Supporting Artificial Social Intelligence With Theory of Mind

Abstract

In this paper, we discuss the development of artificial theory of mind as foundational to an agent's ability to collaborate with human team members. Agents imbued with artificial social intelligence will require various capabilities to gather the social data needed to inform an artificial theory of mind of their human counterparts. We draw from social signals theorizing and discuss a framework to guide consideration of core features of artificial social intelligence. We discuss how human social intelligence, and the development of theory of mind, can contribute to the development of artificial social intelligence by forming a foundation on which to help agents model, interpret and predict the behaviors and mental states of humans to support human-agent interaction. Artificial social intelligence will need the processing capabilities to perceive, interpret, and generate combinations of social cues to operate within a human-agent team. Artificial Theory of Mind affords a structure by which a socially intelligent agent could be imbued with the ability to model their human counterparts and engage in effective human-agent interaction. Further, modeling Artificial Theory of Mind can be used by an ASI to support transparent communication with humans, improving trust in agents, so that they may better predict future system behavior based on their understanding of and support trust in artificial socially intelligent agents.

Computable Artificial General Intelligence

“…[27,28,29] of Deepmind illustrates how a curated vocabulary (in which to construct rules) can provide such an inductive bias (which would make weaker rulesets easier to find in some contexts). Finally, the formulation of tasks employed here has been used to explain how human language functions [3] and lends itself to an artificial theory of mind [4,5]. The more rigorous mathematical treatment given here has also been applied to those derivatives in an upcoming publication.…”

“…There are compelling arguments to be made for the development of this sort of AGI. Such an agent would not only more effectively learn, adapt and even ascribe purpose to what it observes [3,4,5], but may yield social benefits in comparison to methods popular today. For example, only large organisations have the resources to train models that require a lot of data [6].…”

Computable Artificial General Intelligence

“…This limits what sort of tasks it can learn to those for which it has exact language. Finally, the formulation of tasks employed here has been used to explain how human language functions [8] and lends itself to an artificial theory of mind [10,11]. The more rigorous mathematical treatment given here has also been applied to those derivatives in an upcoming publication.…”

“…Second, this description of intelligence addresses not only the ability to learn and adapt, but has been used in philosophical and psychological arguments to explain intent, the meaning of language and the ability to ascribe purpose to what is observed [8,10,11]. A description of intelligence which explains a wider range of human behaviour is more compelling.…”

Computable Artificial General Intelligence

“…There are three reasons we equate intelligence with ability to generalise, instead of adopting the same description of intelligence as AIXI. First, the ability to generalise entails not only learning and adaptation, but has formed the basis of philosophical and psychological research explaining intent, the meaning of language and the ability to ascribe purpose to what is observed (Bennett 2022a;Bennett and Maruyama 2022;Williams, Fiore, and Jentsch 2022). A description of intelligence which explains a wider range of human behaviour is more compelling than one that explains only naive utilitarianism.…”