Re: <nettime> Brain fingerprinting

[nick knouf <nknouf@media.mit.edu>]

On Jun 19, 2006, at 10:21 AM, twsherma@mailbox.syr.edu wrote:

> Here's how it works. A criminal suspect or a terrorist is shown
> pictures of the scene of a crime or a terrorist training site. The
> suspect's  brain waves are monitored, looking for brain waves of
> recognition, signs  in the suspect's memory that links the suspect
> to the scene of a crime or terrorist activity. These brain waves of
> recognition are called P300 waves. The suspect may deny any
> involvement, but a real-time  analysis of his or her brain waves may
> conclusively establish prior criminal  activity.  Previously
> undetectable memories determine guilt.

As someone who has designed, run, and analyzed a number of EEG, MEG,  
and fMRI experiments in the past, I can say that this is still far  
off as an everyday device.  There are a number of practical and  
conceptual problems involved.  Oftentimes the journalistic accounts  
gloss over the difficulties, which are many.  Yet I think it's still  
important to be concerned and develop appropriate responses and  
procedures for using any type of "lie detector".

But onto some of the issues.  First, getting a reliable result within  
a single subject is extremely difficult for these types of  
measurements.  The data are so "noisy" (physical noise from the  
measurement device, as well as physiological "noise" that we may  
someday find out holds interesting data) that results must be  
averaged across many subjects before stable peaks can be found.   
However, it can be done in some cases and for some stimuli in single  
subjects: if you want to get reasonable results from only a single  
subject, you must present him/her with a large number of trials--- 
anywhere from 150-200 trials per condition (in EEG/MEG) and around  
100 (in fMRI) (this is all based on my experience and with the  
analysis tools I've used).

Second, we still do not know enough about the underlying mechanisms  
of memory to know how, in the case of EEG or MEG, the signals we  
measure are related to the underlying mental processes of memory and  
recognition.  All we have here (and with fMRI) is a correlative  
measure; none of these techniques can establish causation.  Perhaps  
the correlation is enough for a court of law; I don't know enough  
about legal standards of evidence to know for sure.

Thirdly, all of the techniques require a willing subject to remain  
still for anywhere between one and two hours.  It's possible to  
sedate a subject, however the sedation process will affect the  
results measured to an uncertain degree.  Also, with fMRI (and  
somewhat so for MEG, but for different reasons), you cannot scan  
people who have certain types of metal in their bodies, suggesting a  
possible (if invasive) countermeasure.

Finally (for now), there are still open questions as to the best way  
to analyze the data.  For example, with fMRI data there are a number  
of tactics to use: you can morph the data for each subject into a  
standard template, allowing direct averaging across subjects; you can  
"localize" areas in individual subjects, and then average the results  
across the localized areas; you can map the data into an agnostic  
"spherical" space to again allow averaging across subjects; and so  
on.  Besides these spatial issues, there is also certain disagreement  
as to the signal and statistical analysis techniques to use, both in  
fMRI and EEG/MEG.

Some journalistic articles:

http://www.reason.com/rb/rb111401.shtml
http://neuroethics.stanford.edu/documents/HankArticleStanfordReport.pdf

Some articles from scientific journals (I have not read these  
articles to know if I agree with their methodology):

http://www.hubmed.org/display.cgi?uids=11835606
http://www.hubmed.org/display.cgi?uids=16161128
http://biology.plosjournals.org/perlserv/?request=get-document&doi=10% 
2E1371%2Fjournal%2Epbio%2E0020150
http://www.hubmed.org/display.cgi?uids=11588589

Cheers,

nick knouf


#  distributed via <nettime>: no commercial use without permission
#  <nettime> is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: majordomo@bbs.thing.net and "info nettime-l" in the msg body
#  archive: http://www.nettime.org contact: nettime@bbs.thing.net