Research
Theme 1: Models of Cognitive & Autonomous Agents
The concept of software agents emerged in the late 1980s within Artificial Intelligence (AI) and related fields such as software engineering. Software agents act as autonomous, intelligent, and interactive entities operating in distributed and dynamic environments. This paradigm marks an evolution from earlier AI systems, which functioned in isolated, static, and controlled settings.
To function effectively, software agents must exhibit the following key traits:
- Proactiveness: Acting toward achieving defined goals.
- Reactivity: Adapting to changes in evolving environments.
- Social Interaction: Collaborating with other agents and entities in their environment.
- Autonomy: Generating goals, planning actions, and learning from experiences to improve performance over time.
Our research in this area aims to address the following critical questions:
- Decision-Making: How should an agent act based on its knowledge, abilities, and goals within its environment?
- Information Utilisation: How can an agent efficiently acquire and use information to improve decision-making?
- Behaviour Specification: How can we provide executable specifications for an agent’s behaviour?
- Learning Capabilities: How can we integrate learning mechanisms into an agent’s behaviour?
- Domain Adaptation: How can general models of intelligent agents be specialised or adapted for specific applications?
For more information, get a flavour of some of our work with the Knowledge, Goals, Plans (KGP) Model of Agency and the Agent Argumentation Architecture (AAA) model for autonomous agents.
Theme 2: Agent Environments
This research focuses on developing models of interaction to enable the effective deployment of intelligent agents in distributed environments. The agent environment is introduced as a first-class abstraction that provides the necessary conditions for agents to operate, mediates their interactions, and regulates access to resources.
The environment serves distinct and well-defined roles, independent of the agents it hosts:
- It acts as the medium where agents interact, enabling observable and evaluable effects.
- It facilitates coordination by connecting individual agents, which on their own are isolated loci of control.
- It provides constraints and mechanisms for meaningful interactions, ensuring practical applicability.
In essence, the environment integrates agents into a coherent system, enabling both interaction and resource sharing while imposing necessary boundaries.
Our research addresses the following key questions:
- Entities and Models: What entities exist in an agent environment, and how should they be modelled?
- Agent-Environment Interaction: How can agents access or affect/influence their environment?
- Interaction Dynamics: How should interactions between agents and other entities in the environment be represented?
- Distributed Environments: How do we design environments that span distributed networks?
- Application Deployment: How can distributed environments be deployed to support specific application domains?
Some of our work in this theme involves the development of two agent platforms: PROSOCS and its successor GOLEM.
Theme 3: Models of Social & Economic Interaction
This research explores the interplay between individual agent actions (whether human or artificial) and the broader social processes that govern and coordinate these actions. A key focus is on the integration of electronic organisations and institutions as foundational structures within a hybrid agent-human society.
- Organisations provide the structure by defining roles that agents (both artificial and human) assume, specifying responsibilities and expectations within a collective system.
- Institutions establish the protocols, construed here as well-defined rules and norms that regulate how agents interact, collaborate, and adapt their behaviour. Together, organisations and institutions form the backbone of this hybrid society, enabling both coordination and accountability.
Within this framework:
- Roles are central to ensuring agents act purposefully within the context of an organisation, balancing individual goals with collective responsibilities.
- Protocols guide and constrain interactions, ensuring agents cooperate effectively while respecting institutional norms.
- The hybrid society supports dynamic relationships, where humans and agents collaborate, adapt to changes, and contribute to achieving shared social or economic goals.
Our research seeks to address the following critical questions:
- Modelling Metaphors: What metaphors best capture the dynamics of social and economic interactions within a hybrid agent-human society?
- Specification and Implementation: How do we model, specify, and implement social and economic actions governed by organisational roles and institutional protocols?
- Monitoring and Management: What mechanisms ensure agents (artificial or human) fulfill their roles and adhere to institutional norms?
- Norm Adaptation: How can the behaviour of agents be influenced or modified through changes in institutional norms or organisational structures?
- Self-Governance: How do we design infrastructures that allow hybrid agent-human systems to self-regulate through dynamic roles and protocols?
- Simulation and Validation: How can large-scale simulations demonstrate the properties, emergent behaviours, and robustness of hybrid agent-human societies?
By addressing these questions, we aim to create systems where artificial and human agents seamlessly interact within organised, institutionalised frameworks, enabling scalable, adaptive, and self-governing environments. This research advances the vision of a hybrid society where structured collaboration and flexibility drive effective decision-making and problem-solving across complex, distributed systems.
Work in this theme involves the development of games as a metaphor for social interaction and the idea of self-managment and the concept of super-agent as concepts to manage social and economic applications of multi-agent systems.
