Social-Behavioral Modeling for Complex Systems

24 Toward Self‐Aware Models as Cognitive Adaptive Instruments for Social and Behavioral Modeling

Levent Yilmaz

Department of Computer Science, Auburn University, Auburn, AL, 36849, USA

Introduction

The use of computational models in systems engineering is pervasive. However, despite the availability of useful tools that assist modelers in routine aspects of system modeling, other stages in evidential reasoning are not yet so helpful, especially for causal modeling. These phases include (i) the generation and prioritizing of modeling assumptions that are ripe for exploration, (ii) hypothesis‐guided automated generation and execution of experiments (Yilmaz et al. 2016, 2017), and (iii) interpretation of results to falsify and revise competing behavioral mechanisms.

A critical obstacle includes the disconnect between model discovery and experimentation. Another is the tendency to delay model justification until after model implementation. We conjecture that addressing these issues requires dynamic coupling of model building and experimentation with continuous feedback between their technical domains. Coupling these domains needs to provide sound explanatory characterization about what alternative and complementary mechanisms are plausible, whether they cohere and under what conditions. This is in contrast with the current practice. Often when experimental results deviate from expected behavior, modelers locate and mitigate the specific issue without looking deeper and ...