Blog Coordinator

Knobe's X-Phi Page

X-Phi Grad Programs

« Registration Open: Experimental Philosophy as Applied Philosophy | Main | Workshop at NYU: Experimental Philosophy Meets the History of Philosophy »

02/12/2016

Comments

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

John Turri

Hello Alex,

Congratulations to you and your co-authors on getting this paper accepted at Cognition and thanks for your contribution to this research question! I like the idea of taking a step back and re-examining the potential consequences of experimental stimuli or procedures in earlier research.

Having read the paper, I had a question about the design of Experiment 1. In particular, I was wondering how the between-subjects design allowed you to rule out the potential confound of blame validation driving results on the lie attribution.

Alex Wiegmann

Dear John,

Thank you for your kind and encouraging words!
Regarding your question: In our first experiment we wanted to empirically test your plausible suggestion that the results of your first experiment (the majority of subjects choosing the subjective option) might be due to subject’s urge to blame the agent for trying to deceive the federal agents. So, in addition to your original scenario, we tested a scenario in which lying clearly seemed to be the right thing to do (- and not blameworthy; there was a merciless dictator who threatened to torture Jacob’s friend Mary). One finding was that in the original scenario as well as in the dictator version most subjects judged that Jacob was not to blame, another finding was that the percentage of subjects not wanting to blame Jacob was significantly higher in the dictator condition. Hence, we knew that these two scenarios differ with regards to the agent’s blameworthiness. We then tested whether this difference in blameworthiness had an influence on subjects ‘ratings of lying. We did not detect a difference: In both scenarios the majority of subjects judged that Jacob was lying. This finding, together with the finding that most subjects did not think that Jacob’s behavior was blameworthy, we took as indicating that subjects’ endorsing of the subjective option cannot be explained by subjects’ urge to blame Jacob.
I hope I got your question right.

Cheers

Alex

John Turri

Hi Alex,

Yes, that directly addressed my question, thank you. It aligns with how I interpreted the paper, but I wanted to make sure I wasn't overlooking something.

One worry we had about our initial studies was that an up-or-down lie attribution, all on its own, might be used as an opportunity to express disapproval of dishonesty or avoid appearing to condone or excuse dishonesty. How does presenting a different group of people with a blame attribution allow one to diagnose whether this happens to the people presented with the up-or-down lie attribution?

There are at least two different aspects to that more general question (the first of which seems more important).

First, granting that the two scenarios differ in the agent's blameworthiness, what does this tell us about how participants interpret the implications of answering "no" to the up-or-down lie attribution?

Second, for each story, most people said Jacob was not to blame, though the majority was stronger for one story than the other. You say that this then allowed for a "test" of whether "the difference in blameworthiness had an influence on subjects‘ ratings of lying." I can see how this might work for a within-subjects comparison. But what test supports this conclusion for a between-subjects comparison?

Alex Wiegmann

Dear John,

Thank you for your response!
Our goal was to empirically test your claim that people try to express blame when judging whether an assertion is a lie. We chose a between-subjects design because in our view this is the cleaner test. Within-subjects designs always have the problem of possible carry-over effect and demand characteristics. In our between-subjects experiment we randomly assigned subjects to conditions. So if subjects in condition A judge a lie as more or less blameworthy, the assumption is that subjects in Condition B who are asked about the lies would have similarly evaluated the lies as subjects in condition A. This inference is statistically warranted because of the random assignments of subjects to conditions. Of course, one could also run the experiment as a within-subject study. Although we think a between-subjects study is cleaner, we don’t expect any differences in this particular case. It would, however, certainly be a worthwhile project. Note that in our later experiments we followed up on your advice to use different test questions that are not potentially contaminated by blame judgments or perspective taking (we used your test question and just added the fact that what Jacob said turned out to be true).

John Turri

Hi Alex,

Thanks for clarifying the reasoning on this point.

You write, "If subjects in condition A judge a lie as more or less blameworthy, the assumption is that subjects in Condition B who are asked about the lies would have similarly evaluated the lies as subjects in condition A. This inference is statistically warranted because of the random assignments of subjects to conditions." That is all true. However, it is not directly informative regarding the perceived implications of answering "no" to the up-or-down lie attribution. I'm not saying that the data are irrelevant, but they do not, as is concluded in the paper, "rule out" interpretations of what the implications are.

The comments to this entry are closed.

3QD Prize 2012: Wesley Buckwalter