The New Science of Autonomous Human-Machine Teams (a-HMT): Interdependence Theory Scales

Abstract

Autonomous submarines. Drone wingmen. Hypersonic missiles. The evolution of autonomous human-machine teams (A-HMT) is occurring when the rapidity of making decisions have become central to military defense, operating complex systems, transportation, etc. Social science, however, offers little guidance for the science of A-HMTs. The problem with social science is its basis in rational methodological individualism (MI), likely at the root of its replication crisis and its inability to make predictions. MI has impeded the generalization of every theory that has used it, e.g., game theory, additive aggregation economics, assembling automata, political science and philosophy. In the laboratory and field, MI’s rational collective theory tellingly fails in the presence of conflict, where interdependence theory thrives. Recently, however, social science has experimentally reestablished the value of interdependence to human team science, especially for the best of science teams, but not theoretically, making the results important but ad hoc. By rejecting MI in favor of interdependence theory, a phenomenon difficult to control in the laboratory, we have hypothesized for teams, found and replicated that the optimum size of a team minimizes its member redundancy. With interdependence theory, we have also found that, proportional to the complexity of the barriers faced by a team to completing its mission, intelligence is critical to a team’s maximum entropy production (MEP); that whereas physical training promotes physical skills and whereas book knowledge promotes cognitive skills, these two skill sets are orthogonal to each other, resolving a long-standing experimental and theoretical conundrum; and, lastly, that the best determinations of social reality, decisions by a team, and decisions for the welfare of a society are based on the interdependence of orthogonal effects: the social harmonic oscillation of information driven by orthogonal pro-con poles, alternatively presenting one argument before an audience of neutral judges countered by its opposing argument. From this foundation, unlike traditional models based on MI, interdependence theory scales to integrate wide swaths of field evidence, e.g., bacteria gene and business mergers seeking MEP, but, if failing, leading to collapse (weak entropy production, WEP). Instead of predictions which fail in interdependent situations, the way forward for autonomous systems is to limit autonomy with checks and balances, similar to how free humans limit autonomy.