4.1. Knowledge of Temporarily Inaccessible Knowledge

KIK says that it is quite ordinary for people to have trouble recalling something yet know that the (temporarily irretrievable) information is stored in memory. Why think that such cases are common? Begin by reflecting on a familiar experience: that of having a definite feeling that one knows the answer to a question but being unable to call that answer to mind. One feels as if the retrieval of a known fact is “blocked.” William James evocatively describes this feeling as an “intensely active” gap in our consciousness which “beckons” us towards the target information (Reference James1890, p. 251). These experiences occur frequently and with many kinds of target information. Contemporary psychologists have studied this metacognitive feeling extensively, dubbing it the “feeling of knowing” (FOK) (Hart Reference Hart1965). A closely related feeling is the “tip-of-the-tongue” state (TOT) (R. Brown and McNeill Reference Brown and McNeill1966). FOK is usually defined as a judgment that one knows the answer to a question (and sometimes, that one could pick it out of a list of alternatives). TOT is the feeling that information stored in memory will soon be recalled or that it is on the tip of one’s tongue.Footnote ¹³ Both feelings have been documented in relation to retrieval of semantic information (e.g., What is the capital city of South Africa?) as well as retrieval of information from episodic memory (e.g., Where did I see my phone last?). Going forward, I won’t make fine distinctions between FOK and TOT. What is important for our purposes is that both are metacognitive judgments about one’s own epistemic states – in particular, they are judgments about whether certain information is known.

An especially fascinating feature of feelings of knowing is that they are generally reliable indicators of what is stored in memory. In his inaugural studies of FOK, Hart (Reference Hart1965, Reference Hart1966) found that subjects who felt that they did not know the answer to a question and would not be able to identify it out of a list of alternatives did no better than chance on a subsequent multiple choice test, whereas subjects who felt strongly that they knew the answer to a question and would be able to identify it out of a list of alternatives were three times more likely to answer the subsequent multiple choice question correctly. Since Hart’s initial research, FOK judgments have been confirmed to be a reliable predictor of the likelihood of recalling the target in the future, producing it in response to clues, and identifying it among distractors (R. Brown and McNeill Reference Brown and McNeill1966; Freedman and Landauer Reference Freedman and Landauer1966; Schwartz and Metcalfe Reference Schwartz and Metcalfe1992).

FOK is not generated by direct access to the sought-after target. Rather, it is hypothesized that the subject’s familiarity with the cue (Metcalfe et al. Reference Metcalfe, Schwartz and Joaquim1993; Reder Reference Reder1987; Schwartz and Metcalfe Reference Schwartz and Metcalfe1992) and the accessibility of information related to the target (Koriat Reference Koriat1993, Reference Koriat1995; Koriat and Levy-Sadot Reference Koriat and Levy-Sadot2001) mediate FOK. For example, in response to the question “Who wrote The Catcher in the Rye?,” both (i) one’s prior familiarity with the cue – in this case, the title of the novel, and (ii) the relevant information that comes to mind on being asked the question – for example, the nationality of the author, the time period of publication, memories of having read the book, etc. – would determine the presence and intensity of a FOK. Once we understand that this is the mechanism by which FOK operates, we can see why a high FOK can sometimes fail to predict the sought-after information: the accuracy of FOK reports will depend on the reliability of the related information that comes to mind when making the judgment, and such information may not always be reliable.

These findings support the interpretation that subjects who have high FOK possess knowledge that is temporarily inaccessible to them. Of course, FOK and TOT sometimes occur in cases in which the target information is not subsequently recalled. In these cases, there is no way to detect whether the target information is really stored in memory, and so one might be reasonably reluctant to call these instances of knowledge. However, in a majority of cases, high FOK predicts successful retrieval. In these cases, it seems clear that the target information was known, but non-occurrent.

KIK claims not only that temporarily irretrievable information is known, but also that subjects often know that they possess this knowledge.Footnote ¹⁴ In support of this secondary knowledge claim, we can draw a parallel with the operation of our senses. Our senses are a reliable but not infallible guide to the world around us. It is widely acknowledged that our senses are sufficiently reliable to furnish us with knowledge of our surroundings, despite the existence of illusory sensory experiences. Like our senses, the mechanisms that generate FOK and TOT yield metacognitive feelings that are a reliable but not infallible guide to whether target information is stored in memory. Although both our senses and the mechanisms that produce our metacognitive feelings are fallible, and can even be deliberately manipulated to produce illusions (Schwartz Reference Schwartz1998; Whittlesea Reference Whittlesea1993), this need not undermine the claim that both are sufficiently reliable to supply knowledge.Footnote ¹⁵

Returning to Emma’s case, if Emma experiences a strong FOK about the location of her keys, then it will be true of her that (i) she knows her keys are in her tennis bag, (ii) she is experiencing difficulty retrieving from memory the knowledge that her keys are in her tennis bag, and (iii) she knows that she knows where her keys are, despite her difficulties in retrieving this information. Put slightly different: Emma can accurately and justifiably represent herself as knowing the answer to the question Where are my keys? even if she cannot currently recall what that answer is.

Of course, persons in cases like Colleague and Keys will not necessarily experience FOK. They may, sometimes, represent themselves as not knowing the answer to the question that they seek. But given the prevalence of FOK, there will be many cases in which subjects in those kinds of situations do experience FOK. And in many of these cases, they will have both temporarily irretrievable knowledge as well as awareness that they possess this knowledge. Therefore, it is a mistake to conceive of subjects like Claire and Emma as necessarily or characteristically mistaken about whether they possess the knowledge that is the target of their inquiries. Rather – as KIK claims – cases are common in which agents know that they possess knowledge that is temporarily inaccessible to them.

4.2. Permissible Curiosity

Now we come to the second claim, Permissible Curiosity (PC):

Metacognitive feelings such as the FOK and the TOT state are remarkable phenomena: they allow humans to know whether a particular piece of information is stored in memory without actually retrieving that information. But why should we have this ability? The leading view that has emerged among cognitive psychologists is that metacognitive feelings and judgments serve to monitor our own cognitive processes and to control our behavior in ways that facilitate learning (Nelson and Narens Reference Nelson, Narens, Metcalfe and Shimamura1994; Son and Schwartz Reference Son and Schwartz2002). In the previous subsection (4.1), we saw that metacognitive feelings serve to “monitor” what is stored in memory, supplying a person with information about what memory contains. Information that is yielded by this monitoring process is then used to guide a person’s behavior: to determine whether to initiate, continue, or stop some action pertaining to the learning process. This “control” function of metacognitive feelings and judgments was first hypothesized in the late 1970s and early 1980s at a time when there was little direct evidence to support it (A. L. Brown Reference Brown, Weinert and Kluwe1987; Flavell Reference Flavell and Resnick1976; Kluwe Reference Kluwe and Griffin1982; Cf. Son and Schwartz Reference Son and Schwartz2002). Since this initial hypothesis, studies have repeatedly confirmed that metacognitive feelings and judgments exert influence over a person’s learning strategies.

For example, it has been shown that the intensity of FOK predicts whether a subject will initiate a search of memory for the target information (Barnes et al. Reference Barnes, Gopher and Koriat1999), as well as how long subjects will continue a search before terminating it (Young Reference Young2004). Feelings of knowing guide subjects’ choices about which inquiries to prioritize. Maciej Hanczakowski and colleagues (Reference Hanczakowski, Zawadzka and Cockcroft-McKay2014) asked participants to study 60 pairs of words (cue and target). Then, in the test phase portion of the experiment, subjects were shown a cue, and in cases where they did not answer with a target, were asked to provide a FOK judgment and asked whether they would like to restudy that cue-target pair. Participants more frequently chose to restudy items for which they had a high FOK – a result that held both when subjects were instructed to choose for restudy pairs that they believed they would be successful in learning, and when subjects were given no directions about which pairs to choose for restudy. Janet Metcalfe and colleagues obtained similar findings with an experiment designed to measure the relationship between being in a TOT state concerning the answer to a question and desiring to see the answer later (2017).Footnote ¹⁶

The findings discussed so far support the “control” function of metacognitive feelings over subsequent inquiry, since they establish clear relationships between the presence of feelings of knowing and subsequent decisions about which inquiries to pursue, and for how long. But how is curiosity involved? PC concerns curiosity – specifically, it assumes that curiosity plays a role in retrieving information from memory in the relevant cases – so, it is crucial to establish the presence of curiosity in these cases.

As with any mental state with motivational force, there are two main approaches to measuring curiosity: observing subjects’ behaviors and relying on self-reports. Both are common in the psychological literature, and both are used in experiments designed to study the relationship between curiosity, metacognitive feelings, and inquiry.

First, behavioral measures. Since curiosity is, by definition, an intrinsic motivation to engage in question-driven inquiry, it is reasonable to infer the presence of curiosity from certain observed search behaviors. Indeed, in the case of nonhuman animals, we have no other way to ascribe curiosity. It is not uncommon for researchers to assume that the conceptual link between curiosity and information-seeking obviates the need for self-report and justifies an exclusive reliance on behavioral measures to gauge curiosity. For example, in a recent study by Gregory Brooks and colleagues, one of the primary goals of the researchers was to learn more about the role played by curiosity in controlling a subject’s information-seeking decisions (2021, p. 153). In their pursuit of this end, the researchers inferred the presence of curiosity from subjects’ decisions to prioritize certain inquiries over others. Brooks et al. defend the decision not to ask subjects to self-report their own levels of curiosity by appealing to the definition of curiosity and to the established norm of using behavioral measures in studies on state curiosity in children and nonhuman animals (2021, pp. 155, 163).Footnote ¹⁷ The researchers are correct to interpret subjects’ choices to restudy or investigate certain items over others as an indication of higher levels of curiosity about the chosen items. When a person chooses, in the absence of any instructions about how they should make a restudy choice and any incentives that would reward greater accuracy in learning, to restudy information that would allow them to answer one question rather than another, a plausible explanation of this fact is that they are intrinsically motivated to obtain the answer to the former question – in other words, that they are curious.

In addition to using behavioral measures to infer curiosity, we can also try to determine the presence of curiosity through self-report, by asking subjects how curious they are about the answer to some question. In a series of experiments, Jordan Litman and colleagues (Reference Litman, Hutchins and Russon2005) found that people self-report higher levels of curiosity about questions when they have a strong feeling that the answer is on the tip of their tongue. In one experiment, participants were given 12 general knowledge questions ranging over topics in history, literature, and science, each of which could be answered with a single word (e.g., Q: “What is the name of the man who began the reformation in Germany?” A: “Luther”). For each question, a metacognitive judgment was elicited by asking participants whether they knew the answer, did not know it, or whether the answer was on the tip of their tongue. Participants were then instructed to write down the answer (if they had reported having known it) or to report the intensity of their metacognitive feeling by indicating on a 5-point Likert scale how confident they were that they could identify the correct answer on a multiple choice test (if they had reported not having known it, or that it was on the tip of their tongue). Finally, subjects were asked to indicate how curious they were to see the answer to each question on a 4-point scale ranging from “not at all curious” to “very curious.” In a subsequent phase of the same experiment, subjects were given an opportunity to see any of the answers to the general knowledge questions that they wished. Each subject was given a stack of twelve envelopes, sealed with wax. Each envelope had one of the general knowledge questions written on the front and contained an index card with the answer to that question printed on it. Subjects were told that they could open as many of the envelopes as they wished (“feel free to look in some, all, or none”) and that they should open only the envelopes corresponding to the questions for which they genuinely wanted to see the answers. Subjects reported the highest levels of curiosity for questions whose answers were felt to be on the tip of their tongue. In addition, the highest levels of exploratory behavior (opening envelopes) occurred for questions for which subjects had reported being in a TOT state.

Both search behavior and self-reported feelings of curiosity increase when metacognitive feelings of knowing are present. These findings support the view that curiosity accompanies feelings of knowing. Having argued that curiosity is often present when agents are trying to retrieve a known proposition from memory, the last step in a defense of PC is to establish that curiosity and the question-driven inquiries it motivates in such cases are permissible.

Why think that it is permissible to allow one’s curiosity to motivate an inquiry that aims at retrieving a known proposition from memory? In short: because it is helpful. Curiosity’s presence in these situations supplies the motivational force that impels a subject to retrieve desired information from memory, helping to secure the epistemic good of the retrieval of dispositional knowledge from memory. Because curiosity contributes to the epistemic good of knowledge in these cases, curiosity is epistemically appropriate. By way of analogy, consider another motivational force: hunger. Hunger-driven food consumption is, in many situations, helpful in ensuring that a subject receives the good of nourishment. In these contexts, we think of hunger and the food consumption it instigates as entirely appropriate. Similarly, in situations where curiosity is helpful in ensuring that a subject obtains the good of occurrent knowledge, we should think of curiosity and the question-driven inquiry it instigates as entirely appropriate.

While it is difficult to establish causal relationships, a reasonable hypothesis is that metacognitive feelings induce curiosity, which serves as the motivational force behind mental search in the relevant cases. Litman and colleagues write that their findings support the conclusion that feelings of knowing “contribute to the arousal of curiosity states to motivate exploratory behavior” (Reference Litman2005, p. 579). Metcalfe interprets her team’s findings as suggesting that the TOT state “prompts” or “induces” curiosity (Reference Metcalfe, Schwartz and Bloom2017, p. 2), and Brooks and colleagues posit that metacognitive feelings generated by unsuccessful retrieval of information from memory “induce” curiosity, which in turn motivates inquiry (Reference Brooks, Yang and Köhler2021, p. 153). Curiosity is crucial in this process because it supplies the motivational force that gives rise to the question-driven inquiry.Footnote ¹⁸

There is empirical evidence that the process by which metacognitive feelings and the curiosity they induce guides subsequent inquiry really does promote epistemic goods for the subject. For instance, in a subsequent analysis of subjects’ test performance on cue-target word pairs, Hanczakowski’s team found that subjects more effectively learned pairs for which they reported a high FOK and a desire to restudy (2014, pp. 1620–1621). The presence of curiosity in the cases under consideration does not appear to be pathological. Rather, cases in which curiosity drives a person to focus on a particular question or to search a little longer for information stored in memory appear to be an adaptive feature of our experience as fallible human knowers.

The data discussed in this section support the conclusion that the FOK and the curiosity that accompanies it exercise a control function over mental search, helping to determine which inquiries to pursue and sustaining those inquiries until target information is retrieved from memory or learned. In other words, both FOK and curiosity play an important role in effective inquiry in the cases under consideration. But, if curiosity plays a beneficial role in effective inquiry, including instances of mental search in which the target information is known, curiosity in these circumstances seems entirely appropriate – not normatively problematic. Therefore, agents who experience curiosity about some question while trying to retrieve its known answer from memory do so permissibly.

5.1. First objection: Temporarily irretrievable information is not knowledge

The first objection I will consider denies that temporarily irretrievable information is knowledge. This move preserves the Ignorance Norm, because it entails that cases where agents are curious about temporarily irretrievable information stored in memory are not cases that involve being curious about a question whose answer is known.

There are a couple of reasons to reject this suggestion. First, it is ad hoc, drawing an arbitrary division in our concept of knowledge. Why should some information stored in memory have less of a claim to be knowledge than information that is more easily retrievable? Second, and more importantly, denying that information stored in memory is knowledge is at odds with widely accepted views of knowledge and belief. Epistemologists standardly draw a distinction between occurrent and dispositional beliefs (sometimes also called “standing” beliefs). A belief that consists of a representation of the propositional content P is dispositional just in case, in the right circumstances, it can be retrieved from memory for use in planning and reasoning. A belief is occurrent when it is conscious and available for use in planning and reasoning (Goldman Reference Goldman1976, p. 88).Footnote ¹⁹ At any given time, most of one’s beliefs are dispositional; only a small proportion are occurrent. The occurrent/dispositional distinction also applies to knowledge, since knowing P entails believing P. On the standard way of thinking about knowledge and belief, then, information stored in memory but retrievable is dispositional knowledge.

One might object that there is an epistemically relevant difference between easily retrievable propositional content stored in memory and propositional content stored in memory that is temporarily inaccessible, such that the former but not the latter counts as knowledge. However, it is hard to see what rationale could be provided for this distinction. In both cases, propositional content that meets the criteria for knowledge (whatever those may be) is stored in memory. Sometimes, that information comes easily to mind when bidden; at other times, retrieval is more challenging. But so long as the target content comes to mind in the end, it is hard to see why functional difficulties with retrieval would make a difference for the stored information’s status as knowledge.

5.2. Second objection: Failure to explain intuitions about cases

A second objection to the view I’ve defended here is that it fails to explain our intuitive reactions to cases like Idle Curiosity 2, Colleague, and Keys. Part of what was so attractive about the Ignorance Norm is that it could explain intuitions that something is going wrong in these cases. So if, as I’ve argued, the Ignorance Norm is false, we will need another explanation. Why is it that these cases seem problematic to many, if not that they involve subjects who inquire into what they know?

In response, we can begin by noting that Idle Curiosity 2, Colleague, and Keys are all framed in a way that encourages the hearer to think of the subject as being conscious of the relevant propositional content that constitutes an answer to their question. Emma, for example, is described as knowing that her keys are in her tennis bag, rather than being described as knowing the answer to the question Where are my keys? Ben is described as knowing that Daniel Craig starred in Layer Cake and Munich. This framing primes us to think of the content that constitutes an answer to the subjects’ respective questions as occurrent for the subject.

This opens the way for an alternate explanation of the cases: that our intuitions that something is amiss are not due to the fact that we see the subjects as being curious about Q^P while knowing P, but because we see them as being curious about Q^P while P is occurrent for them. This hypothesis can be tested by rewriting the cases in such a way that makes it clear that the requisite knowledge is not occurrent and seeing what intuitions result. If curiosity seems unproblematic in cases where we represent the subjects as failing to be conscious of the content that answers their question, this supports the hypothesis that it is being curious about a question whose answer is occurrent that is problematic, rather than being curious about a question whose answer one knows. Consider the following variant of Keys:

Keys 2

While preparing to leave for the gym, Francine put her keys in her tennis bag. Francine knows that she put her keys in her tennis bag, but now, having finished her preparations and ready to walk out the door, despite feeling that she knows where she put her keys, she’s having trouble recalling their location. She spends 10 minutes feeling intensely curious about where she put them. Eventually, Francine recalls what she’s done and locates her keys.

Keys 2 stipulates that Francine knows where she put her keys but that she is having trouble recalling where she put them, suggesting that her knowledge is dispositional, not occurrent. To my mind, Francine’s curiosity in Keys 2 seems entirely unproblematic. To be sure, Francine’s situation is subpar in some respects. For instance, it’s unfortunate for her that her memory is working so poorly right now. But none of this entails that Francine’s curiosity, given her circumstances, is inappropriate. It is plausible, then, that what is going wrong in cases like Idle Curiosity 2 and Keys is that the subjects are curious about a question whose answer is occurrent for them, and not that they are violating the Ignorance Norm. If this line of reasoning is correct, it suggests an easy modification of the Ignorance Norm:

Non-occurrent Knowledge Norm: One ought not: be curious about Q and occurrently know P^Q.

At first glance, this modification to the Ignorance Norm might seem trivial. If there is a prohibition on being curious about something you occurrently know, then there is still some form of incompatibility between curiosity and knowledge – albeit an incompatibility between curiosity and a certain subset of known propositions.

However, if the Non-occurrent Knowledge Norm is correct, I think the alternative it posits is less superficial than it might first appear. I want to conclude by gesturing toward a couple of areas in which I think that this norm, if true, could help us to deepen our understanding of curiosity and curiosity-driven inquiry. These concluding thoughts will necessarily be sketchy, since I don’t have the space to fully develop them here.

Another thing I do not have space for is a thorough evaluation of the Non-occurrent Knowledge Norm. The norm strikes me as plausible, since it both provides a good explanation of intuitions about what is going wrong in cases where agents are curious about something they (occurrently) know, and it accommodates cases involving subjects who are permissibly curious about what they (dispositionally) know. However, there are also reasons for skepticism. Perhaps, it is psychologically impossible to be curious about what one occurrently knows. If so, this would obviate the need for a norm prohibiting agents from being so. Perhaps, there are counterexamples to the norm. For instance, perhaps it is permissible to be curious about a question whose answer one occurrently knows if one fails to recognize that the known proposition constitutes an answer to one’s question. For these reasons, the Non-occurrent Knowledge Norm deserves further scrutiny. However, since a thorough and systematic assessment of the norm is beyond the scope of this paper, the remainder of my remarks should be taken as conditional. If the Non-occurrent Knowledge Norm is correct, it is more than just a superficial revision of the Ignorance Norm.

One interesting feature of the Non-occurrent Knowledge Norm is that it is a product of recognizing the deep commonalities between inquiries that are external to ourselves and those that are internal. Some curiosity-driven inquiries primarily involve seeking information from sources outside of ourselves. Paradigmatic examples include reading a new biography about a favorite actor or visiting restaurants in a new neighborhood. Other inquiries primarily involve seeking information that lies within ourselves. The internally-directed curiosity I have been focusing on in this paper motivates inquiry that is directed at retrieving information from memory. This is an instance where the aim of an inquiry is retrieving existing knowledge. But there may be other kinds of internal curiosity-driven inquiry whose target is new knowledge: working out the entailments of a statement in propositional logic, for instance, or engaging in self-directed questioning to gain insight into one’s own feelings, beliefs, or values.

The contemporary philosophical literature on inquiry and curiosity has, to the best of my knowledge, focused on externally-directed inquiry. This is understandable: there is plenty to probe, think about, and understand even if one limits one’s focus in this way. But the result is that similarities and differences between externally-directed and internally-directed inquiry have been overlooked. At best, bracketing internally-directed inquiry means that we miss out on some interesting data when theorizing about inquiry and question-directed attitudes. At worst, doing so results in a myopic perspective that leads to false conclusions about the nature of inquiry and the norms that govern it.Footnote ²⁰ Examining internally-directed forms of inquiry may yield new insights about the nature and aims of curiosity and question-driven inquiry.

Recall that the Ignorance Norm and the view that curiosity aims at knowledge are believed by some to go hand in hand: if the point of curiosity is to acquire knowledge, it is inappropriate to continue being curious when I know the answer. However, once we recognize that some curiosity-driven searches are properly directed at knowledge an agent already possesses, the view that curiosity aims at knowledge can seem less plausible. If the Non-occurrent Knowledge Norm is correct, it suggests the alternative view that curiosity and curiosity-driven inquiry aim at occurrent knowledge. Perhaps, curiosity aims at putting an agent in a position not merely to know, but to make use of what is known in deliberation and action. Doing so requires not merely that the agent possess relevant knowledge, but that the knowledge be occurrent. Dispositional knowledge is of no use to an agent in planning and deliberation if it does not rise to the level of consciousness.

Another way of reconciling the view that curiosity aims at knowledge with the reality of agents whose curiosity targets what is known would be to regard instances of mental search as occasions on which the knowledge that is the target of one’s curiosity lies within one’s own mind. While this position preserves the view that the aim of curiosity is knowledge, it also raises further questions about how to interpret this claim in a way that does not yield a general prohibition on inquiring into what is known.

Summing up, I have argued that the Ignorance Norm is false. There are circumstances in which it is permissible to be curious about a question whose answer one knows. I have also suggested that the considerations used to establish the falsity of this norm raise interesting questions about the relationship between internally- and externally-directed inquiries, the significance of occurrent knowledge for curiosity-driven inquiries, and for the claim that curiosity aims at knowledge. Exploring the implications of curiosity-driven mental search for these issues will lead to a clearer view of the nature and purpose of curiosity.