bouncing ideas with Ion

[gusl]

These are my notes on the ideas related discussed this Friday at lunch with Ion Juvina. The topic is symbolic cognitive modeling.

* corpus of models

* the idea of "cognitivizing" computational processes
 * information processes can be modeled in two kinds of machines:
human and computer
 * human <-> computer compiler: programming
 * instead of Java / ACT-R, we would need a high-level specification
language: GOMS?
 * cognitivizing is the reverse of task analysis
 * Q: does programming consist of making cognitive models of themselves?
   * A: yes, mostly, except for the cases where the programmer chose
to e.g. scan an array sequentially. The programmer is not aware of how
exactly his vision works (it's probably better than sequential
scanning), and implementing neural networks would be too hard.

* Q: how are conscious productions modeled?
 * A: by having declarative chunks associated with them
 * ...but does activating a production really activate its associated chunk?
 * simplified ontology: everything is chunks

* refining production rules by reflecing on mistakes
 * are they new rules or the same rules? i.e. is the utility
borrowed? if they are different rules, then would we get shadowing
effects?