Does empathy between artificial agents improve agent teamwork?

Abstract

Both everyday experience and scientific studies indicate that emotional intelligence in general, and empathy in particular, improve the effectiveness of human teamwork. Research in affective computing confirms their significance in systems where humans and artificial agents interact. This paper explores the notion of empathy between artificial agents that has so far received little attention, and argues that it could have significant impact on the design of robust and resilient agent teams. Combining the formal framework of Emotional BDI agents with the principles underlying the leading model of empathy in psychology and neuroscience, we define a hierarchy of affective and behavioral responses, integrated into an algorithm that formalizes the interactions between the subject and object of empathy in the domain of artificial practical reasoning agents.

A model of empathy for artificial agent teamwork.

“…The model of empathy introduced in this thesis complements and strengthens some of our earlier results described in [Polajnar et al, 2011]. In th a t paper we adapt an existing model of natural empathy model, introduced by Goubert et al [2005], to explain the formation of affective response in artificial agents from topdown influences (experienced by the subject of empathy) and bottom -up influences (related to the object), and to show how they ultimately lead to behavioural responses of certain types.…”

“…Here, we first present an overview of the problem, clarify our motivation for studying it, and specify the general direction of our research (Sec tion 3.1). After that, we briefly review our previous work in [Polajnar et al, 2011], explain how much progress we have made in th at paper towards designing a model of empathy for artificial agents, and outline what remains to be done (Section 3.2).…”

“…One of the most important points we have made in [Polajnar et al, 2011], is the conjecture that empathy has the potential of improving artificial agent teamwork performance by initiating helpful behaviour in a team. Based on a simple simulation, we have illustrated how helpful behaviour improves the performance of a team under specific circumstances related to environmental dynamics and disturbances.…”

“…In [Polajnar et al, 2011] we have raised the question of whether and how empathy between artificial agents can improve their team performance. In th a t paper we have explored the notion of empathy between artificial agents and argued th a t it can have significant impact on the design of robust and resilient agent teams.…”

“…Such questions have not been investigated much so far and definitely need to be systematically studied. We raise some of them in [Polajnar et al, 2011] and provide a modelling framework for incorporating empathy into artificial agent teamwork. However, the results of th at paper do not include the models of factors and mechanisms by which the empathic responses in artificial agents are formed.…”