset B\) iff \(\mathord{\sim} B \supset \mathord{\sim}A)\), Sosa suggests that it is invalid for counterfactuals, which is why sensitivity and safety are not equivalent. An example of a safe belief that is not sensitive, according to Sosa, is the belief that a distant skeptical scenario does not obtain. If we stipulate that George, discussed above, has never been at risk of being the victim of a Cartesian demon—because, say, Cartesian demons do not exist in George’s world—then George’s belief that he is not such a victim is a safe one, even though we saw in the previous section that it could not be sensitive. Notice that although we stipulated that George is not at risk of deceit by Cartesian demons, we did not stipulate that George himself had any particular access to this fact. Unless he does, safety, like sensitivity, will be an externalist condition on knowledge in the “access” sense. It is also externalist in the “state” sense, since the truth of the relevant counterfactuals will depend on features outside the subject.

Characterizing safety in these counterfactual terms depends on substantive assumptions about the semantics of counterfactual conditionals. [ 19 ] If we were to accept, for instance, David Lewis’s or Robert Stalnaker’s treatment of counterfactuals, including a strong centering condition according to which the actual world is always uniquely closest, all true beliefs would count as safe according to the counterfactual analysis of safety. [ 20 ] Sosa intends the relevant counterfactuals to be making a stronger claim, requiring roughly that in all nearby worlds in which S believes that p , p is not false.

Rather than resting on a contentious treatment of counterfactuals, then, it may be most perspicuous to understand the safety condition more directly in these modal terms, as Sosa himself often does:

In all nearby worlds where S believes that p , p is not false.

Whether a JTB+safety analysis of knowledge could be successful is somewhat difficult to evaluate, given the vagueness of the stated “nearby” condition. The status of potential counterexamples will not always be straightforward to apply. For example, Juan Comesaña (2005) presents a case he takes to refute the requirement that knowledge be safe. In Comesaña’s example, the host of a Halloween party enlists Judy to direct guests to the party. Judy’s instructions are to give everyone the same directions, which are in fact accurate, but that if she sees Michael, the party will be moved to another location. (The host does not want Michael to find the party.) Suppose Michael never shows up. If a given guest does not, but very nearly does, decide to wear a very realistic Michael costume to the party, then his belief, based in Judy’s testimony, about the whereabouts of the party will be true, but could, Comesaña says, easily have been false. (Had he merely made a slightly different choice about his costume, he would have been deceived.) Comesaña describes the case as a counterexample to a safety condition on knowledge. However, it is open to a safety theorist to argue that the relevant skeptical scenario, though possible and in some sense nearby, is not near enough in the relevant respect to falsify the safety condition. Such a theorist would, if she wanted the safety condition to deliver clear verdicts, face the task of articulating just what the relevant notion of similarity amounts to (see also Bogardus 2014).

Not all further clarifications of a safety condition will be suitable for the use of the latter in an analysis of knowledge. In particular, if the respect of similarity that is relevant for safety is itself explicated in terms of knowledge, then an analysis of knowledge which made reference to safety would be in this respect circular. This, for instance, is how Timothy Williamson characterizes safety. He writes, in response to a challenge by Alvin Goldman:

In many cases, someone with no idea of what knowledge is would be unable to determine whether safety obtained. Although they could use the principle that safety entails truth to exclude some cases, those are not the interesting ones. Thus Goldman will be disappointed when he asks what the safety account predicts about various examples in which conflicting considerations pull in different directions. One may have to decide whether safety obtains by first deciding whether knowledge obtains, rather than vice versa. (Williamson 2009: 305)

Because safety is understood only in terms of knowledge, safety so understood cannot serve in an analysis of knowledge. Nor is it Williamson’s intent that it should do so; as we will see below, Williamson rejects the project of analyzing knowledge. This is of course consistent with claiming that safety is a necessary condition on knowledge in the straightforward sense that the latter entails the former.

A third approach to modal conditions on knowledge worthy of mention is the requirement that for a subject to know that p , she must rule out all “relevant alternatives” to p . Significant early proponents of this view include Stine 1976, Goldman 1976, and Dretske 1981. The idea behind this approach to knowledge is that for a subject to know that p , she must be able to “rule out” competing hypotheses to p —but that only some subset of all not- p possibilities are “relevant” for knowledge attributions. Consider for example, the differences between the several models that have been produced of Apple’s iPhone. To be able to know by sight that a particular phone is the 6S model, it is natural to suppose that one must be able to tell the difference between the iPhone 6S and the iPhone 7; the possibility that the phone in question is a newer model is a relevant alternative. But perhaps there are other possibilities in which the belief that there is an iPhone 6S is false that do not need to be ruled out—perhaps, for instance, the possibility that the phone is not an iPhone, but a knock-off, needn’t be considered. Likewise for the possibility that there is no phone at all, the phone-like appearances being the product of a Cartesian demon’s machinations. Notice that in these cases and many of the others that motivate the relevant-alternatives approach to knowledge, there is an intuitive sense in which the relevant alternatives tend to be more similar to actuality than irrelevant ones. As such, the relevant alternatives theory and safety-theoretic approaches are very similar, both in verdict and in spirit. As in the case of a safety theorist, the relevant alternatives theorist faces a challenge in attempting to articulate what determines which possibilities are relevant in a given situation. [ 21 ]

6. Doing Without Justification?

As we have seen, one motivation for including a justification condition in an analysis of knowledge was to prevent lucky guesses from counting as knowledge. However, the Gettier problem shows that including a justification condition does not rule out all epistemically problematic instances of luck. Consequently, some epistemologists have suggested that positing a justification condition on knowledge was a false move; perhaps it is some other condition that ought to be included along with truth and belief as components of knowledge. This kind of strategy was advanced by a number of authors from the late 1960s to the early 1980s, although there has been relatively little discussion of it since. [ 22 ] Kornblith 2008 provides a notable exception.

One candidate property for such a state is reliability . Part of what is problematic about lucky guesses is precisely that they are so lucky: such guesses are formed in a way such that it is unlikely that they should turn out true. According to a certain form of knowledge reliabilism, it is unreliability, not lack of justification, which prevents such beliefs from amounting to knowledge. Reliabilist theories of knowledge incorporate this idea into a reliability condition on knowledge. [ 23 ] Here is an example of such a view:

Simple K-Reliabilism:

S knows that p iff

• S ’s belief that p was produced by a reliable cognitive process.

Simple K-Reliabilism replaces the justification clause in the traditional tripartite theory with a reliability clause. As we have seen, reliabilists about justification think that justification for a belief consists in a genesis in a reliable cognitive process. Given this view, Simple K-Reliabilism and the JTB theory are equivalent. However, the present proposal is silent on justification. Goldman 1979 is the seminal defense of reliabilism about justification; reliabilism is extended to knowledge in Goldman 1986. See Goldman 2011 for a survey of reliabilism in general.

In the following passage, Fred Dretske articulates how an approach like K-reliabilism might be motivated:

Those who think knowledge requires something other than , or at least more than , reliably produced true belief, something (usually) in the way of justification for the belief that one’s reliably produced beliefs are being reliably produced, have, it seems to me, an obligation to say what benefits this justification is supposed to confer…. Who needs it, and why? If an animal inherits a perfectly reliable belief-generating mechanism, and it also inherits a disposition, everything being equal, to act on the basis of the beliefs so generated, what additional benefits are conferred by a justification that the beliefs are being produced in some reliable way? If there are no additional benefits, what good is this justification? Why should we insist that no one can have knowledge without it? (Dretske 1989: 95)

According to Dretske, reliable cognitive processes convey information, and thus endow not only humans, but (nonhuman) animals as well, with knowledge. He writes:

I wanted a characterization that would at least allow for the possibility that animals (a frog, rat, ape, or my dog) could know things without my having to suppose them capable of the more sophisticated intellectual operations involved in traditional analyses of knowledge. (Dretske 1985: 177)

It does seem odd to think of frogs, rats, or dogs as having justified or unjustified beliefs. Yet attributing knowledge to animals is certainly in accord with our ordinary practice of using the word “knowledge”. So if, with Dretske, we want an account of knowledge that includes animals among the knowing subjects, we might want to abandon the traditional JTB account in favor of something like K-reliabilism.

Another move in a similar spirit to K-Reliabilism replaces the justification clause in the JTB theory with a condition requiring a causal connection between the belief and the fact believed; [ 24 ] this is the approach of Goldman (1967, 1976). [ 25 ] Goldman’s own causal theory is a sophisticated one; we will not engage with its details here. See Goldman’s papers. Instead, consider a simplified causal theory of knowledge, which illustrates the main motivation behind causal theories.

Simple Causal Theory of Knowledge:

• S ’s belief that p is caused by the fact that p .

Do approaches like Simple K-Reliabilism or the Simple Causal Theory fare any better than the JTB theory with respect to Gettier cases? Although some proponents have suggested they do—see e.g., Dretske 1985: 179; Plantinga 1993: 48—many of the standard counterexamples to the JTB theory appear to refute these views as well. Consider again the case of the barn facades. Henry sees a real barn, and that’s why he believes there is a barn nearby. This belief is formed by perceptual processes, which are by-and-large reliable: only rarely do they lead him into false beliefs. So it looks like the case meets the conditions of Simple K-Reliabilism just as much as it does those of the JTB theory. It is also a counterexample to the causal theory, since the real barn Henry perceives is causally responsible for his belief. There is reason to doubt, therefore, that shifting from justification to a condition like reliability will escape the Gettier problem. [ 26 ] Gettier cases seem to pose as much of a problem for K-reliabilism and causal theories as for the JTB account. Neither theory, unless amended with a clever “degettiering” clause, succeeds in stating sufficient conditions for knowledge. [ 27 ]

Gettier’s paper launched a flurry of philosophical activity by epistemologists attempting to revise the JTB theory, usually by adding one or more conditions, to close the gap between knowledge and justified true belief. We have seen already how several of these attempts failed. When intuitive counterexamples were proposed to each theory, epistemologists often responded by amending their theories, complicating the existing conditions or adding new ones. Much of this dialectic is chronicled thoroughly by Shope 1983, to which the interested reader is directed.

After some decades of such iterations, some epistemologists began to doubt that progress was being made. In her 1994 paper, “The Inescapability of Gettier Problems”, Linda Zagzebski suggested that no analysis sufficiently similar to the JTB analysis could ever avoid the problems highlighted by Gettier’s cases. More precisely, Zagzebski argued, any analysans of the form JTB+ X , where X is a condition or list of conditions logically independent from justification, truth, and belief, would be susceptible to Gettier-style counterexamples. She offered what was in effect a recipe for constructing Gettier cases:

• (1) Start with an example of a case where a subject has a justified false belief that also meets condition X .
• (2) Modify the case so that the belief is true merely by luck.

Zagzebski suggests that the resultant case will always represent an intuitive lack of knowledge. So any non-redundant addition to the JTB theory will leave the Gettier problem unsolved. [ 28 ] We may illustrate the application of the recipe using one of Zagzebski’s own examples, refuting Alvin Plantinga’s (1996) attempt to solve the Gettier problem by appending to the JTB analysis a condition requiring that the subject’s faculties be working properly in an appropriate environment.

In step one of Zagzebski’s procedure, we imagine a case in which a subject’s faculties are working properly in an appropriate environment, but the ensuing belief, though justified, is false. Zagzebski invites us to imagine that Mary has very good eyesight—good enough for her cognitive faculties typically to yield knowledge that her husband is sitting in the living room. Such faculties, even when working properly in suitable environments, however, are not infallible—if they were, the condition would not be independent from truth—so we can imagine a case in which they go wrong. Perhaps this is an unusual instance in which Mary’s husband’s brother, who looks a lot like the husband, is in the living room, and Mary concludes, on the basis of the proper function of her visual capacity, that her husband is in the living room. This belief, since false, is certainly not knowledge.

In step two, we imagine Mary’s misidentification of the occupant of the living room as before, but add to the case that the husband is, by luck, also in the living room. Now Mary’s belief is true, but intuitively, it is no more an instance of knowledge than the false belief in the first step was.

Since the recipe is a general one, it appears to be applicable to any condition one might add to the JTB theory, so long as it does not itself entail truth. The argument generalizes against all “non-redundant” JTB+ X analyses.

One potential response to Zagzebski’s argument, and the failure of the Gettier project more generally, would be to conclude that knowledge is unanalyzable. Although it would represent a significant departure from much analytic epistemology of the late twentieth century, it is not clear that this is ultimately a particularly radical suggestion. Few concepts of interest have proved susceptible to traditional analysis (Fodor 1998). One prominent approach to knowledge in this vein is discussed in §11 below.

Another possible line is the one mentioned in §2 —to strengthen the justification condition to rule out Gettier cases as justified. In order for this strategy to prevent Zagzebski’s recipe from working, one would need to posit a justification condition that precludes the possibility of step one above—the only obvious way to do this is for justification to entail truth. If it does, then it will of course be impossible to start with a case that has justified false belief. This kind of approach is not at all mainstream, but it does have its defenders—see e.g., Sturgeon 1993 and Merricks 1995. Sutton 2007 and Littlejohn 2012 defend factive approaches to justification on other grounds.

A third avenue of response would be to consider potential analyses of knowledge that are not of the nonredundant form JTB+ X . Indeed, we have already seen some such attempts, albeit unsuccessful ones. For instance, the causal theory of knowledge includes a clause requiring that the belief that p be caused by the fact that p . This condition entails both belief and truth, and so is not susceptible to Zagzebski’s recipe. (As we’ve seen, it falls to Gettier-style cases on other grounds.) One family of strategies along these lines would build into an analysis of knowledge a prohibition on epistemic luck directly; let us consider this sort of move in more detail.

If the problem illustrated by Gettier cases is that JTB and JTB+ analyses are compatible with a degree of epistemic luck that is inconsistent with knowledge, a natural idea is to amend one’s analysis of knowledge by including an explicit “anti-luck” condition. Zagzebski herself outlines this option in her 1994 (p. 72). Unger 1968 gives an early analysis of this kind. For example:

• S ’s belief is not true merely by luck.

The first thing to note about this analysis is that it is “redundant” in the sense described in the previous section; the fourth condition entails the first two. [ 29 ] So its surface form notwithstanding, it actually represents a significant departure from the JTB+ analyses. Rather than composing knowledge from various independent components, this analysis demands instead that the epistemic states are related to one another in substantive ways.

The anti-luck condition, like the safety condition of the previous section, is vague as stated. For one thing, whether a belief is true by luck comes in degrees—just how much luck does it take to be inconsistent with knowledge? Furthermore, it seems, independently of questions about degrees of luck, we must distinguish between different kinds of luck. Not all epistemic luck is incompatible with having knowledge. Suppose someone enters a raffle and wins an encyclopedia, then reads various of its entries, correcting many of their previous misapprehensions. There is a straightforward sense in which the resultant beliefs are true only by luck—for our subject was very lucky to have won that raffle—but this is not the sort of luck, intuitively, that interferes with the possession of knowledge. [ 30 ] Furthermore, there is a sense in which our ordinary perceptual beliefs are true by luck, since it is possible for us to have been the victim of a Cartesian demon and so we are, in some sense, lucky not to be. But unless we are to capitulate to radical skepticism, it seems that this sort of luck, too, ought to be considered compatible with knowledge. [ 31 ]

Like the safety condition, then, a luck condition ends up being difficult to apply in some cases. We might try to clarify the luck condition as involving a distinctive notion of epistemic luck—but unless we were able to explicate that notion—in effect, to distinguish between the two kinds of luck mentioned above—without recourse to knowledge, it is not clear that the ensuing analysis of knowledge could be both informative and noncircular.

As our discussion so far makes clear, one standard way of evaluating attempted analyses of knowledge has given a central role to testing it against intuitions against cases. In the late twentieth century, the perceived lack of progress towards an acceptable analysis—including the considerations attributed to Zagzebski in §7 above—led some epistemologists to pursue other methodological strategies. (No doubt, a wider philosophical trend away from “conceptual analysis” more broadly also contributed to this change.) Some of the more recent attempts to analyse knowledge have been motivated in part by broader considerations about the role of knowledge, or of discourse about knowledge.

One important view of this sort is that defended by Edward Craig (1990). Craig’s entry-point into the analysis of knowledge was not intuitions about cases, but rather a focus on the role that the concept of knowledge plays for humans. In particular, Craig suggested that the point of using the category of knowledge was for people to flag reliable informants—to help people know whom to trust in matters epistemic. Craig defends an account of knowledge that is designed to fill this role, even though it is susceptible to intuitive counterexamples. The plausibility of such accounts, with a less intuitive extension but with a different kind of theoretical justification, is a matter of controversy.

Another view worth mentioning in this context is that of Hilary Kornblith (2002), which has it that knowledge is a natural kind, to be analysed the same way other scientific kinds are. Intuition has a role to play in identifying paradigms, but generalizing from there is an empirical, scientific matter, and intuitive counterexamples are to be expected.

The “knowledge first” stance is also connected to these methodological issues. See §11 below.

10. Virtue-Theoretic Approaches

The virtue-theoretic approach to knowledge is in some respects similar to the safety and anti-luck approaches. Indeed, Ernest Sosa, one of the most prominent authors of the virtue-theoretic approach, developed it from his previous work on safety. The virtue approach treats knowledge as a particularly successful or valuable form of belief, and explicates what it is to be knowledge in such terms. Like the anti-luck theory, a virtue-theoretic theory leaves behind the JTB+ project of identifying knowledge with a truth-functional combination of independent epistemic properties; knowledge, according to this approach, requires a certain non-logical relationship between belief and truth.

Sosa has often (e.g., Sosa 2007: ch. 2) made use of an analogy of a skilled archer shooting at a target; we may find it instructive as well. Here are two ways in which an archer’s shot might be evaluated:

• Was the shot successful? Did it hit its target?
• Did the shot’s execution manifest the archer’s skill? Was it produced in a way that makes it likely to succeed?

The kind of success at issue in (1), Sosa calls accuracy . The kind of skill discussed in (2), Sosa calls adroitness . A shot is adroit if it is produced skillfully. Adroit shots needn’t be accurate, as not all skilled shots succeed. And accurate shots needn’t be adroit, as some unskilled shots are lucky.

In addition to accuracy and adroitness, Sosa suggests that there is another respect in which a shot may be evaluated, relating the two. This, Sosa calls aptness .

• Did the shot’s success manifest the archer’s skill?

A shot is apt if it is accurate because adroit. Aptness entails, but requires more than, the conjunction of accuracy and adroitness, for a shot might be both successful and skillful without being apt. For example, if a skillful shot is diverted by an unexpected gust of wind, then redirected towards the target by a second lucky gust, its ultimate accuracy does not manifest the skill, but rather reflects the lucky coincidence of the wind.

Sosa suggests that this “AAA” model of evaluation is applicable quite generally for the evaluation of any action or object with a characteristic aim. In particular, it is applicable to belief with respect to its aim at truth:

• A belief is accurate if and only if it is true.
• A belief is adroit if and only if it is produced skillfully. [ 32 ]
• A belief is apt if and only if it is true in a way manifesting, or attributable to, the believer’s skill.

Sosa identifies knowledge with apt belief, so understood. [ 33 ] Knowledge entails both truth (accuracy) and justification (adroitness), on this view, but they are not merely independent components out of which knowledge is truth-functionally composed. It requires that the skill explain the success. This is in some respects similar to the anti-luck condition we have examined above, in that it legislates that the relation between justification and truth be no mere coincidence. However, insofar as Sosa’s “AAA” model is generally applicable in a way going beyond epistemology, there are perhaps better prospects for understanding the relevant notion of aptness in a way independent of understanding knowledge itself than we found for the notion of epistemic luck.

Understanding knowledge as apt belief accommodates Gettier’s traditional counterexamples to the JTB theory rather straightforwardly. When Smith believes that either Jones owns a Ford or Brown is in Barcelona, the accuracy of his belief is not attributable to his inferential skills (which the case does not call into question). Rather, unlucky circumstances (the misleading evidence about Jones’s car) have interfered with his skillful cognitive performance, just as the first diverting gust of wind interfered with the archer’s shot. Compensating for the unlucky interference, a lucky circumstance (Brown’s coincidental presence in Barcelona) renders the belief true after all, similar to the way in which the second gust of wind returns the archer’s arrow back onto the proper path towards the target.

Fake barn cases, by contrast, may be less easily accommodated by Sosa’s AAA approach. When Henry looks at the only real barn in a countryside full of barn facades, he uses a generally reliable perceptual faculty for recognizing barns, and he goes right in this instance. Suppose we say the accuracy of Henry’s belief manifests his competence as a perceiver. If so, we would have to judge that his belief is apt and therefore qualifies as an instance of knowledge. That would be a problematic outcome because the intuition the case is meant to elicit is that Henry does not have knowledge. There are three ways in which an advocate of the AAA approach might respond to this difficulty.

First, AAA advocates might argue that, although Henry has a general competence to recognize barns, he is deprived of this ability in his current environment, precisely because he is in fake barn county. According to a second, subtly different strategy, Henry retains barn-recognition competence, his current location notwithstanding, but, due to the ubiquity of fake barns, his competence does not manifest itself in his belief, since its truth is attributable more to luck than to his skill in recognizing barns. [ 34 ] Third, Sosa’s own response to the problem is to bite the bullet. Judging Henry’s belief to be apt, Sosa accepts the outcome that Henry knows there is a barn before him. He attempts to explain away the counterintuitiveness of this result by emphasizing the lack of a further epistemically valuable state, which he calls “reflective knowledge” (see Sosa 2007: 31–32).

Not every concept is analyzable into more fundamental terms. This is clear both upon reflection on examples—what analysis could be offered of hydrogen , animal , or John F. Kennedy ?—and on grounds of infinite regress. Why should we think that knowledge has an analysis? In recent work, especially his 2000 book Knowledge and Its Limits , Timothy Williamson has argued that the project of analyzing knowledge was a mistake. His reason is not that he thinks that knowledge is an uninteresting state, or that the notion of knowledge is somehow fundamentally confused. On the contrary, Williamson thinks that knowledge is among the most fundamental psychological and epistemological states there are. As such, it is a mistake to analyze knowledge in terms of other, more fundamental epistemic notions, because knowledge itself is, in at least many cases, more fundamental. As Williamson puts it, we should put “knowledge first”. Knowledge might figure into some analyses, but it will do so in the analysans, not in the analysandum. [ 35 ]

There is no very straightforward argument for this conclusion; its case consists largely in the attempted demonstration of the theoretical success of the knowledge first stance. Weighing these benefits against those of more traditional approaches to knowledge is beyond the scope of this article. [ 36 ]

Although Williamson denies that knowledge is susceptible to analysis in the sense at issue in this article, he does think that there are interesting and informative ways to characterize knowledge. For example, Williamson accepts these claims:

• Knowledge is the most general factive mental state.
• S knows that p if and only if S ’s total evidence includes the proposition that p .

Williamson is also careful to emphasize that the rejection of the project of analyzing knowledge in no way suggests that there are not interesting and informative necessary or sufficient conditions on knowledge. The traditional ideas that knowledge entails truth, belief, and justification are all consistent with the knowledge first project. And Williamson (2000: 126) is explicit in endorsement of a safety requirement on knowledge—just not one that serves as part of an analysis.

One point worth recognizing, then, is that one need not engage in the ambitious project of attempting to analyze knowledge in order to have contact with a number of interesting questions about which factors are and are not relevant for whether a subject has knowledge. In the next section, we consider an important contemporary debate about whether pragmatic factors are relevant for knowledge.

Traditional approaches to knowledge have it that knowledge has to do with factors like truth and justification. Whether knowledge requires safety, sensitivity, reliability, or independence from certain kinds of luck has proven controversial. But something that all of these potential conditions on knowledge seem to have in common is that they have some sort of intimate connection with the truth of the relevant belief. Although it is admittedly difficult to make the relevant connection precise, there is an intuitive sense in which every factor we’ve examined as a candidate for being relevant to knowledge has something to do with truth of the would-be knowledgeable beliefs.

In recent years, some epistemologists have argued that focus on such truth-relevant factors leaves something important out of our picture of knowledge. In particular, they have argued that distinctively pragmatic factors are relevant to whether a subject has knowledge. Call this thesis “pragmatic encroachment”: [ 37 ]

Pragmatic Encroachment:

A difference in pragmatic circumstances can constitute a difference in knowledge.

The constitution claim here is important; it is trivial that differences in pragmatic circumstances can cause differences in knowledge. For example, if the question of whether marijuana use is legal in Connecticut is more important to Sandra than it is to Daniel, Sandra is more likely to seek out evidence, and come to knowledge, than Daniel is. This uninteresting claim is not what is at issue. Pragmatic encroachment theorists think that the practical importance itself can make for a change in knowledge, without reliance on such downstream effects as a difference in evidence-gathering activity. Sandra and Daniel might in some sense be in the same epistemic position , where the only difference is that the question is more important to Sandra. This difference, according to pragmatic encroachment, might make it the case that Daniel knows, but Sandra does not. [ 38 ]

Pragmatic encroachment can be motivated by intuitions about cases. Jason Stanley’s 2005 book Knowledge and Practical Interests argues that it is the best explanation for pairs of cases like the following, where the contrasted cases are evidentially alike, but differ pragmatically:

Low Stakes . Hannah and her wife Sarah are driving home on a Friday afternoon. They plan to stop at the bank on the way home to deposit their paychecks. It is not important that they do so, as they have no impending bills. But as they drive past the bank, they notice that the lines inside are very long, as they often are on Friday afternoons. Realizing that it wasn’t very important that their paychecks are deposited right away, Hannah says, “I know the bank will be open tomorrow, since I was there just two weeks ago on Saturday morning. So we can deposit our paychecks tomorrow morning”.

High Stakes . Hannah and her wife Sarah are driving home on a Friday afternoon. They plan to stop at the bank on the way home to deposit their paychecks. Since they have an impending bill coming due, and very little in their account, it is very important that they deposit their paychecks by Saturday. Hannah notes that she was at the bank two weeks before on a Saturday morning, and it was open. But, as Sarah points out, banks do change their hours. Hannah says, “I guess you’re right. I don’t know that the bank will be open tomorrow”. (Stanley 2005: 3–4)

Stanley argues that the moral of cases like these is that in general, the more important the question of whether p , the harder it is to know that p . Other, more broadly theoretical, arguments for pragmatic encroachment have been offered as well. Fantl & McGrath (2009) argue that encroachment follows from fallibilism and plausible principles linking knowledge and action, while Weatherson 2012 argues that the best interpretation of decision theory requires encroachment.

Pragmatic encroachment is not an analysis of knowledge; it is merely the claim that pragmatic factors are relevant for determining whether a subject’s belief constitutes knowledge. Some, but not all, pragmatic encroachment theorists will endorse a necessary biconditional that might be interpreted as an analysis of knowledge. For example, a pragmatic encroachment theorist might claim that:

S knows that p if and only if no epistemic weakness vis-á-vis p prevents S from properly using p as a reason for action.

This connection between knowledge and action is similar to ones endorsed by Fantl & McGrath (2009), but it is stronger than anything they argue for.

Pragmatic encroachment on knowledge is deeply controversial. Patrick Rysiew (2001), Jessica Brown (2006), and Mikkel Gerken (forthcoming) have argued that traditional views about the nature of knowledge are sufficient to account for the data mentioned above. Michael Blome-Tillmann (2009a) argues that it has unacceptably counterintuitive results, like the truth of such claims as S knows that p , but if it were more important, she wouldn’t know , or S knew that p until the question became important . Stanley (2005) offers strategies for accepting such consequences. Other, more theoretical arguments against encroachment have also been advanced; see for example Ichikawa, Jarvis, and Rubin (2012), who argue that pragmatic encroachment is at odds with important tenets of belief-desire psychology.

One final topic standing in need of treatment is contextualism about knowledge attributions, according to which the word “knows” and its cognates are context-sensitive. The relationship between contextualism and the analysis of knowledge is not at all straightforward. Arguably, they have different subject matters (the former a word, and the latter a mental state). Nevertheless, the methodology of theorizing about knowledge may be helpfully informed by semantic considerations about the language in which such theorizing takes place. And if contextualism is correct, then a theorist of knowledge must attend carefully to the potential for ambiguity.

It is uncontroversial that many English words are context-sensitive. The most obvious cases are indexicals, such as “I”, “you”, “here”, and “now” (David Kaplan 1977 gives the standard view of indexicals).

The word “you” refers to a different person, depending on the conversational context in which it is uttered; in particular, it depends on the person one is addressing. Other context-sensitive terms are gradable adjectives like “tall”—how tall something must be to count as “tall” depends on the conversational context—and quantifiers like “everyone”—which people count as part of “everyone” depends on the conversational context. Contextualists about “knows” think that this verb belongs on the list of context-sensitive terms. A consequence of contextualism is that sentences containing “knows” may express distinct propositions, depending on the conversational contexts in which they’re uttered. This feature allows contextualists to offer an effective, though not uncontroversial, response to skepticism. For a more thorough overview of contextualism and its bearing on skepticism, see Rysiew 2011 or Ichikawa forthcoming-b.

Contextualists have modeled this context-sensitivity in various ways. Keith DeRose 2009 has suggested that there is a context-invariant notion of “strength of epistemic position”, and that how strong a position one must be in in order to satisfy “knows” varies from context to context; this is in effect to understand the semantics of knowledge attributions much as we understand that of gradable adjectives. (How much height one must have to satisfy “tall” also varies from context to context.) Cohen 1988 adopts a contextualist treatment of “relevant alternatives” theory, according to which, in skeptical contexts, but not ordinary ones, skeptical possibilities are relevant. This aspect is retained in the view of Lewis 1996, which characterizes a contextualist approach that is more similar to quantifiers and modals. Blome-Tillmann 2009b and Ichikawa forthcoming-a defend and develop the Lewisian view in different ways.

Contextualism and pragmatic encroachment represent different strategies for addressing some of the same “shifty” patterns of intuitive data. (In fact, contextualism was generally developed first; pragmatic encroachment theorists were motivated in part by the attempt to explain some of the patterns contextualists were interested in without contextualism’s semantic commitments.) Although this represents a sense in which they tend to be rival approaches, contextualism and pragmatic encroachment are by no means inconsistent. One could think that “knows” requires the satisfaction of different standards in different contexts, and also think that the subject’s practical situation is relevant for whether a given standard is satisfied.

Like pragmatic encroachment, contextualism is deeply controversial. Critics have argued that it posits an implausible kind of semantic error in ordinary speakers who do not recognize the putative context-sensitivity—see Schiffer 1996 and Greenough & Kindermann forthcoming—and that it is at odds with plausible theoretical principles involving knowledge—see Hawthorne 2003, Williamson 2005, and Worsnip forthcoming. In addition, some of the arguments that are used to undercut the data motivating pragmatic encroachment are also taken to undermine the case for contextualism; see again Rysiew 2001 and Brown 2006.

• Almeder, Robert, 1999, Harmless Naturalism. The Limits of Science and the Nature of Philosophy , Chicago and La Salle: Open Court.
• Alston, William P., 1991, Perceiving God. The Epistemology of Religious Experience , Ithaca: Cornell University Press.
• Armstrong, D.M., 1973, Belief, Truth, and Knowledge , Cambridge: Cambridge University Press.
• Bird, Alexander, 2007, “Justified Judging”, Philosophy and Phenomenological Research , 74(1): 81–110. doi:10.1111/j.1933-1592.2007.00004.x
• Blome-Tillmann, Michael, 2009a, “Contextualism, Subject-Sensitive Invariantism, and the Interaction of ‘Knowledge’-Ascriptions with Modal and Temporal Operators”, Philosophy and Phenomenological Research , 79(2): 315–331. doi:10.1111/j.1933-1592.2009.00280.x
• –––, 2009b, “Knowledge and Presuppositions”, Mind , 118(470): 241–294. doi:10.1093/mind/fzp032
• Boh, Ivan, 1985, “Belief, Justification and Knowledge: Some Late Medieval Epistemic Concerns”, Journal of the Rocky Mountain Medieval and Renaissance Association , 6: 87–103.
• Bogardus, Tomas, 2014, “Knowledge Under Threat”, Philosophy and Phenomenological Research , 88(2): 289–313. doi:10.1111/j.1933-1592.2011.00564.x
• Brown, J., 2006, “Contextualism and Warranted Assertability Manoeuvres”, Philosophical Studies , 130(3): 407–435. doi:10.1007/s11098-004-5747-3
• Chisholm, Roderick, 1977, Theory of Knowledge , 2 nd edition, Englewood Cliffs: Prentice Hall.
• –––, 1989, Theory of Knowledge , 3 rd edition, Englewood Cliffs: Prentice Hall.
• Clark, Michael, 1963, “Knowledge and Grounds. A Comment on Mr. Gettier’s Paper”, Analysis , 24(2): 46–48. doi:10.2307/3327068
• Cohen, Stewart, 1988 “How to be a Fallibilist”, Philosophical Perspectives , Epistemology, 2: 91–123. doi:10.2307/2214070
• Comesaña, Juan, 2005, “Unsafe Knowledge”, Synthese , 146(3): 395–404. doi:10.1007/s11229-004-6213-7
• Conee, Earl and Richard Feldman, 2004, Evidentialism: Essays in Epistemology , Oxford: Oxford University Press. doi:10.1093/0199253722.001.0001
• Craig, Edward, 1990, Knowledge and the State of Nature: An Essay in Conceptual Synthesis , Oxford: Oxford University Press. doi:10.1093/0198238797.001.0001
• DeRose, Keith, 1995, “Solving the Skeptical Problem”, The Philosophical Review , 104(1): 1–52. doi:10.2307/2186011
• –––, 2000, “Ought We to Follow Our Evidence?” Philosophy and Phenomenological Research , 60(3): 697–706. doi:10.2307/2653824
• –––, 2009, The Case for Contextualism , New York: Oxford University Press. doi:10.1093/acprof:oso/9780199564460.001.0001
• Dougherty, Trent, 2011, Evidentialism and its Discontents , New York: Oxford University Press. doi:10.1093/acprof:oso/9780199563500.001.0001
• Dretske, Fred, 1981, Knowledge and the Flow of Information , Cambridge, MA: MIT Press.
• –––, 1985, “Precis of Knowledge and the Flow of Information”, in Hilary Kornblith (ed.), Naturalizing Epistemology , Cambridge, MA: MIT Press: 169–187.
• –––, 1989, “The Need to Know”, in Marjorie Clay and Keith Lehrer (eds.), Knowledge and Skepticism , Boulder: Westview Press: 89–100.
• –––, 2005, “The Case Against Closure”, in Matthias Steup and Ernest Sosa (eds), Contemporary Debates in Epistemology , Malden, MA: Blackwell: 13–25.
• Dreyfus, George B.J., 1997, Recognizing Reality: Dharmakirti’s Philosophy and its Tibetan Interpretations , Albany, NY: SUNY Press.
• Dutant, Julien, 2015, “The Legend of the Justified True Belief Analysis”, Philosophical Perspectives , 29(1): 95–145. doi:10.1111/phpe.12061
• Fantl, Jeremy & Matthew McGrath, 2009, Knowledge in an Uncertain World , New York: Oxford University Press. doi:10.1093/acprof:oso/9780199550623.001.0001
• Feldman, Richard and Earl Conee, 1985, “Evidentialism”, Philosophical Studies , 48(1): 15–34. doi:10.1007/BF00372404
• –––, 2001, “Internalism Defended”, American Philosophical Quarterly , 38(1): 1–18. Reprinted in Conee and Feldman 2004: 53–82. doi:10.1093/0199253722.003.0004
• Fodor, Jerry, 1998, Concepts: Where Cognitive Science Went Wrong , New York: Oxford University Press. doi:10.1093/0198236360.001.0001
• Gerken, Mikkel, forthcoming, On Folk Epistemology , Oxford: Oxford University Press.
• Gettier, Edmund L., 1963, “Is Justified True Belief Knowledge?”, Analysis , 23(6): 121–123. doi:10.2307/3326922
• Gibbons, John, 2013, The Norm of Belief , Oxford University Press. doi:10.1093/acprof:oso/9780199673391.001.0001
• Goldman, Alvin I., 1967, “A Causal Theory of Knowing”, The Journal of Philosophy , 64(12): 357–372. doi:10.2307/2024268
• –––, 1976, “Discrimination and Perceptual Knowledge”, The Journal of Philosophy , 73(20): 771–791. doi:10.2307/2025679
• –––, 1979, “What is Justified Belief?” in Justification and Knowledge , George S. Pappas (ed.), Dordrecht: D. Reidel: 1–25.
• –––, 1986, Epistemology and Cognition , Cambridge: Harvard University Press.
• –––, 1999, “Internalism Exposed”, The Journal of Philosophy , 96(6): 271–93. doi:10.2307/2564679
• –––, 2009a, “Replies to Discussants”, in G. Schurz & M. Werning (eds.), Reliable Knowledge and Social Epistemology: Essays on the Philosophy of Alvin Goldman and Replies by Goldman , Amsterdam: Rodopi: 245–288.
• –––, 2009b, “Internalism, Externalism, and the Architecture of Justification”, The Journal of Philosophy , 106(6): 309–338. doi:10.5840/jphil2009106611
• –––, 2011, “Reliabilism”, The Stanford Encyclopedia of Philosophy (Spring 2011 Edition), Edward N. Zalta (ed.), URL = < https://plato.stanford.edu/archives/spr2011/entries/reliabilism/ >.
• Goldman, Alvin I. and Erik J. Olsson, 2009, “Reliabilism and the Value of Knowledge”, in Adrian Haddock, Alan Millar, and Duncan Pritchard (eds.), Epistemic Value , New York: Oxford University Press, pp. 19–41. doi:10.1093/acprof:oso/9780199231188.001.0001
• Greco, John, 2009, “Knowledge and Success from Ability”, Philosophical Studies , 142(1): 17–26. doi:10.1007/s11098-008-9307-0
• Greenough, Patrick & Dirk Kindermann forthcoming, “The Semantic Error Problem for Epistemic Contextualism”, in Ichikawa forthcoming-b: 305–320.
• Greenough, Patrick & Duncan Pritchard, 2009, Williamson on Knowledge , Oxford: Oxford University Press. doi:10.1093/acprof:oso/9780199287512.001.0001
• Hazlett, Allan, 2010, “The Myth of Factive Verbs”, Philosophy and Phenomenological Research , 80(3): 497–522. doi:10.1111/j.1933-1592.2010.00338.x
• Hawthorne, John, 2002, “Deeply Contingent A Priori Knowledge”, Philosophy and Phenomenological Research , 65(2): 247–69. doi:10.1111/j.1933-1592.2002.tb00201.x
• –––, 2003, Knowledge and Lotteries , New York: Oxford University Press. doi:10.1093/0199269556.001.0001
• Ichikawa, Jonathan J., 2011, “Quantifiers, Knowledge, and Counterfactuals”, Philosophy and Phenomenological Research , 82(2): 287–313. doi:10.1111/j.1933-1592.2010.00427.x
• –––, forthcoming-a, Contextualising Knowledge: Epistemology and Semantics , Oxford: Oxford University Press.
• ––– (ed.), forthcoming-b, Routledge Handbook of Epistemic Contextualism , New York: Taylor & Francis.
• Ichikawa, Jonathan J., Benjamin Jarvis, & Katherine Rubin, 2012, “Pragmatic Encroachment and Belief-Desire Psychology”, Analytic Philosophy , 53(4): 327–42. doi:10.1111/j.2153-960X.2012.00564.x
• Kaplan, David, 1977, “Demonstratives”, paper presented at a symposium on Demonstratives at the March 1977 meetings of the Pacific Division of the American Philosophical Association. Printed in Joseph Almog, John Perry & Howard Wettstein (eds.), 1989, Themes From Kaplan , Oxford: Oxford University Press.
• Kaplan, Mark, 1985, “It’s Not What You Know that Counts”, The Journal of Philosophy , 82(7): 350–63. doi:10.2307/2026524
• Kornblith, Hilary, 2002, Knowledge and its Place in Nature , Oxford: Oxford University Press. doi:10.1093/0199246319.001.0001
• –––, 2008, “Knowledge Needs No Justification”, in Quentin Smith (ed.), Epistemology: New Essays , Oxford: Oxford University Press: 5–23. doi:10.1093/acprof:oso/9780199264933.003.0002
• Kripke, Saul A., 2011, “Nozick on Knowledge”, in Philosophical Troubles: Collected Papers, Volume 1 , Oxford: Oxford University Press, pp. 162–224. doi:10.1093/acprof:oso/9780199730155.003.0007
• Lewis, David K., 1973, Counterfactuals , Cambridge, MA: Harvard University Press.
• –––, 1996, “Elusive Knowledge”, Australasian Journal of Philosophy , 74(4): 549–567. doi:10.1080/00048409612347521
• Littlejohn, Clayton, 2012, Justification and the Truth-Connection , Cambridge: Cambridge University Press.
• Lowy, Catherine, 1978, “Gettier’s Notion of Justification”, Mind , 87(345): 105–108. doi:10.1093/mind/LXXXVII.1.105
• Lyons, Jack C., 2009, Perception and Basic Beliefs: Zombies, Modules and the Problem of the External World , New York: Oxford University Press. doi:10.1093/acprof:oso/9780195373578.001.0001
• McGlynn, Aidan, 2014, Knowledge First? , Basingstoke: Palgrave. doi:10.1057/9781137026460
• McKinnon, Rachel, 2013, “Getting Luck Properly Under Control”, Metaphilosophy , 44(4):496–511. doi:10.1111/meta.12044
• Merricks, Trenton, 1995, “Warrant Entails Truth”, Philosophy and Phenomenological Research , 55(4): 841–855. doi:10.2307/2108335
• Miracchi, Lisa, 2015, “Competence to Know”, Philosophical Studies , 172(1): 29–56. doi:10.1007/s11098-014-0325-9
• Myers-Schulz, Blake & Eric Schwitzgebel, 2013, “Knowing that P without Believing that P ”, Noûs , 47(2): 371–384. doi:10.1111/nous.12022
• Nagel, Jennifer, 2010, “Epistemic Anxiety and Adaptive Invariantism”, Philosophical Perspectives , 24(1): 407–35. doi:10.1111/j.1520-8583.2010.00198.x
• –––, 2013, “Knowledge as a Mental State”, Oxford Studies in Epistemology , 4: 275–310. doi:10.1093/acprof:oso/9780199672707.003.0010
• –––, 2014, Knowledge: A Very Short Introduction , Oxford: Oxford University Press.
• Nozick, Robert, 1981, Philosophical Explanations , Cambridge, MA: Harvard University Press.
• Owens, David, 2000, Reason Without Freedom: Problem of Epistemic Normativity , London: Routledge.
• Phillips, Stephen H. and N.S. Ramanuja Tatacharya (trans.), 2004, Epistemology of Perception: Gaṅgeśa’s “Tattvacintāmaṇi” , New York: American Institute of Buddhist Studies.
• Plantinga, Alvin, 1993, Warrant and Proper Function , Oxford: Oxford University Press. doi:10.1093/0195078640.001.0001
• –––, 1996, “Respondeo”, in Jonathan L. Kvanvig. Warrant in Contemporary Epistemology. Essays in Honor of Plantinga’s Theory of Knowledge , Lanham, MD: Rowman and Littlefield.
• Pollock, John J., 1986, Contemporary Theories of Knowledge , Totowa, NJ: Rowman and Littlefield.
• Pritchard, Duncan, 2005, Epistemic Luck , New York: Oxford University Press. doi:10.1093/019928038X.001.0001
• Pritchard, Duncan, Alan Millar, & Adrian Haddock, 2010, The Nature and Value of Knowledge: Three Investigations , New York: Oxford University Press. doi:10.1093/acprof:oso/9780199586264.001.0001
• Radford, Colin, 1966, “Knowledge—By Examples”, Analysis , 27(1): 1–11. doi:10.2307/3326979
• Rose, David & Jonathan Schaffer, 2013, “Knowledge Entails Dispositional Belief”, Philosophical Studies , 166(supplement 1): 19–50. doi:10.1007/s11098-012-0052-z
• Rysiew, Patrick, 2001, “The Context-Sensitivity of Knowledge Attributions”, Noûs , 35(4): 477–514. doi:10.1111/0029-4624.00349
• –––, 2011, “Epistemic Contextualism”, The Stanford Encyclopedia of Philosophy (Winter 2011 Edition), Edward N. Zalta (ed.), URL = < https://plato.stanford.edu/archives/win2011/entries/contextualism-epistemology/ >
• Schaffer, Jonathan, 2004, “From Contextualism to Contrastivism”, Philosophical Studies , 119(1–2): 73–104. doi:10.1023/B:PHIL.0000029351.56460.8c
• Schiffer, Stephen, 1996, “Contextualist Solutions to Scepticism”, Proceedings of the Aristotelian Society , 96(1): 317–333. doi:10.1093/aristotelian/96.1.317
• Shope, Robert K., 1983, The Analysis of Knowing. A Decade of Research , Princeton: Princeton University Press.
• Sosa, Ernest, 1999, “How to Defeat Opposition to Moore”, Noûs , (Supplement: Philosophical Perspectives, Epistemology), 33(s13): 141–153. doi:10.1111/0029-4624.33.s13.7
• –––, 2007, A Virtue Epistemology: Apt Belief and Reflective Knowledge (Volume I), New York: Oxford University Press. doi:10.1093/acprof:oso/9780199297023.001.0001
• Stanley, Jason, 2005, Knowledge and Practical Interests , New York: Oxford University Press. doi:10.1093/0199288038.001.0001
• –––, 2011, Know How , New York: Oxford University Press. doi:10.1093/acprof:oso/9780199695362.001.0001
• Steglich-Petersen, Asbjørn, 2010, “Luck as an Epistemic Notion”, Synthese , 176 (3): 361–377. doi:10.1007/s11229-009-9569-x
• Stine, G.C., 1976, “Skepticism, Relevant Alternatives, and Deductive Closure”, Philosophical Studies , 29(4): 249–261. doi:10.1007/BF00411885
• Sturgeon, Scott, 1993, “The Gettier Problem”, Analysis , 53(3): 156–164. doi:10.2307/3328464
• Sutton, Jonathan, 2007, Without Justification , Cambridge, MA: The MIT Press.
• Unger, Peter, 1968, “An Analysis of Factual Knowledge”, The Journal of Philosophy , 65(6): 157–70. doi:10.2307/2024203
• Weatherson, Brian, 2012, “Knowledge, Bets, and Interests”, in Knowledge Ascriptions , Jessica Brown & Mikkel Gerken (eds), New York: Oxford University Press, pp. 75–103. doi:10.1093/acprof:oso/9780199693702.003.0004
• Williamson, Timothy, 2000, Knowledge and its Limits , Oxford: Oxford University Press. doi:10.1093/019925656X.001.0001
• –––, 2005, “Knowledge, Context, and the Agent’s Point of View”, in Gerhard Preyer & Georg Peter (eds.), Contextualism in Philosophy: Knowledge, Meaning, and Truth , Oxford: Oxford University Press: 91–114.
• –––, 2009, “Reply to Alvin Goldman”, in Williamson on Knowledge , Greenough & Pritchard 2009: 305–312. doi:10.1093/acprof:oso/9780199287512.003.0017
• Worsnip, Alex, forthcoming, “Contextualism and Knowledge Norms”, in Ichikawa forthcoming-b: chapter 14.
• Wright, Crispin, 2004, “Warrant for Nothing (And Foundations for Free)?” Aristotelian Society Supplementary Volume , 78(1): 167–212. doi:10.1111/j.0309-7013.2004.00121.x
• Zagzebski, Linda, 1994, “The Inescapability of Gettier Problems”, The Philosophical Quarterly , 44(174): 65–73. doi:10.2307/2220147

How to cite this entry . Preview the PDF version of this entry at the Friends of the SEP Society . Look up topics and thinkers related to this entry at the Internet Philosophy Ontology Project (InPhO). Enhanced bibliography for this entry at PhilPapers , with links to its database.

• The Epistemology Research Guide

contextualism, epistemic | epistemic closure | epistemology: naturalism in | epistemology: social | epistemology: virtue | justification, epistemic: coherentist theories of | justification, epistemic: foundationalist theories of | justification, epistemic: internalist vs. externalist conceptions of | skepticism: and content externalism

Acknowledgments

For the 2012 revision, we are grateful to Kurt Sylvan for extremely detailed and constructive comments on multiple drafts of this entry. Thanks also to an anonymous referee for additional helpful suggestions. For the 2017 revision, thanks to Clayton Littlejohn, Jennifer Nagel, and Scott Sturgeon for helpful and constructive feedback and suggestions. Thanks to Ben Bayer, Kenneth Ehrenberg, and Mark Young for drawing our attention to errors in the previous version.

Copyright © 2017 by Jonathan Jenkins Ichikawa < ichikawa @ gmail . com > Matthias Steup

• Accessibility

Support SEP

Mirror sites.

View this site from another server:

• Info about mirror sites

The Stanford Encyclopedia of Philosophy is copyright © 2024 by The Metaphysics Research Lab , Department of Philosophy, Stanford University

Library of Congress Catalog Data: ISSN 1095-5054

Conceptual knowledge representation: A cross-section of current research

• Cognitive Neuropsychology 33(3):1-9

• University of Wisconsin–Madison
• This person is not on ResearchGate, or hasn't claimed this research yet.

Discover the world's research

• 25+ million members
• 160+ million publication pages
• 2.3+ billion citations

• Stephanie Brandl
• Jonas Pilot

• COGNITIVE PSYCHOL
• Andrea Tyler

• PSYCHON B REV

• Tijl Grootswagers
• Blake Segula

• COGN NEUROPSYCHOL

• Satoru Saito
• Matthew A. Lambon Ralph

• Vladimir L Cherkassky

• H.M. Wellman

• Samuel J. Gershman
• Samuel Ritter
• Matthew Botvinick

• Jeffrey R. Binder

• Colin J. Humphries
• Rutvik H. Desai
• Recruit researchers
• Join for free
• Login Email Tip: Most researchers use their institutional email address as their ResearchGate login Password Forgot password? Keep me logged in Log in or Continue with Google Welcome back! Please log in. Email · Hint Tip: Most researchers use their institutional email address as their ResearchGate login Password Forgot password? Keep me logged in Log in or Continue with Google No account? Sign up

Knowledge Representation

• Reference work entry
• pp 1689–1692
• Cite this reference work entry

• Pablo Pirnay-Dummer 2 ,
• Dirk Ifenthaler 2 &
• Norbert M. Seel 2  

1013 Accesses

3 Citations

Internal representation

Knowledge representation is a key concept in cognitive science and psychology. To understand this theoretical term one has to distinguish between “knowledge” and its “representation.” Intelligent behaviors of a system, natural or artificial, are usually explained by referring to the system’s knowledge. In other words: The capability of performing intelligent behavior is associated with the existence of applicable knowledge. By relating intelligence and knowledge, the system’s behavior becomes more or less reconstructible and predictable. The most discussed distinction is between declarative (“knowing that”) and procedural (“knowing how”) knowledge (see Anderson 1983 ). Declarative knowledge is defined as factual knowledge, whereas procedural knowledge is defined as the knowledge of specific functions and procedures to perform a complex process, task, or activity.

Modern cognitive science sees cognition and learning as a complex process with...

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Durable hardcover edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Anderson, J. R. (1983). The architecture of cognition . Cambridge, MA: Harvard University Press.

Google Scholar  

Brachman, R. J., & Levesque, H. L. (Eds.). (1985). Readings in knowledge representation . Los Altos: Kaufmann.

Bruner, J. S. (1964). The course of cognitive growth. The American Psychologist, 19 , 1–16.

Article   Google Scholar  

Davis, R., Shrobe, H., & Szolovits, P. (1993). What is a knowledge representation? Artificial Intelligence Magazine, 14 (1), 17–33.

Ifenthaler, D., Pirnay-Dummer, P., & Seel, N. M. (Eds.). (2010). Computer-based diagnostics and systematic analysis of knowledge . New York: Springer.

Larkin, J. H., & Simon, H. A. (1987). Why a diagram is (sometimes) worth ten thousand words. Cognitive Science, 11 , 65–99.

Markman, A. B. (1998). Knowledge representation . Mahwah: Erlbaum.

Minsky, M. (1975). A framework for representing knowledge. In P. Winston (Ed.), The psychology of computer vision . New York: McGraw-Hill.

Mylopoulos, J., & Levesque, H. J. (1984). An overview of knowledge representation. In M. L. Brodie, J. Mylopoulos, & J. W. Schmidt (Eds.), On conceptual modelling. Perspectives from artificial intelligence, databases, and programming languages (pp. 3–17). New York: Springer.

Pylyshyn, Z. (1984). Computation and cognition: Toward a foundation for cognitive science . Cambridge, MA: MIT Press.

Schank, R. (1975). Conceptual information processing . New York: Elsevier.

Seel, N. M. (1991). Weltwissen und mentale Modelle. Göttingen: Hogrefe [World knowledge and mental models].

Download references

Author information

Authors and affiliations.

Department of Education, University of Freiburg, Rempartstr. 11, 4.OG, 79098, Freiburg, Germany

Dr. Pablo Pirnay-Dummer, Dirk Ifenthaler & Norbert M. Seel ( Faculty of Economics and Behavioral Sciences )

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Pablo Pirnay-Dummer .

Editor information

Editors and affiliations.

Faculty of Economics and Behavioral Sciences, Department of Education, University of Freiburg, 79085, Freiburg, Germany

Norbert M. Seel

Rights and permissions

Reprints and permissions

Copyright information

© 2012 Springer Science+Business Media, LLC

About this entry

Cite this entry.

Pirnay-Dummer, P., Ifenthaler, D., Seel, N.M. (2012). Knowledge Representation. In: Seel, N.M. (eds) Encyclopedia of the Sciences of Learning. Springer, Boston, MA. https://doi.org/10.1007/978-1-4419-1428-6_875

Download citation

DOI : https://doi.org/10.1007/978-1-4419-1428-6_875

Publisher Name : Springer, Boston, MA

Print ISBN : 978-1-4419-1427-9

Online ISBN : 978-1-4419-1428-6

eBook Packages : Humanities, Social Sciences and Law Reference Module Humanities and Social Sciences Reference Module Education

Share this entry

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research

• AI Community
• L&D On-Demand

Conceptual Knowledge: Explained | Learnexus

Lauren Goff

Conceptual knowledge is a critical component of the learning and development (L&D) process. It refers to the understanding and comprehension of ideas, theories, principles, and concepts that are abstract and not necessarily tied to physical objects or real-world occurrences. This type of knowledge is essential in various fields, including education, business, science, and technology , where understanding complex concepts is crucial for problem-solving, decision-making, and innovation .

Conceptual knowledge contrasts with procedural knowledge , which involves knowing how to perform specific tasks or activities. While procedural knowledge is about the ‘how’, conceptual knowledge is about the ‘why’. It provides a deeper understanding of why certain procedures work, allowing individuals to apply this knowledge in different contexts and situations. This article will delve into the intricacies of conceptual knowledge, its importance in L&D, and how it can be effectively developed and utilized.

Understanding Conceptual Knowledge

Conceptual knowledge is about understanding the underlying principles and theories that govern various phenomena. It involves recognizing the relationships between different concepts and being able to apply this understanding in different contexts. This type of knowledge is abstract and general, allowing it to be applied across various fields and disciplines.

For instance, understanding the concept of gravity allows us to predict how objects will behave when dropped, regardless of where we are or what the object is. This is an example of conceptual knowledge. It’s not tied to a specific instance or procedure but provides a general understanding that can be applied in various situations.

Characteristics of Conceptual Knowledge

Conceptual knowledge has several key characteristics that distinguish it from other types of knowledge. First, it is abstract, meaning it is not tied to specific instances or examples. Instead, it provides a general understanding that can be applied in various contexts. Second, it is deep, providing a comprehensive understanding of a subject or field. This depth of understanding allows individuals to make connections between different concepts, facilitating problem-solving and innovation .

Third, conceptual knowledge is flexible . Because it provides a general understanding, it can be applied in different contexts and situations. This flexibility is crucial in today’s rapidly changing world, where individuals often need to adapt their knowledge and skills to new challenges and opportunities. Lastly, conceptual knowledge is integrative. It involves connecting different pieces of information and ideas to form a coherent whole. This integrative aspect of conceptual knowledge is essential for critical thinking and decision-making.

Conceptual Knowledge vs. Procedural Knowledge

Conceptual knowledge is often contrasted with procedural knowledge. While conceptual knowledge involves understanding the ‘why’, procedural knowledge is about the ‘how’. Procedural knowledge involves knowing how to perform specific tasks or activities. For example, knowing how to ride a bike or bake a cake involves procedural knowledge.

The #1 place for Learning Leaders to learn from each other.

Get the data & knowledge you need to succeed in the era of AI. We're an invite-only community for L&D leaders to learn from each other through expert-led roundtables, our active forum, and data-driven resources.

While both types of knowledge are important, they serve different purposes. Procedural knowledge is essential for performing specific tasks and activities, while conceptual knowledge provides a deeper understanding that allows individuals to apply their knowledge in different contexts and situations. In the context of L&D, both types of knowledge are crucial. Procedural knowledge allows individuals to perform their jobs effectively, while conceptual knowledge facilitates problem-solving , decision-making, and innovation.

The Importance of Conceptual Knowledge in L&D

Conceptual knowledge plays a crucial role in L&D. It provides the foundation for understanding complex ideas and theories, facilitating problem-solving and decision-making . Moreover, it allows individuals to apply their knowledge in different contexts, enhancing their flexibility and adaptability .

In the context of L&D, conceptual knowledge can be seen as the ‘building blocks’ of learning. It provides the foundation upon which other types of knowledge and skills are built. For instance, understanding the concept of marketing is crucial for learning specific marketing strategies and techniques. Without a solid understanding of the underlying concept, it would be difficult to effectively apply these strategies and techniques.

Facilitating Problem-Solving and Decision-Making

One of the key benefits of conceptual knowledge is that it facilitates problem-solving and decision-making . By understanding the underlying principles and theories, individuals can make informed decisions and solve complex problems. For instance, understanding the concept of supply and demand can help business professionals make informed decisions about pricing and production.

Moreover, conceptual knowledge allows individuals to apply their knowledge in different contexts. For instance, a person who understands the concept of supply and demand can apply this knowledge in various business contexts, such as pricing, production, and marketing. This flexibility is crucial in today’s rapidly changing business environment, where professionals often need to adapt their knowledge and skills to new challenges and opportunities.

Enhancing Learning Transfer

Conceptual knowledge also enhances learning transfer, which refers to the ability to apply knowledge and skills learned in one context to another context. By understanding the underlying principles and theories, individuals can apply their knowledge in various situations and contexts. This ability to transfer learning is crucial for adaptability and lifelong learning.

For instance, a person who understands the concept of conflict resolution can apply this knowledge in various contexts, such as personal relationships, workplace conflicts, and international diplomacy. This ability to transfer learning enhances the individual’s adaptability and versatility, making them more valuable in the workplace and society.

Developing Conceptual Knowledge

Developing conceptual knowledge involves more than just memorizing facts and information. It requires a deep understanding of the subject matter, including the ability to recognize patterns, make connections, and apply knowledge in various contexts. This section will explore some strategies for developing conceptual knowledge.

It’s important to note that developing conceptual knowledge is a process that takes time and effort. It involves active engagement with the material, including critical thinking, problem-solving, and reflection. Moreover, it requires a supportive learning environment that encourages exploration, questioning, and experimentation.

Active Learning

Active learning is a key strategy for developing conceptual knowledge. It involves engaging with the material in a meaningful way, such as through discussion, problem-solving, and application. Active learning encourages learners to take an active role in their learning, rather than passively receiving information.

For instance, instead of simply reading about a concept, learners might discuss it with peers, apply it to real-world situations, or use it to solve problems. This active engagement with the material helps to deepen understanding and facilitate the integration of new knowledge with existing knowledge.

Concept Mapping

Concept mapping is another effective strategy for developing conceptual knowledge. It involves creating a visual representation of the relationships between different concepts. This visual representation can help learners understand the connections between different ideas, facilitating their understanding and retention of the material.

Concept maps can be created using various tools and techniques, including pen and paper, digital tools, and software programs. The key is to create a visual representation that clearly shows the relationships between different concepts, helping learners to see the ‘big picture’ and understand how different ideas fit together.

Reflective Practice

Reflective practice is a process of self-examination and self-evaluation that can help learners deepen their understanding of a subject or field. It involves reflecting on one’s experiences, thoughts, and actions, and using this reflection to gain insights and improve future performance.

Reflective practice can be facilitated through various activities, such as journaling, discussion, and feedback. These activities encourage learners to think critically about their experiences and actions, helping them to make connections between theory and practice and deepen their understanding of the subject matter.

Conceptual knowledge is a critical component of the L&D process. It provides the foundation for understanding complex ideas and theories, facilitating problem-solving and decision-making . Moreover, it enhances learning transfer, allowing individuals to apply their knowledge in different contexts. Developing conceptual knowledge requires active learning, concept mapping, and reflective practice, among other strategies.

By understanding and developing conceptual knowledge, individuals can enhance their problem-solving abilities, make informed decisions , and adapt to new challenges and opportunities. Moreover, organizations can enhance their L&D efforts, fostering a culture of continuous learning and innovation. In the rapidly changing world of today, conceptual knowledge is more important than ever.

Find the best expert for your next Project!

Related blog posts.

What to Know About Augmented Reality for eLearning

Augmented reality (AR) is transforming the landscape of eLearning, offering unique and interactive ways to engage with educational content. ...

Find the Perfect Trainer & Facilitator for Agriculture Training: A Guide for Hiring Managers

Discover the power of collaboration and teamwork in agriculture training. This article reveals winning strategies like group discussions, ca...

BlackBoard Trainers on Learnexus

Discover the benefits of using BlackBoard trainers on Learnexus to enhance your online learning experience....

Join the #1 place for Learning Leaders to learn from each other. Get the data & knowledge you need to succeed in the era of AI.

Join as a client or expert.

We’ll help you get started

Hire an Expert I’m a client, hiring for a project

Find a job i’m a training expert, looking for work, get your free content.

Enter your info below and join us in making learning the ultimate priority 🚀

Join as an employer or L&D expert

I'm an employer, hiring for a project

I'm an L&D expert, looking for work

Already have an account? Login

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

The PMC website is updating on October 15, 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Yearb Med Inform
• v.28(1); 2019 Aug

The Interplay of Knowledge Representation with Various Fields of Artificial Intelligence in Medicine

Laszlo balkanyi.

1 Knowledge Manager, European Centre of Disease Prevention and Control (retired)

Ronald Cornet

2 Associate Professor, Department of Medical Informatics, Academic Medical Center - University of Amsterdam, Amsterdam Public Health research institute, Amsterdam, The Netherlands

Introduction : Artificial intelligence (AI) is widespread in many areas, including medicine. However, it is unclear what exactly AI encompasses. This paper aims to provide an improved understanding of medical AI and its constituent fields, and their interplay with knowledge representation (KR).

Methods : We followed a Wittgensteinian approach (“meaning by usage”) applied to content metadata labels, using the Medical Subject Headings (MeSH) thesaurus to classify the field. To understand and characterize medical AI and the role of KR, we analyzed: (1) the proportion of papers in MEDLINE related to KR and various AI fields; (2) the interplay among KR and AI fields and overlaps among the AI fields; (3) interconnectedness of fields; and (4) phrase frequency and collocation based on a corpus of abstracts.

Results : Data from over eighty thousand papers showed a steep, six-fold surge in the last 30 years. This growth happened in an escalating and cascading way. A corpus of 246,308 total words containing 21,842 unique words showed several hundred occurrences of notions such as robotics, fuzzy logic, neural networks, machine learning and expert systems in the phrase frequency analysis. Collocation analysis shows that fuzzy logic seems to be the most often collocated notion. Neural networks and machine learning are also used in the conceptual neighborhood of KR. Robotics is more isolated.

Conclusions : Authors note an escalation of published AI studies in medicine. Knowledge representation is one of the smaller areas, but also the most interconnected, and provides a common cognitive layer for other areas.

1 Introduction

Artificial intelligence (AI) is becoming increasingly important and its impact is manifold - at least potentially -, but its exact scope is unclear. In this paper we aim to increase understanding of the structure of medical AI as a field of applied science 1 , by investigating the interaction of its constituent fields. The interplay among various fields is studied specifically from the point of view of knowledge representation 1 . Our first objective is to shed light on what exactly AI encompasses, as seen in the medical research literature. The second objective is to analyze how the notion of knowledge representation (KR ) interacts with various fields of AI , i.e., how KR contributes to other fields of AI, and how these contribute to KR. The analysis of the relationships of these fields helps to understand the trends.

2 Background

When addressing AI from a knowledge-representation perspective, an obvious first task is to assess what exactly AI encompasses. Literature does not provide a widely accepted classification or a structural model of (medical) AI and its constituents. Many definitions and descriptions of AI exist, well summarized for example by 2 , but authors think these might not add much to the concept of medical AI for an (already) interested reader. Similarly, there is no single authoritative reference classification of the constituent fields of AI. Library science tools, including catalogue classification systems like DDC (Dewey Decimal Classification), UDC (Universal Decimal Classification), LCC (The Library of Congress Classification) 3 , are of no avail as they don’t provide subcategories. Within the medical and health domain of AI, the hierarchy of Medical Subject Headings (MeSH) provides a good, pragmatic classification of medical AI-related research papers 4 , which is shown in Figure 1 . The definitions of MeSH terms are given in Table 1 .

Full MeSH hierarchy of ‘artificial intelligence’ - MeSH version 2018, October. The MeSH hierarchy levels denoted in italics are used in this paper to investigate the relations of AI fields with KR.

3 Materials and Methods

To achieve our goals we use descriptive metadata, i.e. keywords assigned by authors of papers published in this field and the Medical Subject Headings (MeSH) indexing terms assigned by MEDLINE indexers. We used MEDLINE to retrieve papers, as it consistently specifies MeSH headings. Our method followed these steps:

• Analysis of the proportion of papers in MEDLINE characterized by relevant content metadata for various fields of medical AI. We used PubMed-by-Year 5 to investigate publication frequencies of various medical AI fields over time. This tool is used to visualize the relative proportion of cited publications, tagged by the relevant MeSH index terms. It compares the results for each year to the database as a whole. By entering multiple searches, we may have the results displayed in parallel.
• Determining the interplay between KR and the respective fields of AI by checking co-occurrence of content metadata, as well as detecting interplay among various AI fields themselves. PubVenn 6 was used in this step, a tool that enables PubMed to convert search terms into codified search. As content metadata, i.e., as search terms, we combined our keywords with a series of MeSH terms. PubVenn produces a Venn diagram showing interaction between various AI fields, and provides extraction of the numbers and bibliographic data of citations in the overlapping areas of the Venn diagram.
• Visualization of the interconnectedness, using NodeXL 7 an open-source template for Microsoft® Excel® that makes network graphs.
• Analysis of a limited corpus of medical AI abstracts. A corpus containing the abstracts of the most relevant first thousand papers was established. Relevance was decided according to PubMed “Best Match” ordering. All the abstracts of papers, having MeSH classified “artificial intelligence” keywords, AND the ones, keyworded by authors as “knowledge representation”, were added to the corpus. We used ANTConc 8 to perform phrase frequency and collocation analysis of content metadata labels used as notion labels (words) in the corpus text for better understanding the interplay among fields in general, and between fields and KR.

All search results are based on the numbers extracted from a snapshot of a search performed in October 2018. In order to retrieve papers including knowledge representation and the respective fields in AI, we used a simple search construct: pairs of authors’ keyword ‘ knowledge representation’ and the labels of MeSH index terms pertaining to AI, as shown in Figure 1 . In the same way, we retrieved the MeSH index term pairs, using the same simple search construct e.g., “Biological Ontologies”[All] AND “Natural Language Processing”[All]. As keywords are limited, and may not address all relevant aspects of indexed papers, we exploited text mining to gain insight into the frequencies of phrases that relate to the content descriptive metadata labels. Text mining on the abstracts of the papers followed a Wittgensteinian approach: interpreting the “meaning by usage” - the usage of the content metadata notions as words, referring to AI. 9 . We analyzed occurrences (phrase frequencies and collocation) of content descriptive metadata element labels as words used in the text of papers. We think that this work would provide a deeper understanding of the underlying conceptual structure of the field in research. To this end, a corpus was created consisting of the title, keywords, and abstract of the first 1,000 articles according to PubMed “Best Match” order.

Figure 2 shows the growth of various MeSH-defined AI fields as proportions of MEDLINE-indexed publications. The data regarding “ knowledge representation ” in MEDLINE were collected with the same search query formalism as the queries for those AI fields for which MeSH terms exist. In the last thirty years, research intensified significantly and the growth started in the eighties. The ratio of AI-related research output to all MEDLINE-indexed publications is presently about six times as much as it was at the beginning of the eighties. Some areas like (artificial) neural networks started to grow almost exponentially in the nineties - seemingly levelling out over time, after the year 2000. Other areas like machine learning show very steep growth in the last decade. There is a steady growth in the area of expert systems . The area of knowledge bases (KB) research started to grow with more research on biological ontologies understood by MeSH as a subcategory of KBs. More details and an actualized version with latest data are here: https://goo.gl/j4fvi4

Proportion of MEDLINE-indexed literature on nine areas of AI over time. The envelope curve is not data based but illustrates the cascading growth, the escalation of research.

The changes (and more specifically their relations to knowledge representation ) are further analyzed in this paper by text mining the relevant literature. The results are presented below in two steps.

Step 1: Investigation of the overall interaction among KR and various fields of AI in biomedical literature

Table 2 shows the extracted data. As described in the Methods section, the first level of the MeSH hierarchy classification is used together with ‘ biological ontologies’ - even though ‘ biological ontologies’ falls under the MeSH hierarchy ‘ knowledge bases’ . This is further addressed in the discussion section.

The red and the blue numbers show the two areas ( NeurNet and MachL ) mostly cross-cited with all others. Sums of cross-citations and standard deviations (SD) are calculated from the vertical and horizontal numbers (nine data elements - see as examples the red and blue numbers) for each area. In the case of ‘ heuristics ’, most of the data elements are zero, that is why calculating a standard deviation is not relevant. The standard deviation of these number series indicates how evenly a certain field is connected to others. Obviously, a higher SD means less uniform distribution.

For further visual analysis, overlapping citations among various AI areas are shown as a network diagram in Figure 3 . Nodes represent AI areas by their MeSH designations. Edges represent the overlaps, the cross citations among the nodes. In the depicted network, the nodes are proportionally sized to the number of cited literature areas. The width and the style of the edges correspond to the overlap among them. Widths of edges grow with the magnitude of the overlap. This network visualization helps to see the interconnectedness between the areas and the role of Knowledge Representation in this interdisciplinary arena.

AI fields citations in MEDLINE, viewed as a network, where nodes are the AI fields and edges are the cross-citations.

Step 2: Phrase frequency and collocation analysis of extracted abstracts

The above studied citation data cover over eighty thousand citations. A further, in-depth look a limited corpus containing the abstracts of the most relevant first thousand papers was established. This corpus had 246,308 total words, of which 21,842 are unique word forms. A simple phrase frequency analysis 8 shows that the following five AI fields occur among the most frequent terms in the corpus:

These frequencies show the most researched areas of AI, however they do not shed light to their interaction with the specific aspect of language and meaning, classically discussed as `knowledge representation’. The collocation of AI fields was measured in the same corpus as the phrase frequency , both the left and right window spans were set to the maximum of 20 terms distance. In an earlier paper 10 , authors realized that “... the central role of the term “concept” has been gradually abandoned ….”. The notion of ‘ concept ’ was a term central to what was called the field of research in ‘ knowledge representation’ . Therefore, in order to analyze the current corpus on AI, in addition to the notion of ‘knowledge representation’, the notions ‘language’ & ‘meaning’ were also brought to the collocation study. For the four most studied areas, collocation data found in the corpus are shown in Table 4 .

5 Discussion

Principal findings.

Figure 1 shows that over time the various fields related to medical AI follow a cascading and explicitly escalating evolution. ‘ Expert systems’ studied in the eighties were followed by ‘ computer neural networks’ being in the lead in the nineties and the beginning of the twenty first century. This was followed by even more research focusing on ‘ robotics’ and currently on ‘ machine learning’ . At the same time, research goes on steadily in the other depicted fields. The cascade character might show us how new fields, or new names for old fields, take on and might also incorporate the results of previous areas. However, it is not trivial to see if ‘ machine learning’ will also take on the “cube root” function characteristics of other research fields, levelling out over time. Table 2 and Figure 3 show that although the research in medical AI has branched to a broad spectrum of fields, they are well interconnected. At the same time the interconnectedness varies greatly. ‘ Computer heuristics’ and ‘ biological ontologies’ are somewhat less interconnected to other fields, ‘ machine learning’ and ‘ computer-based neural networks’ are the most interconnected fields with all others. The term “ knowledge representation ” in the MeSH thesaurus itself is not part of an AI field, but is used in three entry terms for AI: Knowledge Representation (Computer) Knowledge Representations (Computer), Representation, Knowledge (Computer). Table 3 shows that the four areas ‘ robotic’ `, ‘ fuzzy logic’ , ‘ neural networks ’, and ‘ machine learning ’ seem to be by far the most mentioned researched areas, while ‘ expert systems ’, although above the limit of 50 citations, scores well below. Table 4 tells us that ‘ fuzzy logic’ seems to be the most collocated notion to the world of ‘ knowledge representation’ , ‘ meaning’, and ‘ language’ . This shows some advantage of the fuzzy approach to represent and to interpret medical knowledge. ‘ Neural networks’ and ‘ machine learning’ are also used in the conceptual neighborhood of knowledge representation. At the same time ‘robotics’ , while an important area in AI, seems to be somewhat isolated from the KR world. These results from text mining show that the various AI fields are well interconnected. It is interesting to see that the lowest standard deviation (SD) of cross citations to different areas occurs for our historically central concept ‘ knowledge representation ’. The relatively lowest SD shows that KR is the most “evenly” referred ‘notion’ till today. This finding provides a quantitative indicator suggesting that studying KR was (and is) at the origin of the wide spreading and branching fields of AI research. We will briefly highlight three interactions.

Interaction between Knowledge Representation and Robotics

Knowledge representation plays a role in robotics, for example for categorizing emotions 11 , learning cognitive robots to count 12 , representing and formalizing knowledge about care 13 . These examples show how knowledge representation can be an integral part of improving the functioning of robots. It apparently is yet too early to exploit the cognitive capacities of robots to contribute to knowledge representation, as no literature was found on this topic.

Interaction between Knowledge Representation and Machine Learning

Interaction between knowledge representation and machine learning is yet limited, but needed. An early acknowledgement of this need, specifically for diagnostic image interpretation, is found in 14 . Already in 1988, it was stated that “Diagnostic image interpretation with learning capability demands a full model of the human expert’s competence, including a considerable variety of knowledge representation schemes and inference strategies, coordinated by a meta-process controller.” A recent approach is to combine graph data (represented in Resource Description Framework and Ontology Web Language) with neural networks to generate embeddings of nodes 15 . This combination results in embeddings that contain both explicit and implicit information. Machine learning can contribute to knowledge representation, e.g., by abstract feature selection, which has been applied for automated phenotyping in 16 . Finally, we notice that natural language processing is among the domains to which machine learning and knowledge representation are applied. For example, MedTAS/P combines these three areas, as described in 17 .

Interaction between Knowledge Representation and Fuzzy Logic

Not surprisingly, most of these overlapping studies focus on the fuzzy nature of our limited knowledge in explaining and understanding particular diseases (e.g. Economou et al. 18 ) in cardiology or in the field of oncology (see D’Aquin et al. 2004 19 ). However, interesting studies compare the “fuzzy” thinking with different approaches, where the “fuzziness” seems to be a connecting notion between the worlds of algorithmic and other approaches interpreting medical data, e.g. Douali et al., in 2014 20 , on fuzzy cognitive maps and Bayesian networks, and Kwiatkowska et al., in 2007 21 , on creating prediction rules using typicality measures. Another typical area for overlapping studies is the high level interpretation of medical knowledge, e.g., Bellamy, in 1997 22 , on “Medical diagnosis, diagnostic spaces, and fuzzy systems” and the work of Boegl et al., in 2004 23 , on knowledge acquisition in a fuzzy knowledge representation framework. Summing up this interaction of these two fields is quite broad and covers many different areas of medical information science.

Limitations

Various widely divergent approaches involving, among others, fuzzy set theory 24 , Bayesian networks 25 , and artificial neural networks 26 27 have been applied to intelligent computing systems in healthcare. Papers concerning AI in the medical domain appear in many literature collections and research events, e.g., events by IEEE - Institute of Electrical and Electronics Engineers, AAAI - Conference on Artificial Intelligence, MLDM - International Conference on Machine Learning and Data Mining, or Intelligent Systems Conference, which may not be indexed in MEDLINE. However, we consider MEDLINE itself as a large enough “sample” of medical AI research to represent the fields and their interplay, so that any limitations of using only MEDLINE will not impact the results.

As mentioned, we found over 80.000 papers that were used in the field interplay analysis. However the more detailed text mining of the corpus had to be limited to the first thousand “best match” papers because of the corpus size limitations of the analytic tools. Having about 250,000 total words and over 20,000 unique word forms, size seems adequate for getting meaningful results for the phrase frequency and the collocation analysis that followed.

For the phrase frequency study, we limited the analysis to phrases occurring at least 50 times. While the tool calculated all phrase frequencies, our opinion is that there has to be a limit in order to judge that a phrase occurs sufficiently frequently in the corpus to demonstrate interest in a research field. While the limit of 50 was chosen in a somewhat arbitrary way, we think there is not much difference among little-mentioned research fields, but there is a clear difference with the leading fields that occur several hundred times. The tables and figures presented in results give some insight in what encompasses AI in the health domain and how the various areas of AI research interact.

Definitions of AI in Related Literature

There is no common agreement on what exactly AI encompasses; thus AI can be considered a “fuzzy” term. In the field of medicine, MeSH provides a good basis for specifying the subdomains of (medical) AI. However, MeSH includes “ knowledge representation ” as an entry term for “ artificial intelligence ”, while “ knowledge bases ” is a subcategory of AI. Outside of the medical domain, attempts to define AI and its field have led to more philosophical answers. Larry Tesler, quoted in 28 , provides a definition that may not be helpful in itself, but does highlight the hype that periodically surrounds AI, stating that “Artificial Intelligence is whatever hasn’t been done yet”. The common aspect of AI is that of computers mimicking intelligent human behavior. Whereas this is sometimes simplified as “thinking machines”, this was demonstrated being an inadequate metaphor by Edsger Dijkstra’s quote “The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.” 29 .

6 Conclusion

The results of our analysis revealed that AI research in medicine occurs in a cascading and escalating way. While neural networks, robotics, and machine learning are the research areas with the largest number of indexed publications, they show the lowest relative interplay with other areas, whereas knowledge representation publications, having one of the smallest numbers of indexed publications, expose the highest interplay of around 45%. This supports the idea that the notion of knowledge representation might play both a historical and foundational role in the various areas, providing a common cognitive layer, a still needed context, even for domains such as machine learning , neural nets , fuzzy logic , and robotics .

1 Authors, chairing the IMIA WG 6, currently called “Language and Meaning in Biomedicine”, formerly “Medical Concept Representation” are continuing the tradition of this WG time to time reaching out for a cross-disciplinary overview with other fields of biomedical information science - in this case with AI. See our WG site for more details: https://imiawg6lamb.wordpress.com/ .