Hacker News Reader

Comment by gwern | Hacker News Reader

> Say every child gets their connectome mapped when they start school to get placed in the best type of learning environment for them (auditory, visual etc.).

There is no reason whatsoever to do this. Aside from the issue that 'multiple intelligences' has not held up and has been abandoned, if you want to measure their intelligence, it's unlikely that brain scans will anytime in the foreseeable future outperform simply giving a child a 10-minute pencil-and-paper IQ test (and highly unlikely that scans will ever be cheaper or easier than a regular IQ test).

View on HN

Scans might be easier to sell to parents when putting the child in a different educational path than an IQ test. IQ tests may be perceived as just an individual rating the child, and parents that aren't completely happy with a result will get in the mindset that the tester "rated the child wrong" or "didn't measure the right things." An actual brain scan may be harder to argue and more effective in the end from the parents' point of view.

While I largely agree with you, I have to take issue with a couple of you points.

> outperform simply giving a child a 10-minute pencil-and-paper IQ test (and highly unlikely that scans will ever be cheaper or easier than a regular IQ test).

For children the test takes, "48–65 minutes to administer" and requires more than pencil-and-paper [1].

> There is no reason whatsoever to do this.

More importantly, I think doing a regular comparison of the standard IQ test results with this new scanning method could be a good way to account for test bias [2].

[1] https://en.wikipedia.org/wiki/Wechsler_Intelligence_Scale_fo...

[2] https://en.wikipedia.org/wiki/Intelligence_quotient#Test_bia...

> For children the test takes, "48–65 minutes to administer" and requires more than pencil-and-paper [1].

For the full-scale WAIS, perhaps. But you can still extract a good estimate from the subscales, and needless to say, the WAIS is far from the only IQ test out there.

> I think doing a regular comparison of the standard IQ test results with this new scanning method could be a good way to account for test bias [2].

Huh? The brain scans are being validated using the existing IQ tests in the first place, and if one doesn't accept all the existing literature on the unbiasedness of current IQ tests (where the concepts are not too difficult - for example, I think many laymen could get through and understand most of _Bias in Mental Testing_, Jensen 1980 http://emilkirkegaard.dk/en/wp-content/uploads/Bias-in-Menta... explaining what biases there would be, how one would test for them, and why IQ tests are unbiased), one is not going to accept the results of the infinitely more complicated & subtler brain imaging pipeline where there's so many details & variability that they had a pretty serious replicability problem for a while, especially when they just yield the same results that the original IQ tests would have.

OK, maybe that was a bad example, but it also wasn't the point. If you do a connectome mapping for one particular purpose, the ability to analyze that mapping data for many other (perhaps more insidious) purpose seems like a real problem.

A better example than the kid example I pulled straight from the article: If someone gets their connectome mapped to help select treatment for some kind of mental disorder, it may also be possible to measure if that person is prone to violence etc. All kinds of conclusions can be drawn from the same data set, limited only by the research that has been done to that point.

Is there a hypothesis that connects particular connectome structures to violent behavior?

Reply on news.ycombinator.com