Modelling Shared Identity and Reputation in Cooperation Systems

Abstract

Cooperation is the process of working together for mutual benefit. Indirect reciprocity is an important form of cooperation because it assumes that a donation to an agent does not guarantee reciprocation. Therefore, understanding how cooperation is incentivised and sustained is of widespread interest. Reputation is known as a key mechanism to support indirect reciprocity because it is a currency through which future donations can be secured based on past behaviour. Conventional models of indirect reciprocity assume that agents have a simple identity that is uniquely defined and not shared with others. This results in a unique reputation for each agent. We generalise this assumption by allowing agents to share elements of their identity with others. This involves composing identity through traits, which can be used to represent group membership. Traits can be shared between agents and we assume that traits carry reputation in their own right, that an agent can inherit. Our investigation of this new framework provides an insight into the effects of sharing identity on cooperation in a number of different ways. Through a breadth of simulation, we identify the extent to which agents can have an element of common identity before cooperation becomes impeded. We also discover a relationship between reputation-based cooperation and cooperation through the evolution of set-based membership, which are previously unrelated alternative perspectives on indirect reciprocity. Finally, we explore the effects of blending personal and group reputations as seen in psychological theories of identity fusion. This allows us to determine the effects of identity-driven agent motivation compared to traditional economic motivation and rational economic decision-making. These findings give new perspectives into previous studies related to identity, such as stereotyping, group identity, whitewashing, identity fusion and intrinsic motivation.