Monday, April 14, 2008

Prediction of choices

Here's something a bit funny. We're told that people's decisions can be predicted by brain scan several seconds before the people think they've made a conscious decision. John Dylan-Haynes, one of the researchers, says: "Our decisions are predetermined unconsciously a long time before our consciousness kicks in". What is the evidence that the study provided for this interesting claim? According to the New Scientist's summary: "By deciphering the brain signals with a computer program, the researchers could predict which button a subject had pressed about 60% of the time – slightly better than a random guess." (The choice is a binary one.)

So the claim which the data supports is that several seconds before the subject thinks she's made a choice, the brain is in a state that makes one of the choices somewhat more likely than the other. Now, we can ask: Which of two hypotheses does this data support better?

  1. Our decisions are predetermined unconsciously a long time before our consciousness kicks in.
  2. Unconscious factors a significant amount of time before the conscious decision indeterministically affect the likelihood that we will consciously make one decision rather than another.
It seems clear that the data moderately supports (2) over (1). Why, then, would the researcher opt for (1) despite the data? Perhaps on the basis of other studies--but if so, then what he should say is that the present data weakens the evidence for his thesis (maybe he did and the New Scientist didn't quote that?). Or maybe the researcher thinks that improvements in the prediction procedure will eventually make it work at 100% accuracy. But the present prediction accuracy gives one very little reason to think this. Maybe the data rules out the hypothesis that there is no unconscious processes involved in decision-making, and hence offers support for (1) over the hypothesis that there is no unconscious component to decision-making at all, but who believes that hypothesis anyway?

It's worth noting that in ordinary situations where we ordinarily take people to have free will, we are often able to predict people's decisions with much better than 60% probability. This isn't to denigrate the brain-scanner. The case here is of a binary choice between button pushes, and our own accuracy would probably be fairly poor. Still, it would be fun to compare the brain-scanner against the accuracy in prediction by a researcher looking through a one-way mirror, seeing which button the person gazes towards, etc.

All that said, I actually have little problem with claim (1): I am not aware of a good argument that freedom (even of the libertarian sort) requires that one be aware of the decision when one makes it. In fact, if anything, it fits better with my preferred model of free will to have the awareness of the decision be explanatorily and maybe even causally posterior to the decision.

4 comments:

Anonymous said...

Alex,

What if the scanner was right 100% of the time, so that we would have (2') Unconscious factors a significant amount of time before the conscious decision deterministically affect the likelihood that we will consciously make one decision rather than another?

Could we then, in our reply, say something other than that the awareness of the decision is plausibly later than the decision?

Vlastimil

Alexander R Pruss said...

Well, the phenomenology of decision making often seems something like this. You think and think about the question. Then eventually you realize: "Hey, I've come to a decision." This realization seems to come after a decision. In fact, sometimes the thought is: "I've known for a long time that I was going to do this." (Whether that's literally knowledge or not is a different question.) So on phenomenological grounds, I don't think we can identify a "time of decision".

If the subjects in the experiment were identifying the "time of decision" with the time of the realization that they have decided, then of course that would come after the actual decision.

It seems to be a part of ordinary experience that sometimes we have already made a decision, but do not know that we have done so.

Anonymous said...

Thanks, Alex.

Assume for now that the scanner-predictions are right 100% of the time. I think that what is then interesting about the experiments like the one mentioned above, as I understand them, is that, *in the specific setting* of frequent choosing between few (say, two) alternatives which are trivial (say, pressing left/right button, or lifting up left/right hand), *the phenomenology suggests* that we make many decisions only a tiny interval before they are manifested (i.e., only a tiny interval before we press the button or lift up the hand). But, as the experiments suggest, the respective brain signals occur earlier (several seconds before the manifestation).

So, the phenomenology plus science plus freedom display some tension.

You claim that "on phenomenological grounds, I don't think we can identify a "time of decision"."
But it seems your talk about decisions in a different setting - as you wrote, about the cases when we "think and think about the question."

Vlastimil

Alexander R Pruss said...

Vlastimil:

You may be right. But then I will make a different move: it seems to me that these kinds of button-press cases are ones we would hardly call "decisions". There are, as far as I know, no significant consequences. They feel more like a mental coin-flip. The phenomenology, for me, is that the decision is as it were made by my hand. Now the 100% data would be incompatible with that, and even the 60% data is somewhat in tension.

However, I am also not sure that in a case like that we identify a conscious decision point, except in contrived circumstances where the experimenter asks us to identify it. The phenomology of such choices need not, I think, include a conscious decision point. So when we're forced to identify one, it's not a surprise that we get it wrong.