[extropy-chat] Neural Engineering

[Anders Sandberg]

fredagen den 19 mars 2004 16.23 wrote Bill Hibbard:
> The really interesting question is whether something like
> this is happening already. I understand that there are
> brain imaging techniques that can be used as effective lie
> detectors. If you show a person a picture, imagers can
> detect whether the person is familiar with the picture's
> subject.

"Brain fingerprinting" as it is called. You look for the P300 "oddball" 
pattern in the EEG when the subject sees images of objects in the same 
category. If you see a number of knives and have nothing to do with a murder, 
you would not react to any in particular. But if you were the murderer (or 
knew something of the murder) there would be a P300 response when you saw 
something you recognized.

This has been used in a few court cases, and is apparently viewed as 
admissible evidence in the US. There is debate about admissibility in India 
right now, and in a few other places.

>From a neuroscience point of view it seems workable, although we might want 
more reliable methods and there is a serious problem of using the right 
questioning technique (bad pictures of the target objects distinguishable 
from other pictures, subconscious influence from testers etc). 

> I wonder if these techniques are being used in questioning
> terrorists. It seems to me that with clever questioning,
> and looking for correlations among the responses of many
> terrorists, that brain imaging could be quite effective.

In principle, yes. In practice this would be very messy. Imagine that you have 
a number of suspected terrorists and try to find out who were in the cell 
planning 911. Doing brain fingerprinting on them for recognition of some 
items that were in the cell, you could probably find them and maybe a few 
people closely connected. But what if you don't have the item and the 
existence and membership of the cell is conjectural? You could start showing 
pictures, finding out who knew who - but that is just a hightech version of 
what intelligence agencies already do. It doesn't help you in proving that 
they did it. Even worse, if you don't have the terrorists to study, you have 
a very hard time finding them. It is the usual intelligence problem. The 
usual suspects are easy to get (unless they hide in a cave somewhere), but it 
is the unusual suspects or remote parts of the network that are dangerous 
because they can strike at any time. 

In the end brain fingerprinting is an adjunct to ordinary intelligence 
gathering. It gets some bits of information that otherwise might be hidden, 
but it is no mind reader.




-- 
Anders Sandberg
http://www.nada.kth.se/~asa
http://www.aleph.se/andart/

The sum of human knowledge sounds nice. But I want more.