Abstract

The purpose of this work is to rethink the process of learning in human evolutionary systems. We take a sober look at how game theory, network theory, and chaos theory pertain specifically to the modeling, data, and training components of generalization in human systems. The value of our research is three-fold. First, our work is a direct approach to align machine learning generalization with core behavioral theories. We made our best effort to directly reconcile the axioms of these heretofore incompatible disciplines -- rather than moving from AI/ML towards the behavioral theories while building exclusively on AI/ML intuition. Second, this approach simplifies the learning process and makes it more intuitive for non-technical domain experts. We see increasing complexity in the models introduced in academic literature and, hence, increasing reliance on abstract hidden states learned by automatic feature engineering. The result is less understanding of how the models work and how they can be interpreted. However, these increasingly complex models are effective on the particular benchmark datasets they were designed for, but do not generalize. Our research highlights why these models are not generalizable and why behavioral theoretic intuition must have priority over the black box reliance on automatic feature engineering. Third, we introduce two novel methods that can be applied off-the-shelf: graph transformation for classification in human evolutionary systems (GT-CHES) and dynamic contrastive learning (DyCon). These models are most effective in mixed-motive human systems. While, GT-CHES is most suitable for tasks that involve event-based data, DyCon can be used on any temporal task.

Degree

PhD

College and Department

Physical and Mathematical Sciences; Computer Science

Rights

https://lib.byu.edu/about/copyright/