Toward Self‐Aware Models as Cognitive Adaptive Instruments for Social and Behavioral Modeling
Levent Yılmaz · 2019
Model development is an incremental and iterative process, but it often prematurely converges to a single supposedly authoritative model. Better use of scientific method involves a dynamic, iterative process in which diverse hypotheses and combinations thereof are compared with evidence. The goal may be to find a set of models and variants thereof with different structures and to have an understanding of when to use. In this chapter, I describe an abstract architecture to guide the dynamic and iterative inquiry. The intent is to identify the different functions that it would be useful to accomplish. These include management functions associated with the designing experiments and composing models for theory–data comparison and scientific discovery as with inferring causal relationships and aiding the researcher along the way in various other ways. I take a systems approach that envisions a special class of "cognitive models" to assist the process. These have self-awareness features. The intent is that this conceptual architecture will assist researchers as they prepare for complicated experiments comparing models with evidence, i.e. that it will alert them to the need for mechanisms to accomplish the functions discussed.