Experimental Philosophy and Epistemology

One of the most richly debated questions in contemporary epistemology is whether or not knowledge attributions are sensitive to conversational context.

Bank Case A:

My wife and I are driving home on a Friday afternoon. We plan to stop at the bank on the way home to deposit our paychecks. But as we drive past the bank, we notice that the lines inside are very long, as they often are on Friday afternoons. Although we generally like to deposit our paychecks as soon as possible, it is not especially important in this case that they be deposited right away, so I suggest that we drive straight home and deposit our paychecks on Saturday morning. My wife says, “Maybe the bank won't be open tomorrow. Lots of banks are closed on Saturdays.” I reply, “No, I know it'll be open. I was just there two weeks ago on Saturday. It's open until noon.”

Bank Case B:

My wife and I drive past the bank on a Friday afternoon, as in Bank Case A, and notice the long lines. I again suggest that we deposit our paychecks on Saturday morning, explaining that I was at the bank on Saturday morning only two weeks ago and discovered that it was open until noon. But in this case, we have just written a very large and very important check. If our paychecks are not deposited into our checking account before Monday morning, the important check we wrote will bounce, leaving us in a very bad situation. And, of course, the bank is not open on Sunday. My wife reminds me of these facts. She then says, “Banks do change their hours. Do you know the bank will be open tomorrow?” Remaining as confident as I was before that the bank will be open then, still, I reply, “Well, no. I'd better go in and make sure.”

On one side of the debate, Contextualists have argued that our intuitive responses to these and similar vignettes show that various features of a conversational context, in particular, what possibilities are made salient, can affect the truth-value of knowledge attributions.¹⁷ On the other side of the debate, subject-sensitive invariantists have argued that conversational context does not affect the truth-value of knowledge attributions and that our intuitive responses to these and similar vignettes can best be explained in other ways, for example, in terms of what is at stake for the person whose epistemic position is described.¹⁸

What has made this debate particularly interesting to experimental philosophers is that both sides of the debate commonly appeal to the very same intuitions about the very same hypothetical cases.

Let's start by considering the view that salience matters. According to many contextualists, whether someone knows something (or, at least, counts as knowing something) depends, at least in part, on what possibilities have been made salient. Contextualists rest their case on our purported intuitions about the two bank cases just considered or other cases having similar structure. In both cases, a husband and wife stop by a bank on their way home from work on a Friday afternoon intending to deposit their paychecks. When they notice that the lines inside are quite long, the husband suggests that he come back the following day to make the deposits, noting to his wife that he knows that the bank will be open on Saturday because he was there two Saturdays ago. The only relevant difference between the two cases is supposed to be that in the second case, unlike the first, the wife explicitly raises the possibility that the husband might be wrong, noting that banks sometimes change their hours.

This argument rests on a testable empirical claim, namely, that our intuitions track what possibilities have been made salient, and a number of experimental studies have recently been conducted in order to test this claim using vignettes designed to ensure that the only difference between vignettes is what possibilities have been made salient. Somewhat surprisingly, the results of some of these studies suggest that salience doesn’t matter: people seem to be just as willing to say that someone knows something when the possibility of being wrong has been made salient as they are to say that someone knows something when that possibility has gone unmentioned.

Consider, for example, the following pair of vignettes (Buckwalter 2010):

Low Standards

Sylvie and Bruno are driving home from work on a Friday afternoon. They plan to stop at the bank to deposit their paychecks, but as they drive past the bank they notice that the lines inside are very long. Although they generally like to deposit their paychecks as soon as possible, it is not especially important in this case that they be deposited right away. Bruno tells Sylvie, “I was just here last week and I know that the bank will be open on Saturday.” Instead, Bruno suggests that they drive straight home and return to deposit their paychecks on Saturday. When they return to the bank on Saturday, it is open for business.

High Standards

Sylvie and Bruno are driving home from work on Friday afternoon. They plan to stop at the bank to deposit their paychecks, but as they drive past the bank they notice that the lines inside are very long.

Unlike Bank Case A and Bank Case B, the only difference between these two cases is that in the High Standards case, unlike the Low Standards case, the possibility that Bruno is wrong has been explicitly raised. If contextualists are right, we should expect that people who are asked to consider the Low Standards case would be more likely to judge that Bruno knows that the bank will be open on Saturday than people who are asked to consider the High Standards case. But, this isn't what happens. Wesley Buckwalter (2010) found no statistically significant difference between intuitive judgments about the two cases: people are just as likely to judge that Bruno knows that the bank will be open on Saturday when they are asked to consider the High Standards case as they are when they are asked to consider the Low Standards case.²⁰

Perhaps salience matters only when stakes matter. Both of Buckwalter's vignettes describe cases in which there is no personal cost associated with being wrong; in both cases, it is not especially important that the paychecks be deposited right away. Maybe the situation would be different were it important that the paychecks be deposited right away. Consider, for example, the following pair of vignettes (May et al. 2010):

High Stakes No Alternative

Hannah and her wife Sarah are driving home on a Friday afternoon.

High Stakes Alternative

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. As they drive past the bank, they notice that the lines inside are very long, as they often are on Friday afternoons. Hannah notes that she was at the bank 2 weeks before on a Saturday morning, and it was open. Sarah points out that banks do change their hours. Hannah says, “I know the bank will be open tomorrow. So we can deposit our paychecks tomorrow morning.”

Unlike Buckwalter's two bank cases, the stakes in both of these cases are quite high. If salience matters more when stakes matter, then we should expect that people who are asked to consider the High Stakes No Alternative case would be more likely to judge that Hannah knows that the bank will be open on Saturday than people who are asked to consider the High Stakes Alternative case. But, again, this expectation is not met. Joshua May, Walter Sinnott- Armstrong, Jay Hull, and Aaron Zimmerman (2010) found no statistically significant difference between intuitive judgments about the two cases.²¹

These studies suggest that contextualists are wrong.

If salience doesn't matter, then what about stakes? According to subject-sensitive invariantists, whether someone knows something depends, at least in part, on facts about the personal costs associated with being wrong. This marks a significant break from more traditional accounts of knowledge, according to which the only facts relevant to whether or not someone knows something are truth- conducive facts, and is part of a growing anti-intellectualist trend in recent epistemology.²³ Like contextualists, subject-sensitive invariantists rest their case on our purported intuitions about pairs of vignettes. Consider, for example, the following two vignettes (Stanley 2005):

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 isn'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 2 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 2 weeks before on 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.”

Assuming that the bank will be open on Saturday and that the truth- conducive facts available to Hannah don’t change from one vignette to the next, then the only relevant difference between the two cases is supposed to be the personal costs associated with being wrong. Since the only relevant differences between the two cases are the personal costs associated with being wrong, and since subject­sensitive invariantists think that our intuitive responses to these vignettes will be that Hannah knows that the bank will be open on Saturday in the Low Stakes case but not in the High Stakes case, they conclude that stakes matter.

Again, the claim that we will intuitively judge that Hannah knows that the bank will be open on Saturday in the Low Stakes case but not in the High Stakes case is a testable empirical claim, and experimental philosophers have recently designed a number of studies aimed at testing this claim, making sure that the only difference between pairs of vignettes was the personal costs associated with being wrong. Interestingly, the results of some of these studies suggest that stakes don’t matter: when truth-conducive facts are fixed across vignettes, people tend to attribute knowledge when the personal costs of being wrong are low and when the personal costs of being wrong are high.

Consider, for example, the following pair of vignettes (Feltz & Zarpentine 2010):

Simplified Low Stakes:

Hannah and her sister 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 do not have an impending bill coming due, it is not 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. Hannah says to Sarah, “I know that the bank will be open tomorrow.”

Simplified High Stakes:

Hannah and her sister 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, 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. Hannah says to Sarah, “I know that the bank will be open tomorrow.”

The only difference between the two cases is the personal cost of being wrong. If stakes matter, then we should expect that people who are asked to consider the Simplified Low Stakes case would be more likely to judge that Hannah knows that the bank will be open on Saturday than people who are asked to consider the Simplified High Stakes case. Interestingly, this doesn't appear to be the case. Adam Feltz and Chris Zarpentine (2010) found no statistically significant difference between intuitive judgments about the two cases.²⁴

In a separate empirical study, Mark Phelan (forthcoming) also found that stakes don't seem to matter. Consider the following pair of vignettes:

Unimportant:

Kate is ambling down the street, out on a walk for no particular reason and with no particular place to go. She comes to an intersection and asks a passerby the name of the street. “Main Street,” the passerby says. Kate looks at her watch, and it reads 11:45 AM. Kate's eyesight is perfectly normal, and she sees her watch clearly. Kate's hearing is perfectly normal, and she hears the passerby quite well. She has no special reason to believe that the passerby is inaccurate. She also has no special reason to believe that her watch is inaccurate. Kate could gather further evidence that she is on Main Street (she could, for instance, find a map), but she doesn't do so, since, on the basis of what the passerby tells her, she already thinks that she is on Main Street.

Important:

Kate needs to get to Main Street by noon: her life depends on it. She comes to an intersection and asks a passerby the name of the street. “Main Street,” the passerby says. Kate looks at her watch, and it reads 11:45 AM. Kate's eyesight is perfectly normal, and she sees her watch clearly. Kate's hearing is perfectly normal, and she hears the passerby quite well. She has no special reason to believe that the passerby is inaccurate. She also has no special reason to believe that her watch is inaccurate. Kate could gather further evidence that she is on Main Street (she could, for instance, find a map), but she doesn't do so, since, on the basis of what the passerby tells her, she already thinks that she is on Main Street.

As before, the only difference between the two cases is the personal cost of being wrong. So, again, if stakes matter, then we should expect that people who are asked to consider the Unimportant case would be more likely to judge that Kate knows that she is on Main Street than people who are asked to consider the Important case. But, again, this doesn't appear to be the case. Phelan found no statistically significant difference between intuitive judgments about the two cases.²⁵

It looks like stakes don't matter after all. When we put these studies together with the experimental work on salience, the results challenge both contextualism and subject-sensitive invariantism.²⁶ People seem to be just as willing to say that someone knows something when the possibility of being wrong has been made salient as they are when that possibility has gone unmentioned, and just as willing to say that someone knows something when the personal costs associated with being wrong are high as they are when those costs are low. In light of this, it might be tempting to reject both contextualism and subject-sensitive invariantism in favor of some form of classical invariantism, according to which the truth conditions of knowledge attributions do not vary once the person, proposition, and truth-conducive facts have been settled. There are growing reasons to worry that this move would be too quick, however. Jonathan Schaffer and Joshua Knobe (2011), for example, have recently argued that experimental studies focused on the role that salience plays in knowledge attribution haven't succeeded in making sufficiently salient the possibility that the bank will be closed on Saturday. And, N. Angel Pinillos (forthcoming) has recently argued that studies focused on the role that stakes play in knowledge attribution haven't made sure that participants are tracking the right details, in particular, the relationship between how much evidence is needed in order to count as knowing something and the personal costs associated with being wrong. Let's look more closely at each of these arguments.

According to Schaffer and Knobe, merely mentioning a possibility does not necessarily make that possibility sufficiently salient, particularly when the possibility mentioned seems strange or improbable. Instead, the possibility must be presented in a particularly concrete and vivid fashion. This raises a concern about previous experimental studies, namely, that the possibility that the bank will be closed on Saturday wasn't made sufficiently salient in these cases, and that, when that possibility is made sufficiently salient, salience will matter. To test this prediction, Schaffer and Knobe had subjects consider the following pair of vignettes:

Revised Low Standard

Hannah and Sarah are driving home on a Friday afternoon. They plan to stop at the bank to deposit their paychecks. As they drive past the bank, they notice that the lines inside are very long, as they often are on Friday afternoons. Hannah says, “I was at the bank two weeks before on a Saturday morning, and it was open. So this is a bank that is open on Saturdays. We can just leave now and deposit our paychecks tomorrow morning.” Sarah replies, “Ok, that sounds good. Let's go on Saturday.”

Concrete High Standard

Hannah and Sarah are driving home on a Friday afternoon. They plan to stop at the bank to deposit their paychecks. As they drive past the bank, they notice that the lines inside are very long, as they often are on Friday afternoons. Hannah says, “I was at the bank two weeks before on a Saturday morning, and it was open. So this is a bank that is open on Saturdays. We can just leave now and deposit our paychecks tomorrow morning.” Sarah replies, “Well, banks do change their hours sometimes. My brother Leon once got into trouble when the bank changed hours on him and closed on Saturday. How frustrating! Just imagine driving here tomorrow and finding the door locked.”

Based on earlier studies that focused on the role that salience plays in knowledge attribution, we should expect no difference: people who are asked to consider the Concrete High Standard case should be no less likely to judge that Hannah knows that the bank will be open on Saturday than people who are asked to consider the Revised Low Standard case. But, this is not what Schaffer and Knobe found. They found that when the possibility that Hannah is wrong about the bank's weekend hours of operation was raised in this more vivid fashion, people who were asked to consider the Revised Low Standard case were, in fact, more likely to judge that Hannah knows that the bank will be open on Saturday than were those people who were asked to consider the Concrete High Standard case.²⁷

Schaffer and Knobe conclude from this that salience matters after all and go on to use these results (together with other empirical studies) to argue in favor of Contrastivism, a version of contextualism according to which claims that someone knows something really amount to claims that someone knows something rather than something else. But, while these results should certainly give pause to anyone who might have been tempted to reject contextualism on the basis of the results obtained by the Buckwalter (2010) and May et al. (2010) studies, the move by Schaffer and Knobe back to contextual-ism, or at least to a version of contextualism, also seems too quick. First, since Schaffer and Knobe tested only Australian subjects, this might only be evidence of cross-cultural diversity: maybe salience does matter... for Australians. (As we shall see in Chapter 4, this isn't an idle concern. The discovery of intuition-based cross-cultural diversity is one of experimental philosophy's most important contributions.) Second, it isn't actually clear that the possibility of being wrong is any more salient in the Concrete High Standards case than it is, for example, in the High Standards case. In both cases, the protagonist is told that banks sometimes change their hours and this fact is used to underscore the possibility that the particular bank under discussion might not be open on Saturday. In fact, the only significant difference between the two cases seems to be that the protagonist is invited to imagine the personal costs associated with being wrong in the Concrete High Standards case. If that is right, then even if we set aside concerns about cross-cultural diversity, it appears that Schaffer and Knobe draw the wrong conclusion. If these cases provide evidence of anything, it seems that they provide evidence that stakes matter.

The idea that stakes might matter after all has also received support from recent work by N. Angel Pinillos (forthcoming). Consider, for example, the following two vignettes (Pinillos, forthcoming):

Low Stakes Typo:

Peter, a good college student, has just finished writing a two-page paper for an English class. The paper is due tomorrow. Even though Peter is a pretty good speller, he has a dictionary with him that he can use to check and make sure that there are no typos. But very little is at stake. The teacher is just asking for a rough draft and it won't matter if there are a few typos. Nonetheless, Peter would like to have no typos at all.

High Stakes Typo:

John, a good college student, has just finished writing a two-page paper for an English class. The paper is due tomorrow. Even though John is a pretty good speller, he has a dictionary with him that he can use to check and make sure that there are no typos. There is a lot at stake. The teacher is a stickler and guarantees that no one will get an A for the paper if it has a typo. He demands perfection. John, however, finds himself in an unusual circumstance. He needs an A for this paper to get an A in the class. And he needs an A in the class to keep his scholarship. Without the scholarship, he can't stay in school. Leaving college would be devastating for John and his family who have sacrificed a lot to help John through school. So it turns out that it is extremely important for John that there are no typos in this paper. And he is well aware of this.

Pinillos asked people how many times Peter or John would need to proofread the paper before knowing that there are no typos. If stakes don't matter, then we should expect no significant difference in the amount of times people think that John would need to proofread the paper before knowing that there are no typos and the amount of times people think that Peter needs to proofread the paper before knowing that there are no typos. This is not what we find; instead, people think that John has to proofread the paper on average 3 more times than Peter in order to know that there are no typos.²⁸ Putting these results together with the Schaffer and Knobe results, it seems like stakes might matter after all. Still, even these results are contested, with Wesley Buckwalter (forthcoming) arguing that these results indicate less about the relationship between stakes and knowledge attributions than they do about the relationship between stakes and rational action.²⁹

What are we to make of the results of all of these studies? On the one hand, we have empirical evidence that suggests that neither stakes nor salience matter. People seem to be just as willing to say that someone knows something when the possibility of being wrong has been made salient as they are when that possibility has gone unmentioned, and people seem just as willing to say that someone knows something when the personal costs associated with being wrong are high as they are when those costs are low. On the other hand, we have empirical evidence that something matters - but we aren't able to determine right now whether it's stakes or salience that matter. Complicating things even more, we don't know whether stakes have an epistemic influence on the truth of knowledge attributions or simply practical influence on the rationality of certain kinds of behaviors.³⁰ We need to be careful drawing conclusions from either set of empirical results at this time.³¹ Maybe earlier studies failed to find statistically significant differences because there were no differences to be found; maybe they found no statistically significant differences because the tools they used to test for such differences weren't sufficiently fine-grained or properly tuned. Maybe later studies found a statistically significant difference because the tools that they used to test for such differences were more fine­grained or well tuned than those used in previous studies; maybe later studies found a statistically significant difference because they were studying a different epistemic population. Maybe later studies found that salience matters; maybe later studies found that stakes matter. Maybe stakes matter for knowledge; maybe stakes matter only for rational action. At this point, we just can't tell. Future research is needed. And, this itself has significant implications for philosophers who have used our intuitive judgments about bank cases to advance certain philosophical arguments and positions. Whatever other reasons we might have for adopting some version of contextualism or subject-sensitive invariantism, it seems like more experimental work is going to be needed before our philosophical intuitions can help settle the debate between these two positions.

4.