![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
Abstract
In the past decade, a number of empirical researchers have suggested that laypeople have compatibilist intuitions. In a recent paper, Feltz and Millan (Citation2015) have challenged this conclusion by claiming that most laypeople are only compatibilists in appearance and are in fact willing to attribute free will to people no matter what. As evidence for this claim, they have shown that an important proportion of laypeople still attribute free will to agents in fatalistic universes. In this paper, we first argue that Feltz and Millan’s error-theory rests on a conceptual confusion: it is perfectly acceptable for a certain brand of compatibilist to judge free will and fatalism to be compatible, as long as fatalism does not prevent agents from being the source of their actions. We then present the results of two studies showing that laypeople’s intuitions are best understood as following a certain brand of source compatibilism rather than a “free-will-no-matter-what” strategy.
Acknowledgements
This research was supported by the National Center of Competence in Research (NCCR) Affective sciences financed by the Swiss National Science Foundation (number 51NF40-104897) and hosted by the University of Geneva.
Notes
1. The notion of “free will no matter what” intuitions originates in Feltz, Cokely, and Nadelhoffer (Citation2009). There the focus is on participants who have entrenched commitments to worldviews that make it difficult for them to imagine or understand a description of a deterministic universe (Feltz, Cokely, & Nadelhoffer, Citation2009, pp.16–17). In this paper, however, we follow Feltz and Millan in classifying as having “free will no matter what” intuitions those participants who “would judge that an agent is free and morally responsible even in fatalistic scenarios” (Citation2015, p. 534) or in any other case in which it would be obvious even to the most die-hard compatibilist that one cannot be free and morally responsible.
2. One might raise a number of issues about this way of defining fatalism. Because we will be following Feltz and Millan in their use of “fatalism”, we take time to address some of these issues and defend our usage in Section 3.2.
3. One possible issue, which we won’t deal with here, is the fact that both Extreme Book and Fries are likely not interpreted by participants as concerning the actual world, but rather present alternative possible worlds. This might have an effect on the results of studies using these scenarios, since there is evidence that participants’ judgments about whether protagonists have free will are sensitive to whether the protagonist’s universe is presented as being the actual universe (Roskies & Nichols, Citation2008). It would be interesting to probe this further in the context of “free will no matter what intuitions.” However, we suspect this would prove difficult as participants would likely have difficulty imagining that the actual world is a fatalistic one—let alone one involving magic books.
4. Thanks to an anonymous referee for pushing us on this point.
5. Here we are not concerned with Frankfurt’s actual view concerning free will and responsibility, but rather the effect that so-called “Frankfurt cases” play in the literature.
6. We take this example from Fischer and Ravizza (Citation1991) as we think it nicely illustrates the distinction. Though Fischer and Ravizza ultimately settle for a form of semi-compatibilism, for which only moral responsibility—but not free will—is compatible with the lack of alternative possibilities, it is still possible to refuse to dissociate free will and moral responsibility and use their argument in favor of a more classical form of compatibilism. For a review of data suggesting that people do not dissociate free will and moral responsibility, see Cova and Kitano (Citation2014).
7. This doesn’t include 14 participants who left the survey without answering a single question.
8. These participants responded to only one of the two scenarios.
9. As Feltz and Millan note, high rates of comprehension failures are the norm in experiments on free will. They draw attention to Sommers (Citation2010) who “notes that people routinely exclude 10–30% of participants in these types of studies” (Feltz & Millan, Citation2015, p. 552, note 8). We excluded around 29% of participants who gave complete answers. These exclusions are necessary to ensure participants understand Extreme Book and Fries adequately. The standard comprehension questions used were (i) “If the universe were re-created with the special book having the same true sentences, would John do the same thing?” (Extreme Book) and (ii) “If Universe A was exactly recreated, is it accurate to say that Bill would do the same thing?” (Fries). Participants were encouraged to take their time on the survey. They were told their answers would not be approved (in Mechanical Turk) if they spent less than three minutes on the survey.
10. Though we used the same texts as Feltz and Millan, two minor changes were made: (i) there was someone called John in both scenarios and, as this could be confusing in a within subjects design so, in Friesm John became Frank; (ii) the opening of our Extreme Book asked participants to “Imagine that there is a special book ….”
11. For a discussion of this scale’s validity, see Björnsson and Pereboom (Citation2014). One issue is that item seven is potentially ambiguous. It could be read as asking about whether the protagonist has actual control or whether the protagonist has regulative control (see Section 3.2). This obviously bears on the extent to which the scale measures attitudes toward bypassing, and we are sympathetic to Feltz and Millan’s suggestion that use of methods such as protocol analysis may be needed to understand the precise nature of participants’ intuitions in this area.
12. We were not personally interested in individual differences. However, Feltz and Millan gave the TIPI to their participants, and we followed them to stay as close to their original experimental design as possible.
13. One item (3 in Extreme Book) showed a significant correlation with age r(87) = .3, p = .004. There were no gender effects. We looked at the relation between personality and FWMR scores. We used the personality traits Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness to Experiences (ascertained using the Ten Item Personality Inventory) and BYPASS scores to predict FWMR scores in linear regression. The full model was a significant predictor in both Fries (F(6,82) = 13.052, p < .0005, R2 = .488) and Extreme Book (F(6,82) = 14.264, p < .0005, R2 = .511). Controlling for BYPASS and the other personality traits, Extraversion continued to predict FWMR in Fries (β = .217, t = 2.395, p = .019) (it did not for Extreme Book). This fits with Feltz and Millan’ findings, and with the literature more widely.
14. Note that this way of looking at the data ignores any within-subjects effects.
15. A referee for this journal notes that the wording of the case used, More Extreme Book, might also tap so-called “deep-self” intuitions in its attempt to articulate strongly to participants the idea that the agent’s mental states are bypassed because it includes the expression “deep down, John did not want to kill his wife.” This might indeed be relevant and worth investigating further.
16. One further participant entered the survey but didn’t consent and progressed no further.
17. We excluded those who didn’t respond to each of the FWMR and BYPASS items.
18. Neither FWMR nor BYPASS showed gender effects or significant correlations with age or any of the five personality traits measured.
19. There are reasons to be cautious about generalizing from participants recruited through Mechanical Turk. However, it has been observed that Mechanical Turk participants are slightly more demographically diverse than standard internet samples, are significantly more diverse than typical American college samples, and that data obtained through Mechanical Turk are at least as reliable as those obtained via traditional methods (Buhrmester, Kwang, & Gosling, Citation2011). Moreover, study 1 replicated the results of Feltz and Millan’s studies, showing that the results obtained through this kind of study can be reliably extended to other populations. Finally, it should be noted that intuitions about free will and moral responsibility seem to be robust and to display very little variation across cultures (Sarkissian et al., Citation2010).