As we live our lives, we gain experiences that are recorded in short-term memory and then transferred to our global search repository – our long term memory. Cognitive psychologists and neuroscientists are exploring these mechanisms. I want to discuss an emulation framework that can be represented and improved in computers.
Hypothesis: We can represent our choices as rows and we can pre-consider the many contexts under which we might make choices. I’d like to do this repeatedly in two dimensional tables. That is one with rows and columns. Such tables are often given labeling information in the first column and the first row, as shown at the top of this post.
The ‘table’ helps us to be more careful about context scope. Many people may not consider changes in situation just as important as strategies and plans that they create, yet they should be tightly interwoven in the creation process. Otherwise, we’ll devolve to same old situation, same option chosen. The literature calls this satisficing.
This idea is in direct opposition to operations research and the field of decision theory which attempts to maximize gain even in only partially known situations.
Intuitively, considering more choices slows the decision process, but often lessens the likelihood of being surprised. There are a number of day-to-day situations where re-thinking our past responses seems worthwhile.
To use a table alone, presupposes that all of the information for context and strategy selection are at the same ‘level.’ Often, there is a zooming in (examining more detail) and zooming out (checking why we are making the current decision that our brains do reasonably well. But computers don’t.
So I will be describing how to build better computational brains – so we humans can perform better.