Blog Coordinator

Knobe's X-Phi Page

X-Phi Grad Programs

« Free Will and Consciousness: Experimental Studies | Main | Gettierized Knobe Effects »

04/24/2012

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Joshua Knobe

Hi Florian,

This is very intriguing indeed. Maybe it will turn out that we were simply wrong about that point, and if it does, that would definitely provide strong support for the claims you make in that new paper looking at intuitions about free will in participants with emotional blunting due to frontotemporal dementia (http://philpapers.org/archive/COVJAM).

Overall, your work on this topic is definitely building a strong case for the claim that the effect is not a matter of anything about emotional reactions but rather a pure effect of concreteness.

Florian Cova

Thanks, Joshua, for taking it this way, and for the advertising.

Nevertheless, it is even more intriguing than you think. I didn't precise it in the post, but my results (and Feltz et al.'s results) suggest that people give mostly incompatibilist answers to the low- and high-affect cases (taxes and rape), but mostly compatibilist answers to the concrete case (the man killing his whole family). Thus, the results cannot be explained by the abstract/concrete distinction, nor by alternative hypothesis such as the NBAR hypothesis. This is why my first move was to ask if I was the only one to find such results.

Adam Feltz

Hey Florian,

Edward and I have had some problems replicating the result from our labs. We have managed to replicate it a few times, but many more times we've failed. From our perspective, this is likely because the paradigm is not very well suited to detecting this difference. Given the number of times we've seen the effect (and given the theory and other data), I'd guess there is likely something to it but again this particular paradigm isn't very reliable (only produces a relatively small effect that is only 3-4% of the variance).

Florian Cova

Thanks, Adam.

Did you observe the same pattern of results than I did, that is: mostly incompatibilist answers for both cases? or did the pattern change from one experiment to another?

Joshua Knobe

This is definitely getting more and more intriguing. I can't speak for Shaun, but my own sense is that -- between this study and your study on people with FTD -- you are really providing some pretty convincing evidence that we were wrong to think the effect could be explained entirely in terms of emotions.

But there is one thing that still troubles me. De Brigard, Mandelbaum and Ripley do get a significant difference between the abstract case and the very brief high affect case. Do you have any idea why that is happening?

(One possibility is that the differences between these different studies arise from differences between the different populations people are examining. Our original study was on undergraduates at University of Arizona, and I could imagine that one might obtain different results if one looked, e.g., at people using mTurk.)

Felipe De Brigard

I feel that in our study the abstract/concrete and high-affect/low-affect were confounded, as back then we thought that both went hand-in-hand. I think that Adam may be onto something when suggesting that the paradigm may not be fine grained enough to tell apart what processes may be driving the distinction. Without controlling for processing time, for instance, it is unclear whether the only difference between conditions is their level of abstractness. It may be possible that, given enough time, both "concrete" and "abstract" conditions would elicit enough details so as to mobilize affective reactions. As you may remember, William Brady and I showed that even with highly concrete and emotionally loaded scenarios you can get both compatibilists and incompatibilists responses when manipulating additional emotionally charged details, such as possible future consequences for innocent third parties. Since the paradigms thus far used don't do not guarantee that participants are only thinking about the information presented in the vignettes, and do not control for the influence of additional thought processes, it is hard to figure out whether the effect is due to an abstract/concrete distinction or something else altogether.

Thomas Nadelhoffer

In my earlier work with Eddy, we used both high and low affect cases and I don't recall that we found any statistically significant differences. The majority of people thought the agent acted of his own free will despite the super computer's predictions and it didn't matter whether he had robbed a bank or cut the grass. This is something that has come up several times since Josh and Shaun published their data and model. I take it Eddy might have a theory or two on this front! Minimally, it seems like the affective performance error model doesn't explain all of the existing data.

Florian Cova

@Joshua:

As said by Felipe, the study you mention does not allow us to distinguish the effects of concreteness from the effects of emotion. Also, I think this result can also be explained by the fact that, in the abstract condition, people imagines the kind of behavior they think likely to be irrepressibly triggered by a neurological disorder, though they have a hard time to believe that a neurological disorder can lead to an irresistible impulse to rape (and thus that, having not resisted though he could, the agent is somewhat responsible.) (More generally, I feel that responsibility ratings are still high in De Brigard et al's studies because people consider that you can be responsible for an action you did not voluntarily initiated if you could have resisted it but did not.)

As to my results, maybe we can offer the following Nahmias-Murray-like account: the description of Universe A leads people to interpret scenarios as "bypassing" cases. This reading is confirmed in the low and high- affect cases because it is said that the agent does something he has often done in the past and there is no mention of his mental states (so it looks like a compulsion): However, in the concrete case (burning his family), there is no mention of repeated action and it is said that the agent acts because he has a certain desire (to be with his secretary). So, in this case, the "bypassing" reading is trumped.

Adam Feltz

When we gave participants both high and low affect scenarios, they tended to be incompatibilists (the Feltz et al paper). However, when we gave them only one of the scenarios and we didn't replicate the effect, they tended to be compatibilists in both scenarios. But affect does do some work pushing around the types of justifications people offer—you can see the effect in one of our forthcoming papers: http://faculty.schreiner.edu/adfeltz/FWDecisions.pdf .

Joshua Knobe

At this point, it is definitely looking like this effect is not actually driven by emotion -- which certainly takes things in an exciting new direction. The major question now, I think, is what actually does explain the highly reliable difference we get between abstract and concrete cases.

A number of people have pointed out that participants who receive the abstract case seem to think that the beliefs and desires of an agent in the deterministic universe have no impact at all on her behavior. This is a very important point, but the natural question to ask is *why* they think this and why they end up reaching a different conclusion in the concrete case.

It should be pretty clear at this point that the effect here is not just the matter of people confusing determinism with something else (say, fatalism). After all, it is not that participants think that nothing in the deterministic universe has any impact on anything else. Rather, they think that all sorts of things can influence subsequent events -- even including emotions influencing facial expressions -- but that there is some specific reason why beliefs cannot influence decisions.

For a brief description of this result, see:

http://agencyandresponsibility.typepad.com/flickers-of-freedom/2011/06/nahmias-and-murray-on-bypassing.html

The fact that people think causal determinism would specifically interfere with the relationship between reasons and actions (while leaving other causal relationships in place) seems to be pointing at something fundamental about how people understand the relationship between reason and action as working under normal circumstances.

But what exactly is it telling us? And why do people's intuitions change so radically when they are presented with a more vivid concrete case?

The comments to this entry are closed.

3QD Prize 2012: Wesley Buckwalter