Telling more than we can know about intentional action

Hi folks,

I wanted to draw your attention to our new manuscript that reports on a study of Joshua Knobe's famous 'Chairman case'. I conducted the study with my colleague Sara Konrath, and we used a method called structural path analysis to uncover some pretty surprising things about people's intuitions in the Chairman case. In short, we argue that normative factors do not drive asymmetric intuitions in the case, though people think they do. The methods used in the study are relatively new in X-Phi, and we hope that people will be receptive to our approach.  We would absolutely love comments and criticisms.

The paper is here.  And here is the abstract reprinted below.

Thanks,
Chandra

---------------------------------------

Abstract

Recently, a number of philosophers have advanced a surprising conclusion – people’s judgments about whether an agent brought about an outcome intentionally are pervasively influenced by normative considerations. In this paper, we investigate the ‘Chairman case’, an influential case from this literature and disagree with this conclusion. Using a statistical method called structural path modeling, we show that people’s attributions of intentional action to an agent are driven not by normative assessments, but rather by attributions of underlying values and characterological dispositions to the agent. In a second study, we examined people’s judgments about what they think drives asymmetric intuitions in the Chairman case and found that people are highly inaccurate in identifying which features of the case their intuitions track. In the final part of the paper, we discuss how the statistical methods used in this study can help philosophers with the tracking problem, the problem of figuring out which features of hypothetical cases our intuitions are responsive to. We show how the methods used in this study have some advantages over both armchair methods used by traditional philosophers and survey methods used by experimental philosophers.  

Comments

[Eddy Nahmias]

This is a really outstanding paper. I highly recommend that people interested in experimental philosophy read it. It offers some very interesting insights into the Knobe effect, but more importantly, it explains statistical methods and path modeling that will be very useful to experimental philosophers. Chandra and Sara explain how such methods can illuminate the psychological sources of people's intuitions and judgments (in a way "more primitive" methods cannot) and how this information can inform philosophical debates. This is a nice example of how Experimental Descriptivism (as Thomas and I called it in our taxonomy of exp phil) plays an essential role in any positive or negative projects in experimental philosophy (and in philosophy in general).

[Richard Holton]

Chandra & Sara

Very interesting paper, both methodologically and substantially. It also led me to read your 'Deep Self Model' pre-print on the Phil Studies site, and the discussion, which I'd missed, when you posted it to this blog a couple of years ago.

The Deep Self model does seem to explain some findings that other accounts don't; and you convincingly argue that it does well on the structural path analysis, given the further questions you ask (not that I have much competence here to say anything useful!). But can you say something about how it explains some of the further findings in the literature? (Thomas alluded to some of these in his response to the original posting.)

For instance, how does it explain the asymmetry in Knobe and Pettit's case of the two subjects one of whom is trying (with little chance of success) to detonate a bomb, and the other of whom is trying (again with little chance of success) to defuse it? In both cases the action looks to be in line with the agents' deep selves; but there is more agreement that the former intends to detonate the bomb than that the latter intends to defuse it. Or take Phelan and Sarkissian's case of the lieutenant who moved his troops into a dangerous position. There he was more likely to be judged as having caused their deaths intentionally if the goal was presented as unimportant; his stated degree of care for his troops had little impact. There one might have expected the deep self to be most concerned with degree of care.

[Thomas Nadelhoffer]

Chandra,

I just wanted to second Eddy's remarks about your very interesting and illuminating paper. I suspect the methods you and Sara adopted to analyse your data will have a lasting impact on experimental philosophy moving forward.

For those of you who are interested in Chandra's earlier paper on the Deep Self View, see here:

http://experimentalphilosophy.typepad.com/experimental_philosophy/2007/11/the-deep-self-m.html

Unfortunately, comment threads now close on this blog after a month to keep down on the spam. But if you have any comments about his earlier paper, this present thread is a good place to post them.

[Chandra Sripada]

@Eddy and Thomas: Thanks so much for the praise!!

@Richard: Many thanks for these thoughts. Here is how I would handle the two cases you mention:

1. Pettit and Knobe’s ‘Detonator/Defuser Asymmetry’

The notion of a Deep Self is a multi-dimensional notion. I hypothesize that people will rate a pro-attitude as having more ‘depth’ based on several often correlated but potentially dissociable features. These features include: importance (very roughly, the object of the pro-attitude’s ranking in a preference ordering), generalizability (the attitude’s tendency to generate a pervasive pattern of behavior across situations and contexts), and ‘commitmentness’ (the attitude is stable and longstanding). There are others as well, but these are key for what follows. Also, let me add that this is a very tentative ‘armchair’ driven list. I am developing studies that identify the dimensions of Deep Self using quantitative methods.

In Pettit and Knobe’s ‘Detonator/Defuser Asymmetry’ case, the pro-attitudes that are the causal source of the agent’s bringing about the outcome are matched between the detonator and defuser in terms of importance only (both very much prefer the bomb’s being detonated [defused] compared to other outcomes). But the detonator’s pro-attitudes score more highly in terms generalizability and commitmentness, as the detonator is intuitively understood to be a typical terrorist. Thus his pro-attitudes relevant to his detonating the bomb are taken to be stable, long-standing, likely to drive behaviors in wide variety of contexts, etc. I predict that in the two versions of the case, differences in imputed depth of the agents' pro-attitudes (and not the normative valence attached to agent or outcome) drives the differences in intentionality judgments. This is a testable hypothesis of course.

2. With Phelan and Sarkissian's lieutenant case, I don’t think the degree of caring manipulation was successful. It strains credibility that a person who genuinely cared about his soldiers would sacrifice them to take a hill he knows he can’t hold, for the stated purpose of ‘simply taking it.’ So the vignette stipulates the lieutenant is caring, but subjects don’t actually believe it. I propose that the degree of importance manipulation actually drives caring judgments (and the vignette’s stipulation of level of caring has relatively little effect on caring judgments). If this is true, then the Deep Self Model does a nice job of predicting the result that importance judgments influence intentionality judgments (with caring judgments, or some closely related deep pro-attitude, playing the role of key mediating variable). Once again, this hypothesis is very testable.

[Jonathan Livengood]

I didn't see a covariance matrix reported in your paper. Did I just miss it? I cannot replicate your analysis without the covariance matrix (or the correlation matrix).

[Chandra Sripada]

Hi Jonathan,

I just added the covariance matrix as an 'Appendix B'. I did not place it in the original to reduce the reader's feeling of being hit with too many numbers. After all, our goal is to publish in a mainstream philosophy journal, though I am not sure how this paper will be received.