Abstract

Abstract

In this talk, I will describe few current methods of approaching the problem of designing incentives for rational agents. I will start by considering the Mechanism Design framework in which the designer defines the rules of the game (e.g., allocation and payments) so as to induce some desired outcome whilst ensuring a given set of properties are satisfied (e.g., agents are not making a loss by being part of the system managed by the designer). Crucially, the agents have private information and are able to communicate with the designer. Then, I will discuss Incentive Design, a problem in which agents do not report information to the designer who is tasked with learning how to set incentives that induce desirable actions after making sequential observations of the agents play. In particular, I’ll describe a project undertaken at Prowler.io the research component of which has been accepted at AAMAS 2019 . Finally, I will introduce Information Design where the the incentives take the form of suggestions.

Speaker’s Bio: Sofia is a Senior Researcher in the multi agent research team at Prowler.io and have a background in game theory and mechanism design. Since her PhD, her work has been focused on designing incentives to motivate people and rational agents to behave as desired in different settings and when different types of incentive are possible. she received a Master’s degree in Computer Science Engineering (2008), and a Ph.D. in Information technology (2013) from Politecnico di Milano (Italy). In 2012, she has been a visiting student at University of Southampton (UK). From September 2013 to August 2015, she was Postdoctoral Researcher in the Networks, Economics, and Algorithms group at Microsoft Research Cambridge. Then, just before joining Prowler.io, she was Research Associate at the Centre for Intelligent Systems and their Applications, School of Informatics at the University of Edinburgh (Scotland) working at the Horizon 2020 project ‘Smart Society’ and at the EPSRC funded project ‘UnBias’.