Optimal Opinion Control: The Campaign Problem

Titelangaben

Hegselmann, Rainer ; König, Stefan ; Kurz, Sascha ; Niemann, Christoph ; Rambau, Jörg:
Optimal Opinion Control: The Campaign Problem.
Universität Bayreuth
Bayreuth , 2014 . - 40 S.

Volltext

	Format: PDF Name: PS.pdf Version: Preprint Verfügbar mit der Lizenz Deutsches Urheberrechtsgesetz. Das Dokument darf zum eigenen Gebrauch kostenfrei genutzt werden. Darüber hinaus bedürfen die Vervielfältigung, Bearbeitung, Verbreitung und jede Art der Verwertung der schriftlichen Zustimmung des jeweiligen Rechteinhabers.
	Download (707kB)

Abstract

We propose discrete-time models for the optimal control of continuous opinions of a finite number of individuals. Given a model of the dynamics of opinions, one individual can freely choose its advertised opinion in each stage, thereby gaining a controlled influence on the dynamics. Special attention is devoted to the bounded-confidence model, for which we point out some numerical instabilities, which is important for assessing the outcomes of simulations. We show that even in a small example system with eleven individuals and ten stages it is highly non-trivial to learn something rigorous about optimal controls. By means of mixed-integer linear programming we can characterize globally optimal controls. In the example case, we can use this to prove the optimality of some controls generated by a tailor-made heuristics. Investigations about the control population utilizing a genetic algorithm show strong evidence that optimal controls can be extremely hard to find among all controls. Controls generated by the popular generic feedback-principle of model predictive control can be significantly far from optimum, which underlines that more research is necessary to understand the structural properties of optimal controls.