A Thermodynamics of Teams: Towards a Robust Computational Model of Autonomous Teams

One of the great puzzles in social science is the failure of rational models of teamwork, a rising concern for Artificial Intelligence researchers. Social learning theory (i.e., rewards and punishments; associations; modelling) works partially with individuals, but not with teams. These theories of methodological individualism, including, but not limited to, game theory, have also failed to advance the field of economics. To address interdependence, the phenomenon central to teamwork, we explain why game theory, the first to study interdependence mathematically, has failed. As an alternative, we offer a non-rational theory composed at this time of three parts: quantum mathematics for interdependence (e.g., interference); biology for population effects; and min-max entropy production as a metric of good and unsatisfactory team performance for humans or artificial agents (with min entropy production as LEP, and maximum entropy production as MEP). We report on three mathematical breakthroughs: First, that the interdependence between an individual’s observations and actions, once measured, breaks the link known as intuition, leading to the measurement problem of incompleteness, accounting for the failure of survey instruments to predict human action; second, that at the team and larger levels of analyses, the ill-effects on min-max entropy production of consensus-seeking rules and authoritarian leadership serve to suppress the search for solutions to the problems that teams work to solve; and third, as a speculation to integrate group decision-making and a team’s emotions: with LEP as a team’s ground state versus a team with internal conflict at an elevated LEP state.