Emma Norling

Folk Psychology for Human Modelling: Extending the BDI Paradigm

BDI agents have been used with considerable success to model humans and create human-like characters in simulated environments. A key reason for this success is that the BDI paradigm is based in folk psychology, which means that the core concepts of the agent framework map easily to the language people use to describe their reasoning and actions in everyday conversation. However there are many generic aspects of human behaviour and reasoning that are not captured in the framework. While it is possible for the builder of a specific model or character to add these things to their model on a case by case basis, if many models are to be built it is highly desirable to integrate such generic aspects into the framework. This paper describes an approach to extending the BDI framework to create an enhanced framework for human modelling. It draws upon the folk psychological roots of the framework to create the extension, maintaining the mapping between the knowledge representation in the framework and the natural means of expressing expert knowledge. The application of this approach is illustrated with an extension to support human decision making.

Multi-Agent-Based Simulation III

Today’s application tend to be more and more decentralised, pervasive, made of autonomous entities or agents, and have to run in dynamic environments. Applications tend to be social in the sense that they enter into communication as human people, and engage into discovery, negotiation, and transactions processes; autonomous programs run their own process, interact with other programs when necessary, but each program lives its life, and a global behaviour emerges from their interactions, similarly to what can be observed in natural life (physical, biological or social systems). Tomorrow’s applications are more and more driven by social interactions, autonomy, and emergence, therefore tomorrow’s engineering methods have to take into account these new dimensions. Traditional software engineering will not be adapted to this new kind of applications: they do not scale, they do not enable the definition of local behaviours and drawing of conclusions about global behaviours. The scope of this paper is to determine today’s and tomorrow’s application domains, where such a sociological behaviour can be observed. Starting from the observation of natural life (natural mechanisms used for self-organisation, for anonymous communication, etc), we then discuss how these natural mechanisms can be translated (or have an artificial counterpart) into electronic applications. We also consider software engineering issues, and discuss some preliminary solutions to the engineering of emergent behaviour.

Embodying the JACK Agent Architecture

Agent-based models of human operators rarely include explicit representations of the timing and accuracy of perception and action, although their accuracy is sometimes implicitly modelled by including random noise for observations and actions. In many situations though, the timing and accuracy of the persons perception and action significantly influence their overall performance on a task. Recently many cognitive architectures have been extended to include perceptual/motor capabilities, making them embodied, and they have since been successfully used to test and compare interface designs. This paper describes the implementation of a similar perceptual/motor system that uses and extends the JACK agent language. The resulting embodied architecture has been used to compare GUIs representing telephones, but has been designed to interact with any mouse-driven Java interface. The results clearly indicate the impact of poor design on performance, with the agent taking longer to perform the task on the more poorly designed telephone. Initial comparisons with human data show a close match, and more detailed comparisons are underway.

Multi-Agent-Based Simulation VIII

MABS Celebrates Its 10th Anniversary!.- MABS Celebrates Its 10th Anniversary!.- Architectures.- System Issues in Multi-agent Simulation of Large Crowds.- Middleware Support for Performance Improvement of MABS Applications in the Grid Environment.- E Pluribus Unum: Polyagent and Delegate MAS Architectures.- A Multi-agent Model for the Micro-to-Macro Linking Derived from a Computational View of the Social Systems Theory by Luhmann.- Teams, Learning, Education.- Agent-Based Simulation of Group Learning.- An Agent-Based Model That Relates Investment in Education to Economic Prosperity.- Economy, Trust and Reputation.- Trust-Based Inter-temporal Decision Making: Emergence of Altruism in a Simulated Society.- Multi-agent Model of Technological Shifts.- Beyond Accuracy. Reputation for Partner Selection with Lies and Retaliation.

Towards the evolution of social structure

To what extent can social structure result from evolutionary processes as popposed to being deliberately organised? To begin to answer this questions five different but releated social simulations are reviewed, and a map of which mechanisms might results in what structures under what conditions being started. These show that different structures can be brought about by evolutionary processes based on the abilities and propensities of the individuals. The article ends with some challenges—to construct a credible simulations of more sophisticated structures: social group selection and self-organised value chains.

Informal Approaches to Developing Simulation Models

This chapter describes an approach commonly taken by most people in the social sciences when developing simulation models instead of following a formal approach of specification, design and implementation. What often seems to happen in practice is that modellers start off in a phase of exploratory modelling, where they don’t have a precise conception of the model they want but a series of ideas and/or evidence they want to capture. They then may develop the model in different directions, backtracking and changing their ideas as they go. This phase continues until they think they may have a model or results that are worth telling others about. This then is (or at least should be) followed by a consolidation phase where the model is more rigorously tested and checked so that reliable and clear results can be reported. In a sense what happens in this later phase is that the model is made so that it is as if a more formal and planned approach had been taken.

Contrasting a system dynamics model and an agent-based model of food web evolution

An agent-based model of food web evolution is presented and contrasted with a particular system dynamics model. Both models examine the effects of speciation and species invasion of food webs, but the agent-based approach focuses on the interactions between individuals in the food web, whereas the system dynamics approach focuses on the overall system dynamics. The system dynamics model is an abstract model of species co-evolution that shows similar characteristics to many natural food webs. The agent-based model attempts to model a similarly abstract food web (in which species are characterised by abstract features that determine how they will fare against any other species). The ultimate aim of this exercise is to explore the many of the assumptions inherent in the system dynamics model; the current challenge is to simply replicate the system dynamics results using agent-based modelling. Preliminary studies have revealed some underlying assumptions in the system dynamics model, as well as some intrinsic difficulties in linking the two different approaches. The paper discusses the key difficulties in linking these different types of models, and presents some discussion of the limits and benefits benefits that each approach may bring to the analysis of the problem.

Integrating learning and inference in multi-agent systems using cognitive context

Both learning and reasoning are important aspects of intelligence. However they are rarely integrated within a single agent. Here it is suggested that imprecise learning and crisp reasoning may be coherently combined via the cognitive context. The identification of the current context is done using an imprecise learning mechanism, whilst the contents of a context are crisp models that may be usefully reasoned about. This also helps deal with situations of logical under- and overdetermination because the scope of the context can be adjusted to include more or less knowledge into the reasoning process. An example model is exhibited where an agent learns and acts in an artificial stock market.

Cross-Disciplinary Views on Modelling Complex Systems

This paper summarises work within an interdisciplinary collaboration which has explored different approaches to modelling complex systems in order to identify and develop common tools and techniques. We present an overview of the models that have been explored and the techniques that have been used by two of the partners within the project. On the one hand, there is a partner with a background in agent-based social simulation, and on the other, one with a background in equation-based modelling in theoretical physics. Together we have examined a number of problems involving complexity, modelling them using different approaches and gaining an understanding of how these alternative approaches may guide our own work. Our main finding has been that the two approaches are complimentary, and are suitable for exploring different aspects of the same problems.

Multi-Agent-Based simulation: why bother?

This years MABS workshop was the sixth in a series which is intended to look at “using multi-agent models and technology in social simulation,” according to the the workshop series homepage [1]. We feel that this is an appropriate time to ask the participants and the wider community what it is that they hope to gain from this application of the technology, and more importantly, are the tools and techniques being used appropriate for achieving these aims? We are concerned that in many cases they are not, and consequently, false or misleading conclusions are being drawn from simulation results. In this paper, we focus on one particular example of this failing: the consequences of the inappropriate use of numbers. The translation of qualitative data into quantitative measures may enable the application of precise analysis, but unless the translation is done with extreme care, the analysis may simply be more precisely wrong. We conclude that as a community we need to pay careful attention to the tools and techniques that we are using, particularly when borrowing from other disciplines, to make sure that we avoid similar pitfalls in the future.