Fear The Mindreading Scientists! March 6, 2007Posted by Johan in Mathematical Cognition, Neuroscience.
Best headline this week: Mindreading Scientists Predict Behavior. Second only to the neurotic Freudian analyst in cliches about psychologists is the idea that psychologists can “read minds.”
Of course, to the extent that psychology is concerned with predicting human behaviour, and your mind is expressed in your behaviour, they are absolutely right. What people do tend to get wrong is the accuracy with which this is possible. As any number of factors influence a psychological trait, you generally end up with probabilistic estimates. In this case, Haynes and colleagues were able to predict with 71 percent accuracy whether the participant was mentally adding or subtracting numbers. This looks good until you realize that the level of chance in such a study is 50 percent. This, incidentally, is unimpressive compared to previous results of Daniel Langleben (the subject of a previous post) who achieved 85 percent accuracy at predicting whether subjects were lying or telling the truth about having viewed a particular playing card.
The two paradigms are not directly comparable, but they both point to a very real development in cognitive neuroscience: as scanner resolutions and methods improve, we are already at the point where predictions can be made about what someone’s internal events are, at a given point in time, and for a given task.
This may have ethical consequences, as Haynes and Rees (2006) outline. Conceivably, the ability to “read minds” could be used for sinister purposes by totalitarian states and others. However, I’m personally not that concerned, for a practical reason: in a real-life context, the predictions are never going to be good enough to be useful.
Psychologists have been able to make predictions about human behaviour for quite some time. What a hundred years or so of experimental psychology has taught us is that you rarely see a measure with enough accuracy to make it safe to conclude anything about the individual, based on their score on that measure alone. We can learn plenty about groups, but keep in mind that psychologists tend to get excited about any correlations over .5 (sometimes far less than that), and no one really gets over .8 outside of perception and other basic processes.
This is in part because psychological measures are not terribly good – direct measures of neural activity may indeed offer a way around the traditional problems of socially desirable responding and limited self-insight. But the limited predictive validity of these measures is also to do with the fact that human behaviour is the result of a complex interplay of disposition and environment. No matter how much you know about someone’s disposition, you cannot predict the environment – save for the gene-environment interactions often reported in behavioural genetics, i.e., the fact that people’s dispositions shape their environment: if you’re bright, you get put in the Magnet class; if you have a short temper, people will enjoy baiting you.
Outside of the laboratory, longer-term predictions based on neural activity will remain about as probabilistic as the behaviour-based predictions that psychologists have been making for decades. Advances in Neuroscience will not bring about Minority Report anytime soon.
Haynes, J.D., & Rees, G. (2006). Decoding Mental States from Brain Activity in Humans. Nature Reviews Neuroscience, 7, 523-534.