Before Damasio discovered Descarte's Error and before Ledoux settled (;-) on The Emotional Brain as the foundation of our rationality, I had a good idea and it felt good because.... well....

Good Ideas Feel Good (Copyright 2003, Stephen R. Deiss, all rights reserved)

"You know that feeling that you get when you are close to the solution of a problem you've been desperately trying to solve? The tension mounts as somehow you inexplicably know you are on the right track... yet you are frustrated because it's just on the 'tip of your tongue.' There is something missing that keeps you from putting it all together... like a puzzle. I am reminded of Stein's book on creativity which gave a synopsis of a famous old theory of the creative process (attributed to Poincare or Helmholz - I don't remember who for sure) as involving immersion (in the problem), followed by dropping it for a while - incubation, then suddenly (out of nowhere?) illumination - aha! - eureka, and then the final stage of verifying and communicating the results. Well, I'm working on such a problem now with the interesting twist that my working on it is the problem. I'll just keep thinking out loud so that you can see what's happening. I have what I think is an unusual cognitive style that involves pursuing analogies in a non-deductive fashion. I was encouraged by Norman and Bobrow's paper on schemata because they confirm a long standing suspicion of mine that thinking is at bottom analogical. Rumelhart and Ortony really hammer this point home showing how inferential reasoning is a type of analogy detection process. In another paper Rumelhart and Abrahamson talk about analogical reasoning as if it involved calculation of a directed distance in semantic n-space. That reminds me of Winograd's discussion of modularity of interactions in systems after Simon. He makes the point that systems based on predicate calculus assume a high degree of linearity as opposed to highly interactive network models (procedural models). This in turn reminds me of general systems theory that emphasizes the way a whole is constituted by interaction of its parts. Systems are better seen as sets of simultaneous differential equations where a change in any variable affects all the others as opposed to a 'summative complex' like a bag of marbles. Another way it's been said is that the variance of the whole system is much less that the sum of the variances of the component subsystems. Winograd's discussion also reminds me of the notion of a linear vector space. As inference rules map axioms to new propositions, linear transformations map from a basis to a vector. Basis elements are linearly independent sort of like axioms. In normed spaces we can talk about the distance between things - which reminds me of Rumelhart again. An orthogonal basis is analogous to a consistent set of axioms is analogous to a non-dissonant cognition. Could there be an isomorphism somewhere in all this? All this talk about analogical reasoning somehow does not seem to fit right off with Pylyshyn's notion of propositional knowledge representation. Analogies are structural animals, and there does not seem to me to be enough structure in semantic nets. That reminds me of a paper I did once on analogical reasoning in (old style) semantic information networks. The problem I tried to solve then still seems to exist, namely, one wants to look at the structure of the network to find analogies, but you can't. You find yourself tracing a path through the network and then hunting for a similar path - an analogous path. What do we 'see' when we 'see' an analogous relationship, and what does it mean in this case to 'see'? Maybe that is what Pylyshyn and Norman are getting at. Yet talk about propositions, frames, descriptions, and active structural networks all sounds somehow too mechanical. Where has the beauty and excitement gone in the discovery?...... Another thing that keeps popping into my head in this context is the way the problem of pattern recognition seems so fundamental in all this. To discover a pattern is analogous to (i.a.t.) performing an induction i.a.t. seeing an analogy i.a.t. pigeonholing i.a.t. the problem of generalization. It seems at times like at every turn we see the same old problem in new clothes. Another example, Wittgenstein's notion of family resemblance. How could we ever come to a conclusion with all these schemas chattering back and forth passing the buck (descriptions) to one another? How can the thing converge? We need a homunculus in there to holler when things look right or feel right or .....but we can't do that - the ego (the 'decision maker') is the RESULT of all this chatter - not the conductor. What to do... Ah, but a thought now comes to mind about the use of emotions as heuristics. Of all things, emotions. The neural network modelers are certainly giving them some thought. When you think about it it sort of fits. Emotions can tell 'us' when things look right, when they feel right..... when they fit together even. When they are analogous? This all reminds me of Pirsig's book in which he refutes the idea that we can reason without value judgments. They are in there, but where.....maybe.....could it be?....but of course! Value judgments - emotions - heuristics - there's the 'minds eye.' All these AI programs are trying to operate without feelings like disembodied minds (in reaction I suppose to Skinnerian diseminded bodies). That's it.....

That is it ! It's got to be it. It feels right. It fits. I think I'm onto a good idea. But now how could I ever begin to communicate this idea so as to really make it stick? I think....uh...........I've got it:

You know that feeling you get when you are close to the solution of a problem you've been desperately trying to solve?..............."

The End (of a beginning)

[I'll spare you the references for this essay. They are pretty obvious from the context, even if somewhat dated now. It was written for a graduate seminar in Human Information Processing (back before they called it Cognitive Science) taught by Don Norman at UCSD (now Emeritus) in 1976. For similar contemporary views I highly recommend reading Johston's readable little book on Why We Feel and for additional background see recent work by Ledoux, Damasio, Hofstadter, Fauconnier and Gardenfors......... and.... well.... stay tuned.....]

Back….