Notes

Introduction

1.   For additional discussion of different ways of dividing up the landscape, see Alexander and Weinberg (2007), Knobe (2007a), Nadelhoffer and Nahmias (2007), Liao (2008), and Alexander et al.

2.   For additional discussion of this issue, see Nadelhoffer and Nahmias (2007), Knobe (2007a, 2007b), Knobe and Nichols (2008), and Sommers (2011).

Chapter 1 Philosophical Intuitions

1.   Bealer (1992, 1996a, 1996b, 1998) and Sosa (1996, 1998) are the most explicit in making this comparison. See, also, Nagel (2007, forthcoming). This way of thinking about things leaves open whether treating intuitions (or perceptions) as evidence involves treating psychological states (or propositions about psychological states) as evidence or treating the contents of those psychological states as evidence. As we will see later on, this way of thinking about philosophy isn't shared by all philosophers, even those philosophers who think that intuitions have a role to play in philosophical practice. For example, Kornblith (1998, 2002, 2007) and Williamson (2004, 2005, 2007) argue against the evidential role of intuitions while maintaining that they still have some role to play in philosophy, and Gendler (2007) has recently argued that the role of philosophical thought experiments is not to generate intuitions that can be used as evidence for, or against, philosophical claims, but to help make philosophical theories cognitively accessible.

2.   Some philosophers take these theories to be telling us something about the psychological world (that is, about our individual or shared psychological concepts of things), while others take these claims to be telling us something about the non-psychological world (that is, about things themselves, platonic forms, Fregean concepts, etc.).

3.   This is not to say that everyone changed his or her mind. Weatherson (2003), for example, argues that our reasons for believing that knowledge is justified true belief trump the evidence provided by the philosophical intuitions generated in response to Gettier's two hypothetical cases.

4.   It is important to keep in mind that when philosophers appeal to intuitions as evidence, they are not (at least typically) appealing merely to their own intuitions. Philosophical analysis is not mere intellectual autobiography and for good reason: an appeal to one's own intuitions would not be dialectically effective. For further discussion, see Alexander and Weinberg (2007).

5.   Descriptivism has seen a revival in recent years associated with the rise of two-dimensional semantics. See, e.g., Chalmers (1996), Jackson (1998a, 1998b), and Stalnaker (1999).

6.   For additional discussion of the doctrine of double effect and various versions of the trolley case, including versions of the trolley case that are thought to generate intuitional evidence inconsistent with the doctrine of double effect, see Foot (1967), Thomson (1985), Kamm (1989, 2007), and Liao et al. (forthcoming).

7.   The first strategy has been endorsed by Lynch (2006), the second strategy by Bealer (1998), Sosa (1998), and Pust (2000), and the third strategy by Williamson (2007). Although we won't press this point here, the third strategy is probably the best. One problem with the first strategy is that it isn't clear that we have any intuitions about what intuitions are (or, at least, whether or not intuitions are the sort of thing about which we have intuitions will depend on just what intuitions turn out to be) and so the strategy is unlikely to provide much help to us in determining which conception is correct.

8.   Of course, this doesn't mean that all beliefs or inclinations to believe are philosophical intuitions; almost all philosophers want to put some restriction on what kinds of beliefs or inclinations to believe could count as philosophical intuitions; for example, most philosophers argue that philosophical intuitions should be immediate or at least not explicitly inferential. It is also worth pointing out here that, while the philosophers mentioned in the section typically remain neutral about whether philosophical intuitions are beliefs or inclinations to believe, others, who endorse some kind of doxastic approach, are more partisan on this matter. So, for example, Sosa (1998) and Earlenbaugh and Molyneux (2009a, 2009b) have advanced conceptions that treat philosophical intuitions specifically as inclinations or dispositions to believe.

9.   It is probably worth noting here that Williamson actually suggests that the practice of appealing to philosophical intuitions as evidence rests on a misunderstanding of the nature of philosophical evidence, arguing that philosophers “might be better off not using the word ‘intuition' and its cognates” since our emphasis on intuitional evidence doesn't help “to answer questions about the nature of evidence on offer but to fudge them, by appearing to provide answers without really doing so” (2007, p.

10.      For additional discussion of this point, see Weinberg and Alexander (forthcoming).

11.      Bealer defends this view over a number of papers. See, also, Bealer (1992, 1996a, 1996b). The view that philosophical intuitions present their propositional contents as necessary has also been defended by BonJour (1992, 1998), Plantinga (1993), Sosa (1996), and Hales (2000). Sosa (2007b), Talbot (2009), and Bengson (forthcoming) have recently advanced phenomenological conceptions that don't involve the claim that philosophical intuitions present their propositional contents as necessary, instead saying only that genuine philosophical intuitions are “conscious states of felt attraction”, “strike us as true without us knowing entirely why they do”, or “strike us in a certain way”, respectively.

12.      Bealer (1998) provides two other reasons for thinking that intuitions are not beliefs. First, he argues that beliefs are plastic in a way that intuitions are not; while you can get someone to believe almost anything at all, it seems to be impossible to get certain propositions to seem to be true in the relevant phenomenological sense of “seems”. Second, he argues that there are no significant limitations to the kinds of propositions about which you can have beliefs while there are limitations to the kinds of propositions about which you can have intuitions. We will focus on Bealer's first two reasons because they seem to be, on balance, stronger reasons for maintaining that philosophical intuitions are not beliefs.

13.      Another point of controversy has been the claim that philosophical intuitions are sui generis propositional attitudes. Both Sosa (1998) and Williamson (2007), for example, argue that they find no intellectual seeming that p beyond their own conscious inclinations to believe that p.

14.      Ludwig (2010) argues that the semantic approach is also too liberal, letting in certain things that we don't want to count as philosophical intuitions. As Ludwig points out, some abstract propositions - for example, the claim that there is no greatest prime number - are simply too complicated for us to form intuitions about them.

15.      For additional discussion of this issue, see Kornblith (2002), Weinberg and Alexander (forthcoming), and Weinberg et al. (forthcoming).

Chapter 2 Experimental Philosophy and Philosophical Analysis

1.   See, for example, Strawson (1986), Kane (1999), O'Connor (2000), Pereboom (2001), and Pink (2004).

2.New Roman">   More specifically, 72% of subjects who were asked to evaluate Jeremy's actions in the Supercomputer case judged that he acted of his own free will when he robbed the bank, while 83% of subjects judged that he was morally blameworthy for his actions.

3.   A quick note about how determinism is described in these studies. Nahmias et al. (2007) suggest that people's philosophical intuitions are also influenced by how determinism is described. In particular, they found that people are more likely to express compatibilist intuitions when determinism is described in nonmechanistic, psychological terms than when determinism is described in mechanistic, neuroscientific terms. Interestingly, they still found that people's emotional responses to affectively charged vignettes influence their intuitive judgments about whether moral responsibility is compatible with causal determinism, regardless of how determinism is described. Roskies and Nichols (2008) also found that people's philosophical intuitions are influenced by how determinism is described. They found that people's philosophical intuitions are sensitive to whether deterministic scenarios are presented as actual or merely possible.

4.   More specifically, 86% of subjects who were asked whether it is possible in Universe A for an agent to be fully morally responsible for her actions judged that it is not possible.

5.   More specifically, 72% of subjects who were asked to consider the Murderous Husband case judged that Bill was morally responsible for his actions.

6.   In order to rule out the possibility that people's judgments are tracking abstractness/concreteness alone, Nichols and Knobe ran a second study involving only high affect (involving habitual rape) and low affect (involving habitual tax evasion) concrete cases. Their results confirmed that people are more likely to say that a person is morally responsible for her actions in high affect cases than in low affect cases even when her actions are causally determined.

7.   For further discussion of this point, see Alexander et al. (2010a, 2010b).

8.   Marr (1982) calls this the computational level, Pylyshyn (1984) calls this the semantic level, and Glass et al. (1979) call this the content level.

9.   Marr (1982) calls these two levels the algorithmic and implementational levels, Pylyshyn (1984) calls them the syntactic and physical levels, and Glass et al. (1979) call them the form and medium levels.

10.      More specifically, 67% of subjects who were asked to consider both cases judged that neither Bill nor Mark is morally responsible for his actions, 25% of subjects who were asked to consider both cases judged that both Bill and Mark are morally responsible for their actions, and 8% of subjects who were asked to consider both cases gave mixed answers.

11.      For additional discussion of this dilemma, see Alexander et al. (2010a).

12."Times New Roman"'>      See, e.g., Christensen (2007), Elga (2006), Feldman (2006), Feldman and Warfield (2007), Kelly (2005, 2007, 2008), and White (2005).

13.      For recent discussions of relativism, see Boghossian (2006) and Hales (2006).

14.      Although, recent work by Nahmias and Murray (2011) presents interesting evidence that might provide the basis for discounting one set of philosophical intuitions about moral responsibility. They found that most people who think that freedom and moral responsibility are incompatible with causal determinism do so because they mistakenly believe that causal determinism involves some kind of fatalism when agent's mental states (including intentions) are bypassed in the relevant causal processes.

15.      Sommers (2010) provides a number of suggestions about how to address some of the worries discussed above, and points to a number of questions worth addressing as experimental research on this topic moves forward.

16.      For other vignettes commonly appealed to in this debate, see Cohen (1998), Fantl and McGrath (2002), and Stanley (2005).

17.      See, for example, DeRose (1992), Lewis (1996), and Cohen (1998).

18.      See, for example, Fantl and McGrath (2002), Hawthorne (2004), and Stanley (2005). Stanley calls his brand of invariantism interest-relevant invariantism.

19.      Actually, the debate between contextualists and subject­sensitive invariantists is a bit more complicated since both groups agree that knowledge is sensitive to what is at stake for the subject. What they disagree about is whether or not making salient the possibility that the subject is wrong affects epistemic standards. Contextualists think that it does; invariants think that it does not. As such, the debate isn't really about whether salience matters or stakes matter (since both sides accept that stakes do matter) but about whether or not salience matters.

20.      More specifically, subjects were asked to rate (using a 5-point Likert scale with 1 = strongly disagree and 5 = strongly agree) the degree to which they agreed or disagreed that Bruno knows that the bank will be open on Saturday. The mean response for subjects who were asked to consider the Low Standards Bank case was 3.83, while the mean response for subjects who were asked to consider the High Standards Bank case was 3.64.

21.      More specifically, subjects were asked to rate (using a 7-point Likert scale with 1 = strongly disagree and 7 = strongly agree) the degree to which they agreed or disagreed that Hannah knows that the bank will be open on Saturday. The mean response for subjects who were asked to consider the Low Stakes No Alternative case was 5.07, while the mean response for subjects who were asked to consider the High Stakes Alternative case was 4.60.

22.      DeRose (2011) has raised several concerns about these studies, including that participants in some studies are asked to rate their willingness to make knowledge attributions rather than being asked to evaluate knowledge attributions, and that when participants are asked to evaluate knowledge attributions, they are asked to do so even in situations where it would be more natural to deny knowledge. Since what conversational context is relevant depends (at least in part) on whether you are being asked to make a knowledge attribution or to evaluate a knowledge attribution, and since being asked to evaluate knowledge attributions when it is natural to deny knowledge introduces a host of pragmatic implications that can influence our judgments, DeRose is worried that these studies fail to actually test the truth or plausibility of epistemic contextualism. While these are legitimate concerns, Buckwalter (forthcoming) finds that salience doesn't seem to matter even when these concerns have been addressed in the experimental design of the studies. DeRose also raises a host of more general concerns about the methods employed by experimental philosophers; these concerns will be addressed

in the epilogue.

24.      More specifically, subjects were asked to rate (using a 7-point Likert scale with 1 = strongly agree and 7 = strongly disagree) the degree to which they agreed or disagreed that Hannah knows that the bank will be open on Saturday. The mean response for subjects who were asked to consider the Simplified Low Stakes case was 3.85, while the mean response for subjects who were asked to consider the Simplified High Stakes case was 3.83. Feltz and Zarpentine (2010) also found no statistically significant differences in people's intuitive judgments about analogous high and low stakes cases. Actually, when they collapsed all of the high stakes and low stakes experiments they ran, they did find a statistically significant difference, but the effect size barely escaped triviality. This should be little solace, then, for those philosophers who think that stakes are an important and pervasive influence on knowledge attributions.

25.      More specifically, subjects were asked to rate (using a 7-point Likert scale with 1 = not confident and 7 = very confident) how confident Kate should be that she is on Main Street. The mean response for subjects who were asked to consider the Unimportant case was 5.11, while the mean response for subjects who were asked to consider the Important case was 4.93.

26.      Although, Brown (forthcoming) has recently argued that the insensitivity of folk knowledge attributions to things like stakes and salience undermines contextualism but not subject­sensitive invariantism.

27.      More specifically, subjects were asked to rate (using a 7-point Likert scale with 1 = disagree and 7 = agree) the degree to which they agreed or disagreed that Hannah knows that the bank will be open on Saturday. The mean response for subjects who were asked to consider the Revised Low Standard case was 5.54, while the mean response for subjects who were asked to consider the Concrete High Standard case was 3.05. This difference was statistically significant at the level p < 0.001. Nagel (2010) also found that salience matters, finding that making error possibilities salient makes people less likely to agree with knowledge assertions and more likely to agree with knowledge denials.

28.      More specifically, the median response for subjects asked to consider the Low Stakes Typo case was 2 and the median response for subjects asked to consider the High Stakes Typo case was 5. This difference was statistically significant at the level p < 0.001.

29.      Pinillos and Simpson (forthcoming) respond to this worry with a series of new studies that focus more directly on knowledge attributions and that still find evidence that stakes matter.

30.      See Hawthorne and Stanley (2008) for an interesting discussion of the relationship between knowledge and rational action.

31.      Weatherson (forthcoming) argues that what might seem like inconclusive evidence that stakes matter is actually conclusive evidence after all. According to Weatherson, subject-sensitive invariantists don't think that stakes always matter; instead, they simply think that stakes sometimes matter. If this is right, then we should be less concerned by the fact that some experimental studies showed that stakes matter while others don't.

Chapter 3 Experimental Philosophy and the Philosophy of Mind

1.   See also Goldman and Pust (1998). Goldman's worry is similar to what has come to be known as Benacerrafs Problem: according to our best theory of mathematical truth, mathematical statements refer to abstract entities; but, according to our best theory of knowledge, a person knows that p only if there is a causal connection between that person and whatever it is that makes p true; since abstract entities cannot stand in causal relationships, there is a problem. This worry leads Goldman to endorse mentalism, the view that philosophical intuitions are evidence only about certain facts (the structure or meaning, for example) of our concepts.

2.   More specifically, 82% of subjects who were asked to evaluate the chairman's actions in the Environmental Harm case judged that the chairman intentionally harmed the environment, while only 23% of subjects who were asked to evaluate the chairman's actions in the Environmental Help case judged that the chairman intentionally helped the environment. This difference is statistically significant at the level p < 0.001. Knobe (2003a) found that similar results obtain for a different pair of vignettes with the same basic structure.

3.   More specifically, 76% of subjects who were asked to evaluate Jake's actions in the No Skill/Immoral Outcome case judged that he hit his aunt intentionally, while only 28% of subjects who were asked to evaluate Jake's actions in the No Skill/Positive Outcome case judged that he hit the bull's-eye intentionally. This difference is statistically significant at the level p < 0.01. In addition to the two studies discussed, Cushman and Mele (2008), Knobe (2004a, 2004b), Knobe and Burra (2006), Knobe and Mendlow (2004), Leslie et al. (2006), Malle (2006), McCann (2005), Mele and Cushman (2007), Nadelhoffer (2004a, 2004b, 2005, 2006a, 2006b), and Young et al. (2006) have found that similar results obtain for different pairs of vignettes with the same basic structure.

4.   Recently, a third explanation has come into fashion; namely, that these results tell us something about the relationship between moral judgments and folk psychology more generally, and is best explained in terms of specific features of the underlying mechanism responsible for a broad range of folk psychological judgments. See, e.g., Knobe (2006), Nadelhoffer (2006b), Alicke (2008), Nado (2008), and Pettit and Knobe (2009).

5.   Grice (1989) introduced the notion of conversational implicature to refer to something that is suggested by a sentence when this is distinct from what is expressed or entailed by the literal meaning of the sentence.

6.   A similar story can be provided to explain why people tend to say that the chairman helped the environment unintentionally. People do not want to praise the chairman for helping the environment in the Environmental Help case. If they said that the chairman helped the environment intentionally, then this would imply that he should be praised for his actions. As a result, when they are asked whether or not the chairman helped the environment intentionally, people tend to say that the chairman didn't help the environment intentionally. Again, people's tendency to say that the chairman didn't help the environment intentionally doesn't reflect the nature of their folk concept of intentional action; it merely reflects their desire to refrain from praising the chairman for helping the environment.

7.   More specifically, using a rating scale ranging from -3 (‘sounds wrong') to +3 (‘sounds right'), people who were asked to consider the Environmental Harm case were asked whether or not it sounded right to say that the chairman harmed the environment in order to increase profits and those who were asked to consider the Environmental Help case were asked whether or not it sounded right to say that the chairman helped the environment in order to increase profits. The average rating for people who were asked to consider the Environmental Harm case was +0.6 and the average rating for people who were asked to consider the Environmental Help case was -1.0. This difference is statistically significant at the level p = 0.01.

8.   Malle (2006) offers a number of his own interpretations of the results. Among the most interesting interpretations is that the side-effect effect might best be interpreted as resulting from the way in which the vignettes are presented. According to Malle, since participants' attention is focused on the evaluative components of the vignette, they might think that they are supposed to use their moral evaluations when making their intentionality judgments. In other contexts where their attention isn't so focused, they might not let their evaluative judgments drive their intentionality judgments.

9. Knobe (2003b) draws this case from Mele (2001).

10.      Malle and Nelson (2003) also argue that these results are best explained in terms of affective bias. They studied couples engaged in heated arguments (arguments in which each party develops a sufficiently negative affective response to the other party) and found that such couples typically judge all of their opponent's actions as intentional - even those that would not otherwise be so judged. These studies suggest to them that negative affective responses distort people's intentionality judgments.

11.     Nadelhoffer (2004a) argues that positive affective responses seem to have a similar influence on people's intentionality judgments. Consider the following vignette:

Academic Help:

Imagine that Steve and Jason are two friends who are competing against one another in an essay competition. Jason decides to help Steve edit his essay. Ellen, a mutual friend, says, “Don't you realize that if you help Steve, you will decrease your own chances of winning the competition?” Jason responds, “I know that helping Steve decreases my chances of winning, but I don't care at all about that. I just want to help my friend!” Sure enough, Steve wins the competition because of Jason's help.

Nadelhoffer found that most people who were asked to consider the Academic Help case have a positive affective response to Jason and judge that Jason decreased his own chances of winning the competition intentionally.

12.     More specifically, using a rating scale ranging from 0 (no blame) to 6 (lots of blame), people who were asked to consider the Police Officer case were asked how much blame the thief deserved for the death of the officer and those who were asked to consider the Car Thief case were asked how much blame the driver deserved for the death of the thief. The average rating for people who were asked to consider the Police Officer case was 5.11 and the average rating for people who were asked to consider the Car Thief case was 2.01. Nadelhoffer associates people's negative affective responses with their willingness to ascribe blame and concludes that people asked to consider the Police Officer case had a stronger negative affective response than those asked to consider the Car Thief case.

13.      More specifically, 37% of subjects who were asked to consider the Police Officer case judged that the thief intentionally brought about the officer's death, while only 10% of subjects who were asked to consider the Car Thief case judged that the driver intentionally brought about the car thief's death. This result is statistically significant at the level p < 0.001.

14.      Knobe and Mendlow (2004) argue that evidence also suggests that people are willing to withhold blame in cases where a morally bad outcome is thought to be intentional.

15.      In fact, Guglielmo and Malle (2010) argue that evidence suggests that people often blame the chairman for harming the environment because they believe that the chairman did so intentionally.

16.      This might be too strong. It is possible that a number of different factors might play a role in generating our intentionality judgments, and that blame and affect might sometimes be among these factors. What the considerations just rehearsed do is put pressure on the idea that these things are the only factors responsible for our intentionality judgments.

17.      For additional discussion, see Alexander et al. (2010a, 2010b).

size=4 color=black face=Arial>18.      Phelan and Sarkissian (2009) argue that there are two problems with the Extra Dollar case. First, if people view paying an extra dollar for a smoothie as a cost, they are likely to view it is a bad thing (although maybe not morally bad). As such, the case doesn't seem to be right for deciding between an explanation of the side-effect effect according to which people's intuitions about whether or not an action was performed intentionally are being influenced by normative considerations and an explanation of the side-effect effect according to which people's intuitions about whether or not an action was performed intentionally are being influenced simply by considerations of costs and benefits. (Phelan and Sarkissian have a similar worry about the subject-oriented reading of the trade-off hypothesis discussed below; namely, that it comes close to Knobe's own hypothesis that people's intuitions about whether or not an action was performed intentionally are being influenced by normative considerations.) Second, it is not clear that the Extra Dollar case actually involves a foreseen side effect. Recall that an outcome is a foreseen side effect just in case an agent chooses to perform an action that she foresees will bring about the outcome but doesn't perform the action for the purpose of trying to bring it about. But, something done in order to bring about a specific outcome isn't a side effect of an action performed to bring about that outcome; it is part of the action performed in order to bring it about. As such, since the agent pays an extra dollar in order to get the Mega-Sized Smoothie, paying the extra dollar isn't a side effect at all.

19.      More specifically, 95% of subjects who were asked to evaluate Joe's actions in the Extra Dollar case judged that he intentionally paid an extra dollar, while only 45% of subjects who were asked to evaluate Joe's actions in the Free Cup case judged that he intentionally bought the commemorative cup. This difference is statistically significant at the level p < 0.001.

20.      More specifically, 56% of subjects who were asked to evaluate John's actions in the Worker case judged that he intentionally caused the death of the workman on the side tracks, while only 23% of subjects who were asked to consider the Dog case judged that John intentionally saved the dog. This difference is statistically significant at the level p < 0.01.

21.      Phelan and Sarkissian (2009) point out that there is a problem with using this study to challenge Knobe's hypothesis. While most people view both actions as morally appropriate, Phelan and Sarkissian speculate that most people would think that the first action (causing the death of the workman on the side track) is bad (although maybe not morally bad), while the second action (saving the dog) is good (probably even morally good). If this is right, then Knobe's hypothesis would actually predict these results.

22.      See Mallon (2009) and Phelan and Sarkissian (2009).

23.      More specifically, 92% of subjects who were asked to evaluate the terrorist's action in the Harmful Terrorist case judged that the terrorist intentionally harmed the Australians, while only 12% of subjects who were asked to evaluate the terrorist's action in the Helpful Terrorist case judged that the terrorist intentionally helped the orphanage. This difference is statistically significant at the level p < 0.0001. Mallon (2009) based these vignettes on vignettes introduced by Knobe and Kelly (2009). Mallon found that similar results obtain for a different pair of vignettes with the same basic structure.

24.      Since the vignette specifies that the lieutenant also views the soldiers' deaths as a cost, this result provides additional evidence against the view that people's intuitions about whether or not an action was performed intentionally are being influenced by the fictional agent's perception of something as a cost.

25.      Phelan and Sarkissian (2009) have another argument against the subject-oriented reading. According to them, if the subject- oriented reading is correct, then we should expect people to be more willing to judge that an action is intentional when they perceive it as a cost incurred in the pursuit of an important goal than when they perceive it as a cost incurred in the pursuit of an unimportant goal. Since they find experimental evidence to suggest that this is not the case, they conclude that the subject-oriented reading is incorrect. The problem is that there is no reason, on any version of the trade-off hypothesis, to expect people to be more willing to judge that an action is intentional when they perceive it as a cost incurred in the pursuit of an important goal than when they perceive it as a cost incurred in the pursuit of an unimportant goal. Costs are costs. For the trade-off hypothesis, nothing hinges on whether the agent was wise to incur the costs. All that matters is that costs incurred are incurred intentionally.

Chapter 4 Experimental Philosophy and Philosophical Methodology

1.   For wonderful discussions of other forms of cultural cognitive diversity, see Nisbett et al. (2001), Nisbett and Norenzayan (2002), and Nisbett (2003). Nisbett and colleagues have found that a variety of different cognitive processes, including perception and memory, display cultural sensitivity. For an early, and influential, study of moral intuitions and cultural diversity, see Haidt et al. (1993).

2.   In particular, they found that 26% of Western subjects, 56% of East Asian subjects, and 61% of South Asian subjects judged that Bob knows that Jill drives an American car. The difference between Western subjects and both East Asian and South Asian subjects is statistically significant at the level p < 0.01.

3.   In particular, descriptivist judgments about the vignettes received a score of 0 while causal-historical judgments received a score of 1. The scores for each vignette were then summed so that the cumulative score could range from 0 to 2. Using this procedure, Machery and his colleagues found that Western subjects were more likely than East Asian subjects to have causal-historical intuitions. The difference between the mean response for Western subjects (1.13) and the mean response for East Asian subjects (0.63) is statistically significant at the level p < 0.05. For recent discussions of this study, including both theoretical and empirical concerns, see Ludwig (2007), Deutsch (2009), Marti (2009), Sytsma and Livengood (20i1), and Ichikawa et al. (forthcoming).

4.   A close look at the empirical results might suggest that this move is rather unique even within Western culture since even

a significant amount of people from Western backgrounds have intuitions about these cases that are inconsistent with the causal-historical account of reference.

5.   More specifically, subjects were asked to rate (using a 7-point Likert scale with 1 = strongly disagree and 7 = strongly agree) the degree to which they agreed or disagreed that it is morally acceptable to redirect the trolley in the two cases. In the Stranger case, the mean response for men was 4.21, while the mean response for women was 4.95. In the Child case, the mean response for women was 4.26, while the mean response for men was 4.87. The difference between responses by men and women to these two cases approached statistical significance with p = 0.07. In addition to these two cases, Zamzow and Nichols asked subjects to consider cases in which the individual on the side track was the participant’s brother and cases in which the individual on the side track was the participant’s sister. Interestingly, they found that men judged killing one’s brother to be less morally acceptable than women (M = 3.41 vs. M = 4.33) and that women judged killing one’s sister to be less morally acceptable than men (M = 3.78 vs. M = 4.40). As Zamzow and Nichols write, “there appears to be a bias in favor of one’s own gender, at least when it comes to siblings and speeding trains.”

6.   More specifically, only 41% of male subjects said that Peter “really knows” that there is a watch on the table, while 71% of female participants said that Peter “really knows” that there is a watch on the table. This difference is statistically significant at the level p < 0.05. In order to rule out the possibility that the effect was being driven by the gender of the protagonist, Starmans and Friedman ran another study using a slightly different vignette whose protagonist was female. If anything, the results were even more striking. In this version, only 36% of male participants said that the female protagonist really knows, while 75% of female participants said that she really knows. This difference is statistically significant at the level p < 0.01.

7.   More specifically, only 35% of male participants said that someone in this world could be free to choose whether or not to murder someone else, while 63% of female participants said that someone in this world could be free to choose whether or not to murder someone else. This difference is statistically significant at the level p < 0.0005.

8.   In fact, there is some evidence that interpersonal intuitional diversity is even finer grained than this, tracking individual personality types. See, e.g., Cokely and Feltz (2009) and Feltz and Cokely (2007, 2009).

9.   Zamzow and Nichols (2009) provide a different reason to think that we should not worry so much about intuitional diversity. Drawing on the analogy between intuitional evidence and empirical evidence, they argue that lessons from the history of science, together with work from the social and cognitive sciences, indicate that evidential diversity is actually an epistemic good to be promoted rather than feared. The idea is that evidential diversity plays a role in epistemic progress. Zamzow and Nichols are certainly right that more attention should be paid to the epistemic benefits of evidential diversity, but for our purposes, it's important to keep in mind that change involves choice - the resolution or at least dissolution of disagreements - and that any claim to progress requires some way of defending those choices as right. It is not clear, then, that their suggestion takes us past the need for a principled way of resolving evidential diversity, although it might make us more sanguine about our prospects for being able to find one.

10.       Moral intuitions also seem to be sensitive to facts about the person who is considering the hypothetical case. Zamzow and Nichols (2009) recently found that moral intuitions demonstrate sensitivity to gender.

11.     Uhlmann et al. (2009) used a linear regression to test for independent and interactive effects between scenario condition (Tyrone Payton vs. New York Philharmonic = -1/Chip Ellsworth III vs. Harlem Jazz Orchestra = 1), political orientation (1 = very liberal, 5 = very conservative) and endorsement of consequentialism (1 = completely disagree, 7 = completely agree). They found a reliable lower-order effect: subjects were less willing to endorse consequentialism when Tyrone Payton was sacrificed than they were when Chip Ellsworth III was sacrificed (b = -0.19, SE = 0.09, t(84) = 2.24, p = 0.03). The effect was even more pronounced for people who self­identified as politically liberal (1 SD below the mean):

b = -0.40, SE = 0.12, t = 3.27, p = 0.002.

12.     Those who identify themselves as politically conservative aren't immune from this kind of intuitional sensitivity. In a separate study, Uhlmann and his colleagues found that people who identified themselves as politically conservative were more likely to think that it is morally permissible for American military forces to engage in an action that involved collateral civilian casualties to stop Iraqi insurgents than they were to think that it is morally permissible for Iraqi insurgents to engage in an action that involved collateral civilian casualties to stop American military forces.

13.      More specifically, subjects were asked to rate (using a 5-point Likert scale with 1 = strongly disagree and 5 = strongly agree) the degree to which they agreed or disagreed that Charles knows that it is 71 degrees in the room. The mean response for subjects who were asked to evaluate the Truetemp case before evaluating any other cases was 2.8; the mean response for subjects who were asked to evaluate the Truetemp case after first being asked to evaluate a clear case of knowledge was 2.4; and the mean response for subjects who were asked to evaluate the Truetemp case after first being asked to evaluate

a clear case of non-knowledge was 3.2. The difference between these conditions is statistically significant at the level p < 0.05. These results were confirmed, replicated, and extended by Wright (2010), who offers an interesting explanation of intuitional sensitivity to presentation order, and by Weinberg et al. (forthcoming), who examined the relationship between people's need-for-cognition (Cacioppo and Petty 1982) and intuitional sensitivity to presentation order.

14.      For a fascinating discussion of this work, together with a discussion of the methodological problems that this kind of sensitivity raises for traditional philosophical methodology, see Sinnott-Armstrong (2008).

15.      Subjects were asked to indicate their willingness to act to save the five track workers using a scale ranging from +5 to -5. The mean when the switch version appeared first was 3.1, the mean when the switch version appeared third was 1.0. The mean when the fat-man version appeared first was -0.86, the mean when the fat-man version appeared third was -1.7. These results are significant at the level p < 0.005.

16.     As before, subjects were asked to indicate their willingness to act to save the five track workers using a scale ranging from +5 to -5. The mean response when the button version followed the switch version was 2.7; the mean response when the button version followed the fat-man version was 0.65. This result is significant at the level p < 0.02.

17.     The Petrinovich and O’Neill study is somewhat limited by the fact that they asked what people would do rather than what people thought was morally permissible or morally right. As Zamzow and Nichols (2009) rightly note, “judgments of what you would do in a situation can come apart from your moral judgments”. This problem has been resolved in more recent studies on the contextual sensitivity of moral intuitions (and, in particular, Trolley intuitions). See, in particular, Liao et al. (forthcoming) and Schwitzgebel and Cushman (2011).

18.      In fact, there are at least two extant philosophical theories according to which some of these things - in particular, cultural background and context - are relevant: namely, relativism and contextualism. According to relativism, the truth or falsity of certain philosophical claims depends - at least, in part - on facts about the person evaluating the claim. And, according to contextualism, the truth or falsity of certain philosophical claims depends - at least, in part - on the context in which those claims are made and evaluated. (For recent discussions of relativism, see Boghossian 2006 and Hales 2006, and for recent discussions of contextualism, see DeRose 1995, Lewis 1996, and Cohen 1998.) While it is somewhat beyond the scope of the present work to evaluate how successful either theory might be at explaining why such things as cultural background, gender, affective content, or context might be relevant to the truth or falsity of philosophical claims, two points are worth considering. First, what has been shown by these studies is that some philosophical intuitions are sensitive to cultural background, gender, affectivity, and context. What the relativist or contextualist needs, however, is some independent argument that philosophical intuitions should be sensitive to these kinds of things. That is, they need some way to move from the descriptive claim to the normative claim. Second, it is not clear that relativism or contextualism - taken individually or jointly - can explain why all of the things are relevant.

Chapter 5 In Defense of Experimental Philosophy

1.   This label comes from Weinberg et al. (2010), who provide the first detailed discussion of the strengths and weaknesses of this defense. Singer (1972, 1982), Devitt (1994, 2006, 2010), Williamson (2005, 2007), Hales (2006), Ludwig (2007), Grundmann (2010), and Horvath (2010) advance versions of the expertise defense. Interestingly, its central tenet (that philosophers should be more concerned with the intuitions of experts than those of the folk) can also be found in Stich and Nisbett's (1980) discussion of the method of reflective equilibrium and the justification of rules of inference. Concerned that a significant number of unacceptable rules of inference stand in reflective equilibrium with folk inferential practices, they argue that a rule of inference is justified just in case it stands in reflective equilibrium to the inferential practices of trained authorities, or experts. It is important to note, however, that Stich (1988) takes back this earlier defense of expert reflective equilibrium.

2.   Jackman (2001), Nichols (2004a), Goldman (2007), and Knobe and Nichols (2008) make similar points.

3.   Ludwig (2007) makes this suggestion, writing about expertise that it “is not a matter of acquiring new concepts! It is a matter of gaining greater sensitivity to the structure of the concepts through reflective exercise with problems involving those concepts” (p. 149).

4.   Ludwig (2007) seems to have something similar in mind in the following example: Mr. Smith, a normal fit man in his twenties, is standing by the window in his living room. He hears the phone ring on the side table by the couch on the other side of the room. He walks to the side table across the open floor between him and the phone and picks it up. Which of the following is true: (A) Mr. Smith tried to walk across the living room; (B) Mr. Smith did not try to walk across the living room; or (C) Neither (A) nor (B). According to Ludwig, we see that (A) is the correct answer since this answer best fits our best theory of intentional action, according to which a person can attempt to do something even in cases in which it is almost assured that he will succeed.

5.   See Livengood et al. (2010) for an interesting empirical study that suggests that philosophers are more reflective than the folk.

6.   Kornblith (2002) provides two nice examples. Kornblith's first example is Nisbett and Wilson's (1977) study on position effects. Nisbett and Wilson found that people's assessment of consumer goods is influenced by the relative position of the goods being evaluated. Even when asked to reflect on what factors contributed to their assessment, people remain unaware of the influence of relative position on their assessments. Kornblith's second example is Tversky and Kahneman's (1974) study on anchoring effects. In this study, subjects were asked to estimate the percentage of African nations that are members of the United Nations. When asked if the percentage was greater or less than 10%, the mean estimate was 25%; when asked if the percentage was greater or less than 65%, the mean estimate was 45%. While it is clear that subjects are influenced by the initial reference point (the “anchor”), subjects are completely unaware of the influence that this reference point has on their estimates.

7.   In fact, Weinberg et al. (forthcoming) give us reason to worry that even reflective intuitions display the same kinds of problematic intuitional sensitivities that were discussed in the previous chapter, although possibly in slightly different ways.

8.   Ichikawa and Jarvis (2009) also explore the similarities between philosophical thought experiments and fiction, building on David Lewis's (1978) work on “truth in fiction”. Camp (2009) also provides a fascinating treatment of the role of imagination in philosophical thought experiments.

9.   Interestingly, what work has been done on expert philosophical intuitions doesn't look particularly promising for proponents of the expertise defense. Studies by Schwitzgebel and Cushman (2011) and by Schultz et al. (forthcoming) seem to suggest that expert philosophical intuitions display some of the same kinds of problematic intuitional sensitivity that folk philosophical intuitions display.

10.      Kauppinen (2007), Cullen (2010), and Bengson (forthcoming) also advance versions of the thickness defense.

11.      Weinberg and Alexander (forthcoming) call these two kinds of conditions veritist conditions and methodological conditions, respectively. They argue that there are two important veritist conditions: an immunity condition - philosophical intuitions shouldn't be subject (to any worrisome degree) to the kinds of evidential diversity and problematic evidential sensitivity displayed by the kinds of mental states studied by experimental philosophers - and a Hippocraticity condition - philosophical intuitions shouldn't be subject (to any worrisome degree) to other kinds of problematic evidential sensitivity. They also argue that there are two important methodological conditions: a manifestability condition - we can tell which mental states are genuine philosophical intuitions - and a current practice condition - we can do this using the methods commonly associated with traditional analytic philosophy.

12.      Ichikawa (forthcoming) makes the same point, arguing that it is hard to come by examples of philosophers making explicit appeal to our philosophical intuitions in philosophical discussions, except in cases where philosophers (for example, Goldman 2007) are engaged in studying human psychology, namely, our shared or individual philosophical concepts.

13.      Actually, it isn't clear that Williamson would be entirely happy with this way of talking about things. Williamson's preferred reconstruction of the Gettier argument is this:

(A)                                 lang=EN-US>It's possible for someone to stand in relation to some proposition p just as the protagonists of Gettier's cases stand to the relevant propositions.

(B)                                 If someone were to stand in relation to some proposition p just as the protagonists of Gettier's cases stand to the relevant propositions, then anyone who stood in relation to p just as the protagonists of Gettier's cases stand to the relevant propositions would have a justified true belief that p that isn't knowledge.

(C)                                Therefore, it is possible for someone to have a justified true belief that isn't knowledge.

Although (A) and (C) are the same as (1) and (3), it is important to note that (B) is not the same as (2). In fact, Williamson rejects something like (2) on the grounds that there are ways of filling out the details of Gettier's thought experiments such that the vignettes don't present cases of justified true beliefs that aren't knowledge. Since this point is rather controversial, and since nothing turns in what follows on which formulation we use, we can stick with (1)-(4).

14.      It is important to keep in mind that when philosophers appeal to intuitions as evidence, they are not (at least typically) appealing merely to their own intuitions. Philosophical analysis is not mere intellectual autobiography and for good reason: an appeal to one's own intuitions would not be dialectically effective. For further discussion, see Alexander and Weinberg (2007).

15.      Ichikawa (forthcoming) introduces an interesting distinction between different kinds of metaphilosophical claims:

(1)                               Intuited contents are (often) taken as important evidence/ reasons/data/input in armchair philosophy.

(2)                               Intuited contents are (often) taken as important evidence/ reasons/data/input in armchair philosophy because they are intuited.

(3)                               Intuition states, or facts about intuition states, are (often) taken as important evidence/reasons/data/input in armchair philosophy.

In these terms, the view being defended here is either (2) or some hybrid of (2) and (3).

16.      Here, we are assuming that one of the functions of evidence is to provide an epistemic basis for our beliefs. On this view of evidence, if Williamson is right that the evidence we have for believing that (2) consists in the fact that (2), then the epistemic basis for that belief would also be the fact that (2). Although this is not one of the functions that Williamson explicitly assigns to evidence (see, e.g., Williamson 1997, 2000), nothing that he says about evidence rules out its functioning in this capacity (see, e.g., Williamson 1997, p. 728). Having said that, there are places where it seems that Williamson has a different basis in mind for our belief that (2): namely, our “capacity for applying epistemological concepts” (Williamson 2007, p. 189). It is hard to see, however, that viewing things in this way actually helps Williamson’s case against the evidentiary status of intuitions. After all, most philosophers engaged in the debate about whether or not philosophical intuitions count as evidence view this capacity precisely as the capacity for producing epistemic intuitions. This would make (2) a philosophical intuition (at least in the sense relevant to any debate about their evidentiary status) and would, subsequently, mean that intuition plays a direct evidentiary role in the argument for the truth of (4).

17.      In addition to this worry about the dialectical standard of evidence, Williamson also claims that appealing to the dialectical nature of philosophy doesn't favor treating philosophical intuitions as evidence. According to Williamson, there are some contexts in which there is (or would be) controversy about what counts as a genuine philosophical intuition, what counts as the proper expression of a philosophical intuition, why philosophical intuitions should count as evidence, whether or not there are such things as philosophical intuitions, etc. (see, for example, Williamson 2004, pp. 119-123; 2007, pp. 235-237.) Of course, even if Williamson is right, the fact that there might be contexts in which we can't appeal to philosophical intuitions as evidence doesn't mean that we can never appeal to them. To support this stronger claim, Williamson would have to make the case that such problems will arise in all contexts. Since Williamson doesn't argue for this stronger claim, it doesn't seem like this can be reason to reject a dialectical picture of philosophy that treats philosophical intuitions as evidence.

18.      We may need to adjust this revised dialectical standard somewhat to deal with people whose skeptical attitudes are local rather than global - people who deny only that certain kinds of reasons are good enough reasons for believing that p. The important point, however, is that we can distinguish between being unpersuaded and being (locally or globally) unpersuadable and that it seems possible to adopt a dialectical standard of evidence that doesn't require that evidence be capable of persuading the unpersuadable.