Reviewed by
Gennaro Di Tosto
ISTC-CNR, Rome
It was 1950 and Alan Turing had just published the paper "Computing machinery and intelligence" (Turing 1950). There we find the idea for a test to evaluate the performance of a computer program attempting to replicate human intelligence: a human judge is allowed to communicate with two parties - one a machine, the other a person - by exchanging text messages with them. The task of the judge is to distinguish the person from the machine; if he fails, the machine will have passed the test. This, in short, is the Imitation Test, or Turing Test, as it is commonly called. In 1991 it was turned into an annual AI competition, the Loebner Prize for Artificial Intelligence, awarded to the first computer program whose responses are indistinguishable from a human's. Turing himself made the following prediction: "that in about fifty years' time it will be possible to programme computers [...] to make them play the imitation game so well that an average interrogator will not have more than 70 per cent chance of making the right identification after five minutes of questioning". Although that time has passed, still no one has won the prize. Imagine, though, that a computer able to pass the test exists. Would you be willing to say that its ability to convincingly imitate humans in conversation is proof that a machine can "think"? And what cognitive abilities must be reproduced in order to build an artificial intelligence able to compete with humans? Starting from the original argument expressed in "Computing machinery and intelligence", and compiling the list of objections discussed in there by Turing, Jay Friedenberg - who is director of the cognitive science program at Manhattan College - provide an introduction to the philosophical problems of building a thinking machine, while giving examples of the achievements of artificial intelligence and robotics in modelling natural cognition.
"Artificial psychology" follows the organization of a regular textbook in introductory psychology. In each one the contents are presented avoiding any technical language, and the arguments are often illustrated by clear examples or references to popular culture, like movies, novels, etc. The book does not presuppose any previous knowledge from the reader and thus represents an ideal starting point for undergraduate students and laymen interested in pondering the questions related to the uniqueness of human intelligence and behaviour, the possibility of building artificial systems able to display them, and the consequent philosophical, social or ethical implications.
Following an engineering approach, the book explores the attempts made to reproduce human cognitive abilities, and uses them to offer an operational definition of the process involved in the areas of perception and action, learning, memory, language, intelligence, consciousness, creativity, motivation and emotion. Along the way the reader is introduced to the idea of functionalism and the argument of the "Chinese room"; the perspective on the development of cognitive faculties and the role of the body as a mediator between the environment and the mind given by the principles of situatedness and embodiment. The overall goal, expressed by the author in the introduction, is to compose all these elements in the portrait of an "artificial person". Updating the Turing Test with the advancements made in so far by the science of the artificial, an artificial person should be able to display human-like abilities not only in use of language, but in all kinds of interactions with its (social) environment.
The last chapter concerns social behaviour, with two sections about virtual and robots societies, respectively. Computational social scientists, however, will be disappointed to discover that it only contains references to "boids" agents (Reynolds, 1987) as an example of flocking behaviour and swarm intelligence, to stigmergy as a principle for the emergence of intelligent behaviour, and to the analysis of coordination rules as constraints over an actions repertoire made by Shoham and Tenneholtz (1995). Given the emphasis usually reserved to cognitive plausibility in other parts of the book, the choice of examples confined to the behaviouristic level is noticeable.
The rest of the chapter continues with the analysis of two issues:
* Human-Machine Interaction: In the design of machines able to move in an open environment and to interact with people researchers are confronted with both engineering and psychological problems. Studies have shown that the perception of technological devices is affected by subtle cues that can alter the emotional response of the user. The social perception of computers is also influenced by our tendency to attribute human-like qualities to artefacts; this attribution usually extends to mental states, as it implies the prediction of the actions of a system based on concepts like beliefs, desires, and intentions (Dennett 1981). The current attempts to endow robots with social skills described in the book are an important step toward the acceptance of intelligent technology among non-expert users. Since the growing number of elderly people in western societies have increased the demand of assistive technology, these issues are vital to ensure trust when people with particular needs delegate parts of their daily tasks and activities to automated entities.
* The relationship between ethics and AI: the issue of norms --legal, social and/or moral-- has always received great attention in the fields of multi-agent systems (Conte and Castelfranchi 1994) and in robotics (Wallach and Allen 2008). On this regard the book explore two directions: a) the implementation of machines capable of moral decision-making; and b) the extension of moral or ethical rights to machines capable of autonomous behaviour. This second perspective might appear farfetched given the present technological scenarios, but its philosophical interest is directly connected to the interpretation of the concepts of autonomy and responsibility.
So, should we be prepared to observe the birth/creation of the first artificial person? Even with a technology capable of reproducing the computing power of the human brain, we might still be unable to build a machine as complex as ourselves. However, we could still gain something from the endeavour: building an artificial person might turn out to be an impossible task, but it can help to develop a better understanding of the mind. An epistemological perspective that concerns all the sciences of the artificial.
ReferencesDENNET D (1981) 'Intentional systems." In Haugeland J (Ed.), Mind design: Philosophy, psychology, artificial intelligence. Cambridge, MA: MIT Press, pp. 220-242.
TURING A (1950) Computing machinery and intelligence, in Mind, 49, pp. 433-460.
WALLACH W and ALLEN C (2008) Moral Machines. Teaching Robots Right from Wrong. New York, USA: Oxford University Press.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2009