In article <20970 at castle.ed.ac.uk> pck at castle.ed.ac.uk (P C Knox) writes:
>andrews at syma.sussex.ac.uk (Andrew Shires) writes:
> <on network pathology>
>>I am under the impression that biological data in this area is collected from
>>willing human subjects who have suffered injuries, etc. How much does this
>>area rely upon vivisectory data? I'd be uncomfortable about taking this work
>>on if it involved the use of such findings. Please, withold any flames about
>>my being concerned about this, but I am quite happy to hear people tell me
>>I can't get on in neural nets work if I'm unwilling to use such information.
>>No flames. I think you'll find that most folk (including myself)
>involved in animal work appreciate the dilemas and are quite happy to
>discus the issues flamlessly. I think the points made elsewhere about
>the reliance of psychology and neurophysiology on animal work are right.
Thank you. I would like to take this opportunity to applaud readers who've
responded to me, on both sides of the debate -- everyone has approached this
question without bringing heavy ideological baggage along, and there have
been no flames. As a supporter of those who wish to ban needless animal testingI nonetheless have my doubts about the necessity of rejecting animal testing
for medicine, and other areas where results may reduce others' suffering.
I have little knowledge of neuroscience and exactly what is done with animals
or what kinds of findings are sought by neuroscientists. However, I am
not favourably disposed towards collecting data for its own sake.
For my own sake, I am quite pleased with what I have learned. Network
pathology can obviously be a purely mathematical discipline, but respondents
have told me that biological data used in this research is, largely, found by
examination of consenting human subjects who have suffered brain damage in one
way or another and who agree to examination when it might possibly improve their
health. Some neuroscientists have pointed out in electronic mail that a lot
of such evidence is anecdotal, and that, because of the legal restraints
upon examinations, some of the important aspects of damage cannot be
investigated, and that for these reasons animal tests prove useful for
understanding damage. This seems to me a logical argument, if predicated upon
an accepted speciesism, which I am not going to debate here. However, I feel
that there is too little similarity between artificial neural networks and
biological neural networks (particularly those of another species) for AI
researchers to justify commissioning their own experiments, even if they
support vivisection. Vivisection data already available is another matter,
of course, and some may feel it justifiable to use such data.i
>Bear in mind
>too that a lot of research relevant to neural nets is done on simpler
>animals, invertebrates such as insects and crustaceans. Here, the
>rampant anthropomorphism which so marks these discussions breaks down.
>But even here, where it is not clear what suffering means, precautions
>will be taken.
I can agree with this.
> So I sympathise with the dilemma. But examine the facts as
>opposed to the propaganda (even including mine :-) !). But don't ignore
>or undervalue a vast body of careful and important work for wrong
> I wish you success in you deliberations!
Thanks to all again.
Andrew Shires, COGS, University of Sussex, Falmer, Brighton, ENGLAND
email: andrews at uk.ac.sussex.cogs and andrews at uk.ac.sussex.syma