Wednesday, November 14, 2007

a relatively naive gripe about brain research

First, look at this article on what happened to some people who got fMRI'd and received politics-related stimuli. Or maybe just read this quotation, which gives the tenor of the piece:

When we showed subjects the words “Democrat,” “Republican” and “independent,” they exhibited high levels of activity in the part of the brain called the amygdala, indicating anxiety. The two areas in the brain associated with anxiety and disgust — the amygdala and the insula — were especially active when men viewed “Republican.” But all three labels also elicited some activity in the brain area associated with reward, the ventral striatum, as well as other regions related to desire and feeling connected.

Or this one:

Voters who rated Mrs. Clinton unfavorably on their questionnaire appeared not entirely comfortable with their assessment. When viewing images of her, these voters exhibited significant activity in the anterior cingulate cortex, an emotional center of the brain that is aroused when a person feels compelled to act in two different ways but must choose one. It looked as if they were battling unacknowledged impulses to like Mrs. Clinton.

Granted, there is some hedging in these claims. But I think it's clear that the intended interpretation of the data is: Activity in the insula indicates that men are disgusted by the Republican party. And likewise, mutatis mutandis, for the other trials.

Drawing conclusions about people's mental states from brain imaging is tricky for a lot of reasons, but isn't there a very basic confusion going on here? I mean that what (say) fMRI studies give us is information about brain activity given some emotional state (or whatever), where the emotional state is identified independently by subject report or reasonable assumption. So the studies tell us that the probability of B given E is very high. But the claims above are of the reverse form: that the probability of E given B is high, enough so for us to be confident that the subjects really are anxious, disgusted, etc., based upon their patterns of brain activity. And absent background assumptions, P(B|E) and P(E|B) are completely independent, so the one tells us nothing about the other.

Is it possible that these scientists are making such a basic confusion of reasoning? Are there important and reasonable assumptions that I don't know about? Because to my eye, articles like this are really just embarrassing exercises in stamping the imprimatur of brain-scanning techniques onto the messy phenomena of human choice.


Mike T. said...

Hmm. Surprised no neuro-types have replied to this. Maybe this will stoke the fire a bit.

I think that many scientists either make exactly that mistake or don't care if their silence on the matter implies it, and that most science journalists are actually scientism-ists who either don't notice it or care to.

Perhaps this is too Science Studies for people, though. How about this: if you think that B=E then P(B|E) and P(E|B) are the same and your question doesn't come up. I think lots of neuro-types think that B=E or at least that B is negligibly close to E.

Whatever the case about scientists and philosophers, my biggest worry is with science journalism. The flood of very interesting research about the relationship between concrete behavior and emotions, and fMRI scans has potentially very significant consequences for social and political institutions. Don't believe me? Google "neurolaw." There's a lot of loosy-goosey speculation about what fMRI's show presently morphing into "scientific knowledge" out there. People and politicians are more likely to make these kinds of mistakes without realizing it, and they can become ensconced into legal practice and legislation.

Theron said...

Hey Adam,
First, I'd like to note that, as some type of neuro-type, it's a shame that some studies (especially some fMRI studies) receive so much popular attention. Some of them are embarrassingly sloppy. So I am very sympathetic to what you and Mike are getting at on this score. Regarding the P(B|E) and P(E|B) problem, it seems to me that Mike is right that many think inferences between the two are legitimate because B=E, (or in more functionalist terms, because E is the computational pattern B instantiates).

But the answer, I think, to your question, "Are there important and reasonable assumptions that I don't know about?" is that neuroimaging is just one source of evidence. There are also intervention studies -- such as lesion studies and micro-stimulation studies. In these kinds of studies, information about the particular experiential states of subjects is obtained given some particular brain manipulation. In inverention studies we get information about P(E|B). I am not defending the particular study you cited, but I believe there is a reasonable amount of converging evidence linking the amygdala with fear and disgust processing. If you are interested, I can provide some relevant citations.