(12c) Control with Soft Feedback in Social Systems: Mathematical Principles, Empirical Evidence, and Applications

Feedback control has long been regarded as essential in engineering, biological, and ecological systems. Control in social systems, a recent addition, reflects further advances of this discipline. Feedback here includes opinions, consensus, polling results, market signals, rumors, product ratings, etc. Compared with the same concept in process control, feedback in social systems is “soft” and need not be followed. This unique property makes control in social systems challenging and interesting. To quantify the extent to which individuals adopt feedback, we introduce a new quantity—degree of social influence: It critically determines the closed-loop dynamics of a social control system. In this presentation, we will first describe a social system with multiple intelligent agents. Each agent has an optimization problem to solve. That problem, for instance, can be improving one’s health or finding the optimal state-wise taxation depending on the context. Our objective is to understand whether soft feedback makes the problem-solving process more efficient for the multi-agent system. If so, what is the optimal degree of social influence to achieve the most improvement? How can we make intelligent crowds even smarter? We will show via mathematical proofs and empirical evidence that it is indeed possible to make crowds smarter. Our work has a diverse set of potential applications in areas such as healthcare, revenue management, and policy-making.