Joanna J. Bryson

Semantics derived automatically from language corpora contain human-like biases

Machines learn what people know implicitly AlphaGo has demonstrated that a machine can learn how to do things that people spend many years of concentrated study learning, and it can rapidly learn how to do them better than any human can. Caliskan et al. now show that machines can learn word associations from written texts and that these associations mirror those learned by humans, as measured by the Implicit Association Test (IAT) (see the Perspective by Greenwald). Why does this matter? Because the IAT has predictive value in uncovering the association between concepts, such as pleasantness and flowers or unpleasantness and insects. It can also tease out attitudes and beliefs—for example, associations between female names and family or male names and career. Such biases may not be expressed explicitly, yet they can prove influential in behavior. Science, this issue p. 183; see also p. 133 Computers can learn which words go together more or less often and can thus mimic human performance on a test of implicit bias. Machine learning is a means to derive artificial intelligence by discovering patterns in existing data. Here, we show that applying machine learning to ordinary human language results in human-like semantic biases. We replicated a spectrum of known biases, as measured by the Implicit Association Test, using a widely used, purely statistical machine-learning model trained on a standard corpus of text from the World Wide Web. Our results indicate that text corpora contain recoverable and accurate imprints of our historic biases, whether morally neutral as toward insects or flowers, problematic as toward race or gender, or even simply veridical, reflecting the status quo distribution of gender with respect to careers or first names. Our methods hold promise for identifying and addressing sources of bias in culture, including technology.

Agent-based modelling as scientific method: a case study analysing primate social behaviour

A scientific methodology in general should provide two things: first, a means of explanation and, second, a mechanism for improving that explanation. Agent-based modelling (ABM) is a method that facilitates exploring the collective effects of individual action selection. The explanatory force of the model is the extent to which an observed meta-level phenomenon can be accounted for by the behaviour of its micro-level actors. This article demonstrates that this methodology can be applied to the biological sciences; agent-based models, like any other scientific hypotheses, can be tested, critiqued, generalized or specified. We review the state of the art for ABM as a methodology for biology and then present a case study based on the most widely published agent-based model in the biological sciences: Hemelrijks DomWorld, a model of primate social behaviour. Our analysis shows some significant discrepancies between this model and the behaviour of the macaques, the genus used for our analysis. We also demonstrate that the model is not fragile: its other results are still valid and can be extended to compensate for these problems. This robustness is a standard advantage of experiment-based artificial intelligence modelling techniques over analytic modelling.

The Insufficiency of Formal Design Methods " The Necessity of an Experimental Approach - for the Understanding and Control of Complex MAS

We highlight the limitations of formal methods by exhibiting two results in recursive function theory: that there is no effective means of finding a program that satisfies a given formal specification; or checking that a program meets a specification. We exhibit a simple MAS which has all the power of a Turing machine. We argue that any pure design methodology will face insurmountable difficulties in todayýs open and complex MAS. We recommend instead a methodology based on experimental method ¿ scientific foundations for MAS construction and control.

The behavior-oriented design of modular agent intelligence

Agent UML is a graphical modeling language based on UML. Like UML, Agent UML provides several types of representation covering the description of the system, the components, the dynamics of the system and the deployment. Multiagent system designers already use Agent UML to represent interaction protocols [13, 2]. Since agents and objects differ on several points, UML class diagram has to be modified for describing agents. The aim of this paper is to present how to extend UML class diagrams in order to represent agents. We then compare our approach to Bauers approach [1].

Agent-Based Composite Services in DAML-S: the Behavior-Oriented Design of an Intelligent Semantic Web

Many researchers are working towards the goal of a semantic Web — a Web that is unambiguously computer interpretable, and thus very accessable to artificial intelligence. A semantic Web would allow artificial agents to do the work of searching for and utilizing services required by humans or organizations. DAML-S is a Web service ontology intended to facilitate the semantic Web by describing the properties and capabilities of Web-accessible services in an unambiguous, computer-interpretable form. In this chapter, we propose that an important new perspective on the semantic Web can be obtained by regarding its content as behavioral intelligence. The services encoded in DAML-S can then be viewed as specifications either for extensions of the user-owned agents attempting to exploit the services, or as independent, collaborative agents that can be ‘awakened’ to assist the user agents. We draw on our experience in agent development to elaborate the specification, particularly of the process ontology of DAML-S, in order to support this vision.

Agent Architecture as Object Oriented Design

Improving the development of agent intelligence requires improving the mechanisms of that development. This paper explores the application of an established software methodology, object-oriented design, to agent development in two ways. We present a distributed agent architecture, Edmund, and describe first its own object-oriented structure. Then we relate the methodology for developing agent behaviors under Edmund. We explain how this methodology exploits key aspects of object-oriented design, particularly the development of the class hierarchy, as a prototype for agent design.

Introduction. Modelling natural action selection

Action selection is the task of resolving conflicts between competing behavioural alternatives. This theme issue is dedicated to advancing our understanding of the behavioural patterns and neural substrates supporting action selection in animals, including humans. The scope of problems investigated includes: (i) whether biological action selection is optimal (and, if so, what is optimized), (ii) the neural substrates for action selection in the vertebrate brain, (iii) the role of perceptual selection in decision-making, and (iv) the interaction of group and individual action selection. A second aim of this issue is to advance methodological practice with respect to modelling natural action section. A wide variety of computational modelling techniques are therefore employed ranging from formal mathematical approaches through to computational neuroscience, connectionism and agent-based modelling. The research described has broad implications for both natural and artificial sciences. One example, highlighted here, is its application to medical science where models of the neural substrates for action selection are contributing to the understanding of brain disorders such as Parkinsons disease, schizophrenia and attention deficit/hyperactivity disorder.

Architectures and Idioms: Making Progress in Agent Design

This chapter addresses the problem of producing and maintaining progress in agent design. New architectures often hold important insights into the problems of designing intelligence. Unfortunately, these ideas can be difficult to harness, because on established projects switching between architectures and languages carries high cost. We propose a solution whereby the research community takes responsibility for re-expressing innovations as idioms or extensions of one or more standard architectures. We describe the process and provide an example -- the concept of a Basic Reactive Plan. This idiom occurs in several influential agent architectures, yet in others is difficult to express. We also discuss our proposals relation to the the roles of architectures, methodologies and toolkits in the design of agents.

Learning a real time grasping strategy

Real time planning strategy is crucial for robots working in dynamic environments. In particular, robot grasping tasks require quick reactions in many applications such as human-robot interaction. In this paper, we propose an approach for grasp learning that enables robots to plan new grasps rapidly according to the objects position and orientation. This is achieved by taking a three-step approach. In the first step, we compute a variety of stable grasps for a given object. In the second step, we propose a strategy that learns a probability distribution of grasps based on the computed grasps. In the third step, we use the model to quickly generate grasps. We have tested the statistical method on the 9 degrees of freedom hand of the iCub humanoid robot and the 4 degrees of freedom Barrett hand. The average computation time for generating one grasp is less than 10 milliseconds. The experiments were run in Matlab on a machine with 2.8GHz processor.

Toward behavioral intelligence in the Semantic Web

Realizing the Webs full potential will require the development and support of agents that function as schedulers, planners, and searchers who, with minimal direction, can serve as an omnipresent staff of advisers, secretaries, brokers, and research assistants. Electronic commerce has brought this capability tantalizingly near. Organizations and individuals have connected an enormous variety of products and services to the Internet, making them accessible to other programs through simple communication protocols. Now the AI community must determine how it can build intelligent agents to exploit these services. One strategy would change the Web itself, making it accessible to existing AI modeling, and reasoning techniques. In this semantic Web, service and content providers would mark pages in accordance with standardized conventions designed to reduce ambiguity and make automated reasoning easier. The paper considers the development of a distributed intelligence and bringing agents to the Web. It discusses DAML-S which provides support for composite services, combinations of simpler services, or behaviors, and the coordination mechanisms, or reactive plans, used to combine those behaviors.