Perspective Intentional Action 1 Running Head: PERSPECTIVE INTENTIONAL ACTION Perspective in Intentional Action Attribution: Reversing the Side-Effect Effect

Empirically minded researchers (e.g., experimental philosophers) have begun exploring the “folk” notion of intentional action, often with surprising results. We extend these lines of research and present evidence from a new paradigm in experimental philosophy using some methods of behavioral economics. Our experiments indicate that in some circumstances people make strikingly different judgments about intentions and intentionality as a function of whether they bring about or observe an event. In some of these circumstances, a well-known asymmetry in intentional action attribution, the side-effect effect, can be reversed. Implications for traditional action theory and the experimental study of folk intuitions are discussed. Perspective Intentional Action 3 Determining one's intentions or whether one behaves intentionally has been a central concern for philosophers of action (Mele, 1992). Many of these philosophers take themselves to be exploring the everyday or “folk” conception of intentional action (Adams, 1986; McCann, 1986, 2005; Mele, 1992). Some of these philosophers even hold that “a philosophical analysis of intentional action that is wholly unconstrained by that [folk] concept runs the risk of having nothing more than a philosophical fiction as its subject matter” (Mele, 2001, 27). Empirically minded researchers (e.g., experimental philosophers) have helped shed light on the folk notion of intentional action, often with surprising results. We present evidence from a new paradigm in experimental philosophy using some of the methods of behavioral economics. Our results suggest that in some circumstances people make strikingly different judgments about intentions and intentionality partially as a function of whether a person brings about or observes an event. Using these new methods can reverse a well-known phenomenon—the side-effect effect. Implications for traditional action theory and the study of philosophically relevant folk intuitions are discussed. Experimental Philosophy and Action Theory Experimental philosophy is a relatively new movement in philosophy. Experimental philosophy is characterized by the use of experimental methodologies of psychology and cognitive science to help shed light on philosophically important questions (see Feltz, 2009; Nadelhoffer & Nahmias, 2007). Arguably the best known studies in experimental philosophy are Knobe's (2003a) harmful (underlined) and helpful (bracketed) chairman cases: The vice-president of the company went to the chairman of the board and said, “We are thinking of starting a new program. It will help us increase profits, but [and] it will also harm [help] the environment.” The chairman of the board answered, “I don't care at all about harming [helping] the environment. I just want to make as much profit as I can. Let's start the new program. They started the program. Sure enough, the environment was harmed [helped]. (191) The main difference between the two cases is the moral valence of the consequence of the chairman's decision. Remarkably, this shift in the moral valence of the consequence drastically changed people's intentionality judgments about the consequence: 82% of participants judged that the chairman brought about the harm to the environment intentionally whereas only 23% Perspective Intentional Action 4 judged the chairman brought about the help intentionally. This general effect (the side effecteffect or Knobe effect) has been replicated using similar scenarios (Cushman and Mele, 2008; Knobe, 2003a, 2003b, 2004a, 2004b) across cultures (Knobe and Burra, 2006), as well as across ages (Leslie, Knobe, and Cohen, 2006). Knobe-style cases feature side effects. If a consequence of an intended action is foreseen but not intended, then that consequences is a side effect of the intended action. Side effects have been considered important test cases of some theories of intentional action. Just to take one example, Knobe-style cases have been argued to challenge a prominent view in intentional action —the Simple View (SV). According to the SV, if one intentionally performs an action A then one intends to A. Some philosophers have argued that the SV is supported by folk intuitions (Adams, 1986; McCann 1986, 2005). However, as judgments in Harm suggest, sometimes the folk make judgments that are contrary to the SV. If the harm to the environment is a side effect, then it is not intended. But, most people think that the harm is brought about intentionally. Hence, in some circumstances, the folk have the intuition that one can harm the environment intentionally without intending to do so. This pattern of intuitions seemingly falsifies that the SV is supported by folk intuitions (Nadelhoffer, 2006; for alternative interpretations, see Adams & Steadman, 2004a, 2004b). In the next two sections, we suggest that folk intuitions surrounding intentional action may be much more complicated than originally thought and may be influenced by a variety of factors including one's perspective. Actor-observer differences Actor-observer differences refer to a common effect where people who engage in behaviors (actors) judge things differently than those who watch behaviors (observers). The traditional conception of the actor-observer asymmetry posits that the “actor's view of his behavior emphasizes the role of environmental conditions at the moment of action. The observer's view emphasizes the causal role of stable dispositional properties of the actor” (Jones & Nisbett, 1972, 80). While it is debatable whether this traditional conception is completely accurate (Malle, Knobe, & Nelson, 2007), some actor-observer asymmetries have been revealed 1 See Feltz, 2007b for a more detailed overview. 2 See Cushman & Mele, 2008 for a detailed definition of a side effect. Perspective Intentional Action 5 in decisions made in risky environments (Fernandez-Duque & Wifall, 2007), moral judgments (Nadelhoffer & Feltz, 2008), and action explanations (Malle & Knobe, 1997; Malle, Knobe, & Nelson, 2007). To illustrate, consider one case where an actor-observer asymmetry was found (Nadelhoffer & Feltz, 2008): A trolley is hurtling down the tracks. There are five workers on the track ahead of the trolley, and they will definitely be killed if the trolley continues going straight ahead since they won’t have enough time to get out of harm’s way. There is a spur of track leading off to the side where another person is working. The brakes of the trolley have failed and there is a switch which can be thrown to cause the trolley to go to the side track. Imagine that you are an innocent bystander who happens to be standing next to the switch. You realize that if you do nothing, five people will definitely die. On the other hand, you realize that if you throw the switch, you will definitely save the five workers. However, you are also aware that in doing so the worker on the side track will definitely be killed as the result of your actions. Observers received the same scenario except 'you' was replaced with 'John' (along with appropriate verb conjugations). Participants were asked if flipping the switch is morally permissible and rated how much control over the situation one has. People who were given the 'John' version were more likely than those given the 'you' version to judge that (a) flipping the switch was “morally permissible” and (b) John had control over the events. Hence, one's viewpoint influenced moral judgments in this example (Nadelhoffer & Feltz, 2008). But why do people sometimes display actor-observer asymmetries? According to Malle, Knobe, & Nelson (2007), one reason is that “we can expect that actors normally have better access to their own reasons than observers do and that they are normally more motivated to portray themselves as active, conscious, and rational agents” (508). Because actors may be motivated to (a) portray themselves in a positive light and (b) have special access to their own reasons, they are prone to judge their own behaviors differently from others' behaviors. This explanation could account for asymmetry in the Trolley example. Because actors are motivated 3 These scenarios are modified cases used by Petrinovich and O'Neil (1996), but Trolley cases are well known in the literature. 4 Ninety percent in the 'John' version thought it was permissible versus 65% in the 'you' version. The mean control rating in the 'John' version was 3.72 and 2.88 for the 'you' version (on a 7 point ascending scale). Perspective Intentional Action 6 to portray themselves in a positive light and flipping the switch results in the awful killing of a person, they are less likely to judge it permissible for themselves to flip the switch. However, because actors are relatively less interested in portraying others in a positive light, they judge that it is permissible for others to flip the switch. But actors who realize that flipping the switch is the optimal decision even if it kills a person may excuse themselves by judging they had little control over the situation. Given that there are actor-observer differences in a wide variety of related contexts, we thought that similar actor-observer differences would be found in judgments involving intentions and intentionality. To explore these possible differences, we used a new method in experimental philosophy borrowed from experimental economics. We had participants engage in a real decision making process with real rewards and penalties. Because participants actually became actors, we hypothesized this methodology would have the greatest chance of revealing actorobserver differences in intuitions about intentional action. Experiment 1a We constructed a decision making environment where participants could (a) engage in helpful or harmful behaviors and (b) observe others' helpful or harmful behaviors. We call Actors those who generate a behavior. We call Observers those who watch a behavior. In the Harm condition, an actor generates a harm to one other person. In the Help condition, an actor generates a benefit to one other person. We hypothesized that actors would judge behaviors as (a) less intended and (b) less intentional than when they judge behaviors as observers. Participants Participants (N = 45) were recruited via email at a southern university. Participants were tested in 6 groups consisting of no more than 12 participants and no fewer than 4. Participants received $10 for attending. They also had the opportunity to earn an additional $10 depending on their performance in the experiment (Range = $16-$20). Participants were told that they would be paid as a function how many Experimental Currency Units (ECUs) they earned in the experiment. The payoff function was not disclosed. Each participant was an actor and an observer (counterbalanced for order). However, each participant was in only one of the Help or the Harm conditions. Because we were interested 5 The expense of the experiment necessitated a small sample size. Perspective Intentional Action 7 in intuitions about actions, all participants who did not perform the desired action (contributing to Account A, see below) were excluded. Seven participants were thereby excluded in Harm. For the purposes of analyses, there were 20 participants in Help and 18 in Harm. Materials Participants completed the experiment on a computer programed using Z-Tree software (Fischbacher, 2007). Actors in the Harm condition were instructed to indicate how many of their 10 “tokens” they wished to invest in an account called 'Account A'. They were told that for every token they invested in Account A, they would earn 12 ECUs. For every token they did not invest in Account A, they would earn 10 ECUs. However, for every token invested in Account A, they would generate a 3 ECU penalty to one other person in the experiment. Actors in Help were given the same instructions as Actors in Harm but instead of generating a 3 ECU penalty, the actor generated a 3 ECU bonus by contributing to Account A. Observers in Harm read a display stating that somebody else had contributed 10 tokens to Account A generating a 30 ECU penalty to them. Observers in Help read a display stating that another participant contributed 10 tokens to Account A generating a 30 ECU bonus for them. There was one unpaid practice round followed by one paying round in each condition. After each instance of acting or observing, participants were asked to rate on a 7 point scale (1 = disagree, 7 = agree) their level of agreement with the appropriate version of each of the following statements: 1. You/the other participant intended to generate the penalty/bonus 2. You/the other participant intentionally generated the penalty/bonus 3. You/the other participant are/is blameworthy/praiseworthy for generating the penalty/bonus. Participants were also given the opportunity to explain their answers in a few sentences. So, each participant answered 3 actor questions and 3 observer questions in only one of Harm or Help conditions and had the opportunity to explain their answers in each condition. Results and Discussion The means and standard deviations are reported in Table 1. Table 1 Actor Intend M = 3.26 SD = 2.13 Perspective Intentional Action 8 Actor Intentional M = 3.5 SD = 2.20 Observer Intend M = 4.16, SD = 2.1 Observer Intentional M = 4.26, SD = 2.17 To test our hypothesis, a mixed-model Analysis of Variance (ANOVA) was performed with Harm/Help and observer order as between participants variables and answers to the Actor Intended/Intentional and Observer Intended/Intentional prompts as within participants variables. The predicted difference in actor/observer judgments was found for intention F (1, 34) = 6.01, p = .02, ηp = .15. Neither order (F (1, 34) = 1.82, p = .19, ηp = .05) nor condition (F < 1) interacted with judgments. A similar mixed-model ANOVA was conducted on intentionality judgments. The predicted actor/observer difference was found F (1, 34) = 5.55, p = .02, ηp = .14, and order did not interact with judgments (F < 1). Theoretically, there should be differences in people's Harm and Help judgments (Knobe, 2003a) and a moderately sized near-significant interaction for Harm/Help was observed for Intentional, F (1, 34) = 3.66, p = .06, ηp = .10. Because of this interaction, each condition (Harm or Help) was selected and four mixed-model ANOVAs were conducted with order as between participants factors and judgments about (1) Actor Intention/Observer Intention and (2) Actor Intentional/Observer Intentional as within participants factors. Means and standard deviations are reported in Table 2. Table 2 Harm Help Actor Intend M = 2.33, SD = 1.57 M = 4.1, SD = 2.25 Actor Intentional M = 2.67, SD = 2.11 M = 4.25, SD = 2.05 Observer Intend M = 3.67, SD = 2.2 M = 4.6, SD = 1.96 Observer Intentional M = 4.11, SD = 2.3 M = 4.4, SD = 2.09 In Harm, predicted differences were found for Actor Intention/Observer Intention F (1, 16) = 7.99, p = .01, ηp = .33. Order did not reliably interact with judgments F (1, 16) = 2.04, p = .17, ηp = .11. The predicted differences were also found for Actor Intentional/Observer Intentional judgments F (1, 16) = 12.62, p = .003, ηp = .44. Order did not interact with judgments (F < 1). Perspective Intentional Action 9 In Help, no actor-observer differences were detected (F's < 1). This experiment also allowed tests of some other interesting actor-observer differences. We thought that actors would display a reversed side effect-effect while observers would display the traditional side effect-effect. That is, actors would tend to judge their own harmful behavior as not intentional and their own helpful behavior as intentional. Observers would tend to think that harmful behaviors are intentional and helpful ones not. As side effects can occur when a behavior is judged intentional but not intended, we selected only those participants who did not judge the behavior in the relevant condition to be intended. After excluding those who did not intend the behavior (responding 4 or less), 16 participants remained in Harm and 10 remained in Help conditions. An ANOVA found the predicted shift in judgments in Harm that trended toward significance: Harm M = 2.31, SD = 1.89, Help M = 3.1, SD = 1.97, F (1, 26) = 2.94, p = .10, ηp = .12. However, order appeared to interact with judgments F (1, 26) = 2.94, p = .10, ηp = .12. To eliminate any possible order effect, only first responses were analyzed. After eliminating those who were in the actor condition second, did not contribute to Account A, and responded that they did not intend the bonus or penalty, a very large, significant difference was observed: Harm (N = 9, M = 2.33, SD = 1.73), Help (N = 3, M = 5.0, SD = 1.0) F (1, 10) = 6.15, p =.03, ηp = .38. We did not find the traditional side effect-effect for observers (F's < 1). Finally, previous research indicates that the Knobe effect is predictable by the global personality trait extraversion (Cokely & Feltz, 2009a). Extraversion is a member of the Big Five personality model and is represented in almost all modern personality models (John, 1999). Within the Big Five model, an extravert is defined as one who is a “communicative, sociable, energetic person who thrives on social contact and who does not regulate tightly his/her emotional reactions” (Akert & Panter, 1998, p. 966). Extraverts enjoy social interaction, find it rewarding, and actively seek out opportunities to be socially and emotionally engaged (Ashton, Lee, & Paunonen, 2002; Lucas & Fujita, 2000). Extraversion is one of the best understood and most thoroughly investigated personality traits in the Five Factor model of personality and is correlated with unique emotionally expressive behaviors, socially focused judgments and 6 Participants could not go back to the previous condition after they had entered their answers. 7 Because the small sample size and unequal cells are problematic, we are currently conducting follow up studies. The small sample size in Help was anticipated because it is unlikely that good behaviors would be judged unintended by actors (see Feltz, 2007a and Nadelhoffer, 2007 for a discussion). Perspective Intentional Action 10 memory retrieval processes (Akert & Panter, 1988; Chamorro-Premuzic, Furnham, & Ackerman, 2006; Lucas & Fujita, 2000; Rusting, 1999; Zelenski & Larsen, 2002). The current experiment allowed us to test for actor-observer differences in relation to extraversion. To this effect, participants also completed the Brief Big Five Inventory at the end of the experiment (Gosling, Rentfrow, & Swan, 2003). Excluding those who did not contribute and did not intend the harm/help, extraversion was negatively correlated with actors' intentional harm judgments, r (N = 16) = -.55, p = .03, and was not significantly correlated with actors' intentional help judgments r (N = 10) = .40, p = .25. To illustrate the relation of extraversion with the reversal of the Knobe effect, a rough median split was constructed where all those who scored higher than 9 were classified as extraverts and those scoring 9 or below were classified as introverts. The predicted interaction between extraversion and the reversal of the Knobe asymmetry was found F (1, 22) = 6.43, p = .02, ηp = .23 (See Figure 1). Figure 1: Extraversion Median Intentionally Means Experiment 1b One worry with Experiment 1a was that observers were not "true" observers. There are at least two ways that not being a true observer could have influenced participants' judgments. First, during the experiment, some were first harmed or helped by another person in the experiment. That is, they either received a penalty or a bonus depending on the other participant's Perspective Intentional Action 11 action. They may have reacted to this element of the experiment in unintended and unknown ways. Second, observers in Experiment 1a were also actors. Because participants were also Actors, the decisions they made as actors may have had some effect on their judgments as Observers. Indeed, Experiment 1a indicated that there were order effects on some judgments. Hence, a different pattern of results may emerge if participants were true observers who were neither affected by others' decisions nor made those decisions themselves. To test for this possible difference, a follow-up study was conducted. Participants Eighty-four participants were recruited from Amazon's Mechanical Turk and were redirected to a survey hosted on Surveymonkey.com. After completing the survey, participants received a small cash award ($0.15). One participant did not indicate sex and two did not indicate their age. Of the remaining, 30 (36%) were male and 53 (64%) were female. The mean age was 36.66, SD = 13.39 ranging from 18-81. Materials Participants read a scenario that described the decisions that participants made in Experiment 1a. Participants were thereby pure Observers because nobody's decision influenced their reward and none of the participants made decisions as Actors. Participants received one of the following scenarios: Economic Help/Harm The following actually happened a couple of weeks ago. Two people were in an experiment where they received money based on how many Experimental Currency Units (ECUs) they acquired:In the experiment, Person 1 had to decide what to do with 10 tokens. Person 1 could either invest some tokens into Account A or do nothing. If Person 1 did nothing, Person 1 would receive 10 ECUs for each token not invested and nothing would happen to Person 2. For any token Person 1 invested in Account A, Person 1 would receive 12 ECUs. However, Person 2 would receive a 3 ECU penalty/bonus for each token Person 1 invested in Account A. Person 1 decided to invest 10 tokens into Account A thereby receiving 120 ECUs. As a result, Person 2 received a 30 ECU penalty/bonus. 8 Evidence indicates that recruiting participants from Amazon's Mechanical Turk is no worse than typical recruitment strategies (e.g., from Universities' Subjects Pool), and in some cases may actually be superior (Buhrmester, Kwang, & Gosling, in press). Perspective Intentional Action 12 Remember, the situation just described actually happened a few weeks ago. Please rate your agreement with the following statements on the scale provided. After reading one of these scenarios, participants answered the following two questions: 1. Person 1 intended to generate the 30 ECU penalty/bonus for Person 2. 2. Person 1 intentionally generated the 30 ECU penalty/bonus for Person 2. Participants responded using a 7 point scale where 1 = strongly disagree, 7 = strongly agree, and 4 = neutral. Participants also answered the following comprehension question: 3. In the experiment, what could Person 1 decide to do? All who failed the comprehension question were excluded from analyses. Results and Discussion Means and standard deviations are reported in Table 3. Table 3 Harm (N = 43) Help (N = 41) Intend M = 3.56, SD = 1.76 M = 4.15, SD = 1.78 Intentional M = 4.14, SD = 1.81 M = 4.32, SD = 1.72 ANOVAs were conducted comparing the Actor results in Experiment 1a versus the results of the Observer cases in Experiment 1b. The same pattern of results were obtained as in Experiment 1a. In Harm, there were significant actor-observer differences for Intend F (1,66) = 9.41, p = .003, ηp = .13 and Intentionally: F (1, 66) = 12.19, p = .001, ηp = .16. Importantly, there was a qualitative shift in judgments for Intentionally where Actors in Experiment 1a judged that they did not bring about the harm intentionally and Observers in Experiment 1b did not think the harmful behavior was intentional. No actor-observer differences were found in Help for Intend or Intentional (F's < 1). Oddly, there was no asymmetry in intentional judgments typical in Knobe style cases (F < 1). Experiment 1b addressed a possible worry with Experiment 1a that observers were somehow illegitimately influenced by other participants' decisions or by performing the action themselves. Observers in Experiment 1b displayed the same pattern of responses as Observers in Experiment 1a. At least in the present experiments, it did not appear that being influenced by another decision or making a similar decision as an actor played a large role in observers' Perspective Intentional Action 13 intention and intentionality judgments. Experiment 1c It is unclear why in Experiments 1a and 1b we did not find the traditional side-effect effect for observers. One difference between the original Knobe cases and those used in Experiment 1b is that the scenario in Experiment 1b did not expressly state that the participant did not care about the consequence of the behavior as Knobe's original chairman cases did. Presumably, participants assume that actors had some positive attitude toward the helpful consequence, thereby inflating their Intention and Intentionality judgments. To test this hypothesis, we conducted a follow up study where we explicitly say that the actor did not care about the consequence of contributing to Account A. Given the actor's declaration, the traditional side-effect effect should be present. Participants One hundred and five participants were recruited from Amazon's Mechanical Turk and redirected to a survey hosted on Surveymonkey.com. Ten participants were excluded because they indicated that their first language was not English. Of the remaining participants, 28 (30%) were male and 67 (70%) were female. The mean age was 34.12, SD = 12.44 ranging from 18-69. After completing the survey, participants received a small cash award ($0.20). Materials Participants were given exactly the same scenario as participants in Experiment 1b. However, there was one additional sentence added to the Help and Harm versions. Before the second to last sentence, the appropriate version of the following sentence was inserted: In a written comment, Person 1 stated “I didn't care at all about generating the 30 ECU penalty/bonus. All I wanted to do was maximize my profits.” Participants received only one scenario. Participants answered the same two questions participants in Experiment 1b answered. Results and Discussion Means and standard deviations are reported in Table 4. Table 4 Harm (N = 48) Help (N = 47) Intend M = 3.29, SD = 2.34 M = 1.66, SD = 1.37 9 We would like to thank Joshua Knobe for bringing this to our attention. Perspective Intentional Action 14 Intentional M = 3.75, SD = 2.36 M = 2.34, SD = 1.74 ANOVAs found the predicted differences between the two cases: Intend F (1, 93) = 17.08, p < . 001, ηp = .16; Intentionally F (1, 93) = 10.98, p = .001, ηp = .11. As in Experiment 1b, treating participants in Experiment 1c as Observers revealed the predicted actor-observer differences for Harm Intend F (1, 71) = 4.19, p = .04, ηp = .06 and Intentional F (1, 71) = 5.40, p = .02, ηp = . 07. Additionally, there were actor-observer differences for Help Intend F (1, 65) = 29.78, p < . 001, ηp = .31 and Intentionally F (1, 65) = 15.22, p < .001, ηp = .19. The results of Experiment 1c indicated that the relatively high observers' judgments for the helpful behavior in Experiments 1a and 1b may have been the result of some observers assuming that actors who bring about the helpful consequence have favorable attitudes toward bringing about that consequence. These results help clarify a debate in the intentional action literature about folk intuitions about helpful behaviors. Under normal conditions, people may be inclined to think that if an actor knows that a behavior will have a beneficial consequence, that consequence becomes part of the actor's reason for acting (Nadelhoffer, 2006, 2007). However, when one explicitly states not having a desire to bring about the known, helpful consequence, it is clear that consequence does not even form part of the person's reason for acting. This could partially explain why Help judgments in Experiment 1a and 1b were inflated compared to traditional Knobe-style cases—some participants may have thought bringing about the beneficial consequence forms part of the actor's reason for acting. However, the data from Experiments 1a and 1b also suggest that not all people thought this way. Some people did not judge that the beneficial consequence formed part of the actor's reason for acting. Hence, not all people judge known, beneficial consequences form part of a person's reason for acting (Feltz, 2007a). Experiment 2 Experiment 1a suggested that providing the right environment facilitates Actor-Observer differences in judgments about intentions and intentionality. However, a question remains whether Actor-Observer differences can occur in pencil-and-paper surveys. If Actor-Observer differences occur in pencil-and-paper surveys, the Actor-Observer asymmetry observed in Experiment 1a is not necessarily encouraged by the testing environment. Rather, the difference would be a more general phenomenon. Experiment 2 was designed to address the worry that Perspective Intentional Action 15 Actor-Observer differences could be found when participants are merely asked to imagine that they are the chairman. Participants One hundred and one participants were recruited from Amazon's Mechanical Turk and were redirected to complete the surveys at SurveyMonkey.com. After completing the survey, participants received a small reward ($0.15). Participants were excluded if they reported that their first language was not English or if they failed the comprehension question. After excluding these participants, 95 remained. Materials There were four different scenarios: 1. Harm Observer, 2. Harm Actor, 3. Help Observer, 4. Help Actor. The following were the Help and Harm cases in the Actor condition: Imagine that you are the chairman of the board. The vice-president of a company comes to you and says, “We are thinking of starting a new program. It will help us increase profits for this year’s balance sheet, but in ten years it will start to [harm/help] the environment.” Imagine that you answered, “I don’t care at all about [harming/helping] the environment. I just want to make as much profit for this year’s balance sheet as I can. Let’s start the new program.” The program was started. Sure enough, ten years later, the environment started to be [harmed/helped]. Immediately following the scenario, participants were asked to rate their level of agreement with the following statements (1 = strongly disagree, 4 = neutral, 7 = strongly agree): 1. You intended to [harm/help] the environment. 2. You intentionally [harmed/helped] the environment. 3. You are [blameworthy/praiseworthy] for [harming/helping] the environment. Participants were also asked the following comprehension question: 4. How long did it take before the [harm/help] began? The following were the Help and Harm scenarios in the Observer condition: The vice-president of a company went to the chairman of the board and said, “We are thinking of starting a new program. It will help us increase profits for this year’s balance sheet, but in ten years it will start to [harm/help] the environment.” The chairman answered, “I don’t care at all about [harming/helping] the environment. I just want to Perspective Intentional Action 16 make as much profit for this year’s balance sheet as I can. Let’s start the new program.” They started the new program. Sure enough, ten years later, the environment started to be [harmed/helped]. Immediately following the scenario, participants were asked to rate their level of agreement with the following sentences (1 = strongly disagree, 4 = neutral, 7 = strongly agree): 1. The chairman intended to [harm/help] the environment. 2. The chairman intentionally [harmed/helped] the environment. 3. The chairman is [blameworthy/praiseworthy] for [harming/helping] the environment. Participants were also asked the following comprehension question: 4. How long did it take before the [harm/help] began? Each participant was an Actor and an Observer in only one of the Harm or Help conditions. The order of presentation was counterbalanced. Once participants completed their responses to one condition, they could not go back and change their answers. 4.3 Results and Discussion Means and standard deviations are reported in Table 5. Table 5 Harm (N = 46) Help (N = 49) Actor Intend M = 4.94, SD = 1.65 M = 1.14, SD = 0.54 Actor Intentional M = 5.83, SD = 1.36 M = 1.35, SD = 0.88 Observer Intend M = 4.94, SD = 1.65 M = 1.96, SD = 1.94 Observer Intentional M = 5.83, SD = 1.24 M = 2.04, SD = 1.99 ANOVAs found a large Knobe-like effect for Actor Intention judgments F (1, 93) = 231.92, p < . 001, ηp = .72. Order did not reliably interact with judgments (p > .22). There was also a Knobelike effect for Observer judgments about intentions F (1, 93) = 54.54, p < .001, ηp = .41. Order did not interact with judgments (F < 1). Similar effects were found for judgments about Actor intentional F (1, 93) = 369.8, p < .001, ηp = .80 and Observer intentional F (1, 93) = 122.29, p < .001, ηp = .57. Order did not reliably interact with judgments (p > .09). A mixed-model ANOVA with Actor-Observer judgments as within-participants factors and order as between participants factor did not reveal a reliable Actor-Observer asymmetry for Actor-Observer Perspective Intentional Action 17 judgments in Harm (F's < 1). A significant Actor-Observer difference was found for Intention Help F (1, 47) = 9.77, p = .003, ηp = .17 and Intentional Help F (1, 47) = 8.46, p = .006, ηp = . 15. The results of Experiment 2 suggest that the Actor-Observer asymmetry produced in Experiment 1a is not likely to be found in pencil-and-paper surveys encouraging participants to imagine they are the chairman. This reinforces the notion that only when the relevant conditions are made salient to the participants do Actor-Observer asymmetries emerge in intention and intentionality judgments. General Discussion Consistent with and extending previous research, Experiments 1a-c suggested that in some circumstances people judge their own behaviors differently than they judge the identical behavior of others. Experiment 2 suggested this result was not likely to be found using the standard survey methodology in experimental philosophy. In addition, our evidence suggests that a well-known result in experimental philosophy—the side-effect effect—can be reversed. Finally, replicating previous work (Cokely & Feltz, 2009a), extraversion was systemically and predictably related to this reversed asymmetry. These results provide further evidence that impression management can play a key role in people's intention and intentionality judgments (Malle, Knobe, & Nelson, 2007). An important clue for this interpretation comes from the results of the Harm case. Participants were much less likely to judge that they intended the harm or intentionally brought about the harm compared to their judgments as observers. Presumably, participants did not want to be a “bad guy” by bringing about the bad consequence whereas they were relatively less interested in managing their impression of others. Hence, they were more motivated to respond that they did not intend or intentionally bring about the harm. In addition, extraverts were more likely to display this 10 Replicating previous work (Cokely & Feltz, 2009a), after removing those who received Actor first we found a near significant correlation of extraversion with observer judgments in Harm, r (23) = .39, p = .065. We removed the observer first condition to more accurately re-create Cokely and Feltz's design. 11 In Help, those who received Actor first had higher ratings in the Observer condition (Intention M = 2.68, SD = 2.44, Intentional M = 2.27, SD = 2.49) than those who received Actor second (Intention M = 1.21, SD = 0.66, Intentional M = 1.33, SD = 0.87). These results were statistically significant: intention F (1, 47) = 8.78, p = .005, ηp = .16 and intentionality F (1, 47) = 8.46, p = .006, ηp = .15. While this was a statistically significant difference, all of the mean responses were still moderately disagreeing. Unlike Experiment 1a, overall Help judgments were qualitatively the same. 12 Some of the written, open-ended responses provide some evidence for this interpretation. For example, one Actor Perspective Intentional Action 18 pattern in Harm. Because extraverts are socially minded individuals, they would be relatively more concerned with possible social aspects of their behavior. However, because the behavior in Help is beneficial, there is less motivation to mitigate possibly negative implications of that behavior. So in Help, the responses between actors and observers would be more similar. These data also provide some important insights into the side effect-effect. We found strong correlations between people's intention and intentionality judgments. Those who favor the SV may take these as supporting data. However, defenders of the SV should be cautious for two reasons. First, correlation indicates that there is some relation between intention and intentionality judgments. These correlations do not necessarily indicate that an intention to A is a necessary condition for A-ing intentionally. These results are equally consistent with intending to A is a sufficient condition for A-ing intentionally (when one A's)—a condition that most theories of intentional action would endorse under normal conditions (e.g., no causal deviance). Second, we have some evidence that a new but equally problematic side effect-effect exists. For actors who did not think they intended to bring about the penalty or bonus, the moral valence of the consequence influenced their intentionality judgments. Specifically, actors were more likely to judge they brought about the beneficial consequence intentionally than the harmful consequence. These results suggest that at least some folk do not treat an intention to A as necessary for A-ing intentionally as the SV holds. Third, our results reinforce the importance of individual differences in judgments about intentions and intentionality and provide more evidence that philosophically relevant intuitions are systematically fragmented (Feltz, & Cokely, 2008). Extraverts were more likely to display the reversed Knobe effect. Importantly, we were able to predict a priori who were likely to make those judgments. If there are predictable and systematic differences in intuitions regarding intentionality, then perhaps there is not a single folk concept of intentional action, but several. That there are individual differences in intentional action intuitions is important for several reasons. As already mentioned, individual differences in intentional action intuitions may in harm wrote, “I was not trying to generate a penalty to the rest of the group, I was entering money to further my own investment and income.” One Observer in harm wrote, “The other person knew going into this that whatever they were going to invest would have a penalty on me. You can blame them because they had the knowledge. They may not have intended to cause penalty, but the fact is they did.” These responses are consistent with other written reports about intentional action (Nichols & Ulatowski, 2001). Perspective Intentional Action 19 indicate that there is more than one folk concept of intentional action. If that is so, then any account of intentional action that respects the folk concept of intentional action must take into account the diversity of folk intuitions or provide some explanation why folk intuitions contrary to that account are wrong (e.g., error theory) (see Nahmias, Morris, Nadelhoffer, and Turner, 2006; Vargas, 2005 for a similar argument in the free will debate). In addition, it is widely accepted that in most cases one cannot make inferences about proximal judgments processes based on mean data when there are individual differences (Newell, 1973; Estes, 1956; Molenaar & Campbell, 2009). There is gathering evidence that there are individual differences in the Knobe effect (Cokely & Feltz, 2009a) and in the present studies concerning the reversal of the Knobe effect. There have been a host of models that have attempted to explain the Knobe effect (e.g., Knobe, in press, 2006b; Malle & Guglielmo, 2006; Nadelhoffer, 2006b; Phelan & Sarkissian, in press). Our data put pressure on these models because all of the models are based on mean responses and do not take into account individual differences. These individual differences provide important and necessary opportunities to refine these models to make them more accurate and precise (Cokely & Feltz, 2009b). Finally, work in experimental philosophy has been dominated by the use of surveys. These surveys have been helpful and have uncovered a number of interesting phenomena. However, these methods can be limited. As the current series of studies illustrate, pencil and paper surveys are not likely to uncover actor-observer asymmetries in intentional action attributions. There are likely to be a number of philosophically relevant topics yet to be explored by the methods of behavioral economics. For example, the exploration of virtuous decisions making or notions of blame and praise could be well-suited for these alternative methods. Using these alternative methods, along with individual differences approaches, can help get closer at the proximal judgment processes involved in philosophically relevant intuitions (Cronbach, 1957). Hence, moving beyond pencil and paper surveys offers a great opportunity for theoretical and practical progress in experimental philosophy. Conclusion Actors tend to judge their own helpful or harmful behaviors differently from the same behaviors they observe in others. This generally well-known result has been applied to intentional action ascriptions using a new method. Using this new method provided evidence that Perspective Intentional Action 20 people ascribe intentions and intentionality differently depending on their perspective. In some instances, this effect is so pronounced that a well-known result in experimental philosophy—the side-effect effect—is reversed. Moreover, we have extended and provided further evidence that these intuitions are fragmented and predictable by global personality traits. Our results not only open up and new avenues for experimental philosophy, they also provide evidence for the necessity of taking these new avenues. Perspective Intentional Action 21 Authors' Note We would like to thank Shaun Nichols, Joshua Knobe, Ron Mallon, Jennifer Wright, Mark Phelan, Mark Isaac, and participants at the Mid-South Philosophy conference, NEH Summer Institute for Experimental Philosophy, and the 32 Annual Conference of the Cognitive Science Society for very helpful comments. Perspective Intentional Action 22 ReferencesAdams, F. (1986). Intention and intentional action: The Simple View. Mind and Language, 1,281-301.Adams, F., & Steadman, A. (2004a). Intentional Action in Ordinary Language: Core Concept orPragmatic Understanding. Analysis, 64, 173-181.Adams, F., & Steadman, A. (2004b). Intentional Actions and Moral Considerations: StillPragmatic. Analysis, 64, 264-267.Akert, R., & Panter, A. (1988). Extraversion and the ability to decode nonverbal communication.Personality and Individual Differences, 9, 965-972.Ashton, M., Lee, K., & Paunonen, S. (2002). What is the Central Feature of Extraversion? SocialAttention Versus Reward Sensitivity. Journal of Personality and Social Psychology,83, 245-252.Buhrmester, M., Kwang, T., & Gosling, S. D. (in press). Amazon's Mechanical Turk: A newsource of inexpensive, yet high-quality, data? Perspectives on Psychological Science.Chamorro-Premuzic, T., Furnham, A., & Ackerman, P. (2006). Ability and personality correlatesof general knowledge. Personality and Individual Differences, 41, 419-429.Cokely, E. T., & Feltz, A. (2009a). Individual differences, judgment biases, and Theory-of-Mind:Deconstructing the intentional action side effect asymmetry. Journal of Research inPersonality, 43, 18-24.Cokely, E.T. & Feltz, A. (2009b). Adaptive variation in folk judgment and philosophicalintuition. Consciousness and Cognition, 18, 355-357.Cushman, F., & Mele, A. (2008). Intentional action: Two-and-a-half folk concepts? In J. Knobe& S. Nichols (Eds.), Experimental Philosophy (pp. 171-188). Oxford: Oxford UniversityPress.Cronbach, L. J. (1957). The two disciplines of scientific psychology. American Psychologist, 12,671-684.Estes, W. K. (1956). The problem of inference from curves based on group data. PsychologicalBulletin, 53, 134-140.Feltz, A (2007a). Knowledge, moral praise, and moral side effects. Journal of Theoretical andPhilosophical Psychology, 27, 123-126. Perspective Intentional Action 23 Feltz, A. (2007b). The Knobe effect: A brief overview. Journal of Mind and Behavior, 28, 265-277.Feltz, A. (2009). Experimental philosophy. Analyse & Kritik, 31, 201-219.Feltz, A. & Cokely, E.T. (2008). The fragmented folk: More evidence of stable individualdifferences in moral judgments and folk intuitions. In B. C. Love, K. McRae & V. M.Sloutsky (Eds.), Proceedings of the 30th Annual Conference of the Cognitive ScienceSociety (pp. 1771-1776). Austin, TX: Cognitive Science Society.Feltz, A. & Cokely, E.T. (2009). Do judgments about freedom and responsibility depend onwho you are?: Personality differences intuitions about compatibilism andincompatibilism. Consciousness and Cognition, 24, 342-350.Fernandez-Duque, D., & Wifall, T. (2007). Actor/observer asymmetry in risky decision making.Judgment and Decision Making, 2, 1-8.Fischbacher, U. (2008). z-Tree: Zurich Toolbox for Ready-made Economic Experiments.Experimental Economics, 10, 171-178.Gosling, S. D., Rentfrow, P. J., & Swann, W. B. (2003). A very brief measure of the Big-Fivepersonality domains. Journal of Research in Personality, 37, 504-528.John, O. (1999). The Big-five Trait Taxonomy: History, Measurement, and TheoreticalPerspectives. In L. Pervin & O. P. John (Eds.), Handbook of Personality: Theory andResearch (2nd ed., pp. 66-100). New York: Guilford.Jones, E., & Nisbett, R. (1972). The actor and the observer: Divergent perceptions of the causesof behavior. In E. Jones, D. Kanouse, H. Kelly, R. Nisbett, S. Vallins & B. Weiner (Eds.),Attribution: Perceiving the Causes of Behavior (pp. 79-94). Morristown, NJ: GeneralLearning Press.Knobe, J. (2003a). Intentional action and side-effects in ordinary language. Analysis, 63, 190-194.Knobe, J. (2003b). Intentional action in folk psychology: An experimental investigation.Philosophical Psychology, 16, 309-323.Knobe, J. (2004a). Folk psychology and folk morality: Response to critics. Journal ofTheoretical and Philosophical Psychology, 24, 270-279.Knobe, J. (2004b). Intention, intentional action and moral Considerations. Analysis, 64, 181-187. Perspective Intentional Action 24 Knobe, J., & Burra, A. (2006a). Experimental philosophy and folk concepts: Methodologicalconsiderations. Journal of Cognition and Culture, 6, 331-342.Knobe, J. (2006b). The concept of intentional action: A case study in uses of folk psychology.Philosophical Studies, 130, 203-231.Knobe, J. (in press). Action trees and moral judgment. Topics in Cognitive Science.Leslie, A., Knobe, J., & Cohen, A. (2006). Acting intentionally and the side-effect effect:Theory of mind and moral judgment. Psychological Science, 17, 421-427.Lucas, R., & Fujita, F. (2000). Factors Influencing the Relation Between Extraversion andPleasant Affect. Journal of Personality and Social Psychology, 79, 1039-1056.Malle, B., & Guglielmo, S. J. (2006). Directions and Challenges in Studying Folk Concepts andFolk Judgments. Journal of Cognition and Culture, 6, 321-329.Malle, B., & Knobe, J. (1997). Which behaviors do people explain? A basic actor-observerasymmetry. Journal of Personality and Social Psychology, 72, 288-304.Malle, B., Knobe, J., & Nelson, S. (2007). Actor-observer asymmetries in explanations ofbehavior: New answers to an old question. Journal of Personality and Social Psychology,93, 491-514.McCann, H. (1986). Rationality and the range of intention. Midwest Studies in Philosophy, 10,191-211.McCann, H. (2005). Intentional action and intending: Recent empirical studies. PhilosophicalPsychology, 18, 737-748.Mele, A. (1992). Recent work on intentional action. American Philosophical Quarterly, 29,199-217.Mele, A. (2001). Acting intentionally: Probing folk intuitions. In B. F. Malle, L. J. Moses & D.A. Baldwin (Eds.), Intentions and Intentionality (pp. 27-43). Cambridge: MIT Press.Molenaar, P. C., & Campbell, C. G. (2009). The new person-specific paradigm in pscyhology.Current Directions in Psychological Science, 18, 112-117.Nadelhoffer, T. (2006). Desire, Foresight, Intentions, and Intentional Action: Probing FolkIntuitions. Journal of Cognition and Culture, 6(1-2), 133-157.Nadelhoffer, T. (2007). Fringe benefits, side effects, and intentional actions: A reply to Feltz. TheJournal of Theoretical and Philosophical Psychology, 27, 801-809. Perspective Intentional Action 25 Nadelhoffer, T. (2006a). On trying to save the simple view. Mind and Language, 21, 565-586.Nadelhoffer, T. (2006b). Bad acts, blameworthy agents, and intentional actions: Some problemsfor juror impartiality. Philosophical Explorations, 9, 203-220.Nadelhoffer, T., & Nahmias, E. (2007). The Past and Future of Experimental Philosophy.Philosophical Explorations, 10, 123-149.Nadelhoffer, T. & Feltz, A. (2008). The actor-observer bias and moral intuitions: Adding fuelto Sinnott-Armstrong's fire. Neuroethics, 1, 133-144.Nahmias, E., Morris, S., Nadelhoffer, T., & Turner, J. (2006). Is incompatibilism intuitive?Philosophy and Phenomenological Research, 73, 28-53.Newell, A. (1973). You can't play 20 questions with nature and win: Protective comments on thepapers of this symposium. In W. G. Chase (Ed.), Visual Information Processing (pp.283-308). New York: Academic Press.Nichols, S., & Ulatowski, J. (2007). Intuitions and Individual Differences: The Knobe EffectRevisited. Mind & Language, 22, 346-365.Petrinovich, L., & O'Neill, P. (1996). Influence of wording and framing effects on moralintuitions. Ethology and Sociobiology, 17, 145-171.Phelan, M., & Sarkissian, H. (in press). The Folk Strike Back: Or, Why You Didn't Do itIntentionally, Though It Was Bad and You Knew It. Philosophical Studies.Rusting, C. (1999). Interactive effects of personality and mood on emotion-congruent memoryand judgment. Journal of Personality and Social Psychology, 77, 1073-1086.Vargas, M. (2005). The revisionist's guide to responsibility. Philosophical Studies, 125, 241-252.Zelenski, J., & Larsen, R. (2002). Predicting the future: How affect-related personality traitsinfluence likelihood judgments of future events. Personality and Social PsychologyBulletin, 28, 1000-1010.