Scientists Predicting Intentions 105
An anonymous reader writes to tell us German scientists claim to have the means of predicting decisions of high level mental activity. "In the past, experts had been able to detect decisions about making physical movements in advance. But researchers at Berlin's Bernstein Center for Computational Neuroscience claim they have now, for the first time, identified people's decisions about how they would later do a high-level mental activity _ in this case, adding versus subtracting."
I randomize lots of things (Score:3, Interesting)
I'm now seeing that this was a very wise decision....
I do a lot of sub-optimal things, but at least I'm not predicatable
Re:Suspicion (Score:5, Interesting)
Here's how this stuff works. Step 1, scientist do incremental, meaningful, but boring (to those outside their specialty) work. Step 2, media picks up on story and puts overreaching spin on story. (Alternatively, the scientists, the journal, or the university's PR office puts out a press release supplying overreaching spin to credulous journalists.) Step 3, everybody sits back in wonderment at a finding that essentially establishes what we already knew: that mental processes take place in the physical brain.
Parent poster is right about the special demands of individual prediction. The basic science might be incrementally useful - trying to ultimately understand how future planning/intentions take place in the brain. (And given the breadth of mental operations that could be considered "intentions," there are probably hundreds of more studies that need to be done before that question can begin to be answered.) But going from a scientific explanatory mode, where you have potentially large samples and budgets and cooperative subjects, to prediction of individual behavior is a huge leap. Just look at a much older psychometric approach, the TAT, which is okay for research [psychologicalscience.org] but lousy for individual prediction [psychologicalscience.org]. Brain scanning may well turn out to be the next TAT, for precisely the same reasons.
Part of the problem is that a lot of this work is being done by medical researchers and neuroscientists who have no basic training in psychometrics. They're just reinventing old mistakes (but wasting a hell of a lot more money this time around).
Re:Suspicion (Score:3, Interesting)
Well, sort of. My impression is that this has little to do with a lack of training in psychometrics, but a lot to do with the more general problem, evident to anyone who reviews the occasional fMRI article, that researchers like to make unquantified (or improperly quantified) observations. Most often the data are there, just not analyzed properly. This is really just a basic issue with the use and reporting of inferential statistics.
That said, I don't honestly see that it's a big issue here. It seems like the authors did something sort of reasonable and drew mostly reasonable conclusions (I say this without having given it the close reading I reserve for research I really care about). My sense is that the desire to overextend the results is coming more from the reporting of the article and less from the reporting in the article. In other words, it's not clear to me who needs the training in psychometrics.