2009 IsAnAgentTheoryofMindToMValuabl

(Berg-Cross, 2009) ⇒ Gary Berg-Cross. (2009). “Is An Agent Theory of Mind (ToM) Valuable for Adaptive, Intelligent Systems?.” In: Proceedings of the 9th Workshop on Performance Metrics for Intelligent Systems. doi:10.1145/1865909.1865936

Subject Headings: Theory of Mind.

Notes

Cited By

Quotes

Author Keywords

adaptation; developmental robotics; theory of mind

Abstract

This paper serves as a short introduction for the special PerMIS session on Theories of Mind (ToM). The session intends to explore the viability of the ToM concept for R&D and if the ToM hypothesis is mature & relevant to the goal of highly competent systems able to achieve goals in a relatively autonomous way. The question can be considered from philosophical, research and robotic implementations as well as critiques central to the topic. The introduction is organized into 4 parts. Part 1 briefly reviews some of the history of the ToM idea and its recent reformulations. Part 2 discusses the widening use of the concept as an explanatory device within a few areas using developmental studies as a focus. Part 3 introduces the idea that particular types of robotics offer a new kind of tool to investigate cognitive development and the validity of some theories such as a ToM. The paper concludes with an outline of some issues that remain to be explored and advanced to show the value of a ToM theory in general and especially within the domain of intelligent systems.

1. INTRODUCTION and History

A Theory of Mind (ToM) refers to reasoning about the mental states of self and others. Empathy, the concept of putting yourself in another person’s shoes and relating to his situation, is a good example of theory of mind at work. This idea fits our everyday understanding of others, or what is called a folk psychology explaining why some things happen in the world. The practice of folk psychology has been a recurrent topic in philosophical and psychological discussions for a long time. In philosophy conceptualizing intentional states such as beliefs and desires has often been seen as dependent upon our linguistic abilities. That is, language ability seems to provide a representational medium for describing our own and others people’s actions in an intentional way. Recently, a new perspective on folk psychology has emerged in philosophy of mind and psychology. Such conceptualizations achieved a new purchase when Premack & Woodruff (1978) asked, ‘Does the chimpanzee have a theory of human intentions?’ or as they put it do they have a “theory of mind”. How would we know? The resulting research on this concept included its use as an explanation for autistic children’s cognitive deficits building on evidence suggesting that autistic individuals lack an ability to maintain a theory of other minds and to reflect on their own thought processes (Baron-Cohen, 1995). Since this early work the concept of a ToM has come to be broadly used in a developmental perspective of how children come to understand the social and psychological world. Formalized as a ToM theory these propose alternative inherited or acquired paths by which a particular cognitive capacity may arise in a cognitive agent (children) so they understand and predict external behavior of others by attributing unobservable mental states, such as beliefs, desires and intentions. Such a theory fits the observation that one human can predict how another human will behave in familiar surroundings because they are maintaining a theory of other people’s beliefs. Indeed everyday behavior seems largely based on what an adult person thinks others know, believe or want. But this idea might apply to highly social animals, such as chimpanzees and young children, in general since they need to compete and cooperate effectively with others in their family or group. It would be advantageous to not only to react to what others are doing but also to anticipate what they will do. A practical way of accomplishing this is to act like a junior scientist - observe what others do in particular situations and construct a set of ‘behavioral rules’ that fit the pattern. The reward is that behavioral prediction in similar situation. But a more flexible way to anticipate others behaviors is posit what the goals of others might be and what state of affairs they are trying to bring about. This goes beyond behavioral prediction in similar situations but may also apply to novel situations.

Scientific evidence for a ToM comes from several sources with a particular favorite being the false belief task described by Wimmer and Perner (1983) which seemed to show that a full-fledged TOM doesn’t develop before the age of 3/4. They set up a series of experimental tests in order to check whether children between 3 and 5 years of age were able to attribute a false belief to someone else. In one experiment a child (Maxi), puts chocolate in a blue cupboard and leaves to play. Following this an adult uses part of the chocolate placing the remaining part into a different color (green) cupboard. Wimmer and Perner reported that when asked where ”Maxi” would look for the chocolate after returning most children under four years of age attributed to Maxi what they understood and had seen themselves, that the chocolate is now in the green cupboard. This suggests that a full-fledged TOM doesn’t develop before the age of 3/4. Younger children believe what it is the case, while older kids show a capacity of distinguishing what another mind might believe that is in reality false.

References

,

	Author	volume	Date Value	title	type	journal	titleUrl	doi	note	year
2009 IsAnAgentTheoryofMindToMValuabl	Gary Berg-Cross			Is An Agent Theory of Mind (ToM) Valuable for Adaptive, Intelligent Systems?				10.1145/1865909.1865936		2009