Discipline

Today's rant will be about the overly attached emotions one has for the artificial construct of a ``discipline of knowledge''.

I was in a discussion with a friend not too long ago about the state of affairs of machine thought versus human thought regarding the issue of medical decisions. It started off innocently enough, but it rapidly devolved into a slinging match between the two of us as to how humans will always remain ``superior'' to the thinking machines because the former can make judgement calls while the latter cannot. I vehemently disagreed to the premise, of course, since I was of the opinion that any form of human judgement is not completely devoid of stimuli, and can be codified into a system of logics combined with the balancing of probabilities, sort of a mix of regular first (or even second) order logic with associated Bayesian prior distributions over various theorems that get updated with new information. My friend's argument was that machines are too limited and are unable to make judgement calls nonetheless, wherein I brought in the point that the way in which human judgement calls and machine calls are used are completely different.

A human has continuous stimuli in the form of faster feedback loops, while the machine is often forced to make a single judgement given only one chance at getting ``observations'' in the form of some type of sensory/test output. Such a test protocol is obviously unfair and imbalanced.

We wisely ceased continuing the conversation when it was clearly obvious that each of us thought that we were right while the other was a complete moron and completely wrong.

I reflected upon the conversation and came to a much simpler conclusion as to what was going on. We were talking from too deep within our own disciplines. Had the domain of discourse been substituted from medical decisions so something more neutral (in comparison anyway) like that of a judicial ruling, a more productive outcome would have ensued, with some of the points of differing test protocols, types of codified knowledge, relative likelihood assignment and the like be debated on a more neutral territory.

But I was coming from the machine learning/knowledge representation/probabilistic modelling perspective, while my friend was coming in at full speed from the perspective of a medical doctor in training. There are turfs to guard, and that's what the notion of a discipline in human knowledge teaches us. That I was advocating for a more fair assessment of how machine intelligence can help provide medical decisions was seen, rightly or not, as an encroachment on the sacrosanct turf of the learned practitioner of the medical arts, who of course see the computer scientist who looks at data as ``that silly guy who does not have a medical degree'', which prompted the robust response, be it good or bad.

That got me thinking even more about the whole concept of a discipline.

Having disciplines in human knowledge is a good thing. It partitions the vastness of human knowledge into mostly self-consistent chunks, with dogma codifying the founding characteristics of that particular chunk of knowledge. Great advances in human knowledge have come from the explorations of researchers who are a part of each of their discipline, contributing much from their perspective, and sometimes even challenging the dogma to update it with the newly acquired data.

But defining disciplines, like all forms of discrimination, have a dark side. It makes experts in their own discipline arrogant. It makes them complacent; many feel that since they are masters of their discipline, it makes them qualified to criticise seriously on the work of others from other disciplines at best, and at worst, find that it is in their perogative to downplay or even slam the work of others outside of their chosen discipline. Hence the attack on developing machine intelligence for medical decision making by my friend. The world view was strongly distorted from the long periods of time that my friend had spent in the medical world.

Then there's the rise of so-called ``inter-disciplinary'' work. So much talk about marrying engineering with science and what-not. I'm too tired to write more on this, but I'll leave with this Feynman quote from the first book of the lectures:

``...because the separation of fields, as we have emphasized, is merely a human convenience, and an unnatural thing.''