Research objectives
To properly manage the high complexity of this research, we frame the problem in three different ways. First, we focus on how collaborative decision processes can be modelled for dynamic environments, while disregarding the specifics of actors and agents, and the organizational dynamics. Second, we address the cognitive aspects of collaboration between actors and agents, while disregarding the specifics of the collaboration framework. Third, we assume a collection of entities endowed with the proper cognitive and collaborative capabilities, and study the mechanisms of self-organization in dynamic environments. This way, highly generic results will be achieved for these three research problems, which form the basis for a sound design of Actor-Agent Communities. We therefore conduct research in the following major areas:

Dynamic collaborative problem solving.

This research will provide a sound theoretical basis for modelling collaborative distributed problem-solving. In a first instance, we restrict the scope of problem solving to decision making in crisis situations. The specific aspects of crisis response (environmental dynamics, time pressure, incomplete or unreliable information, limited computational resources, etc.) will impose additional requirements on the decision making process, and will provide a realistic model of decision making entities with bounded rationality. In a later phase we will generalize these principles for a wider range of problems.

Definition and design of autonomous complementary-cognitive systems.

The agents in an AAC will complement the capabilities of human experts in two ways: they will perform those (information processing and analysis) tasks where humans perform poorly, and will simulate, verify and validate the decisional options generated heuristically by humans. For attaining these objectives, the following three research challenges will be addressed:
a)    First, the degradation of the quality of decision making by humans will be analyzed under a selection of critical circumstances (such as time pressure, incomplete, unreliable, or conflicting information, information overflow, communication delays, non-deterministic changes in hypothesis, high complexity of options space, etc.), for some of the most demanding activities in typical decision making processes (such as information filtering and fusion, options generation, and estimation of options’ utility);
b)    A selection of those activities that are most prone to error, or experienced as bottlenecks by humans, will be modelled as complementary-cognitive process primitives;
c)    Finally, complementary cognitive systems will be designed that: (i) approximate one or more complementary cognitive primitives; (ii) are capable of identifying sub-problems that they can solve, and (iii) adjust their behaviour (e.g. by “tuning” the primitives’ algorithms) based on different local and global feedback signals.

Principles and mechanisms for purposeful self-organization.

In order to ensure quick response, high efficiency, and flexibility of AACs, the functional relationships between entities are not defined at design time. Instead, a collection of entities behave autonomously in pursuit of a long-term goal, expressed typically as a state vector. Purposeful self-organization will enable entities to discover deviations from the goal due to events in the environment, and spontaneously start collaborating with those entities that could contribute to restoring (or approaching) the target state. This research is especially challenging since the self-organization has to take place between entities with limited capabilities and partially different world models.

A successful proof of the theoretical concepts developed in this research will lead to novel approaches to the development of complex hybrid (i.e. human-machine) systems, whose main characteristic is a fast and flexible integration of specialised systems (or human experts) for attaining, at runtime, the problem-solving capability that is required for dealing with otherwise intractable problems. These new concepts will eventually lead to a paradigm shift whose essence lies in a fundamental change of the role and positioning of computing systems with respect to human activities. Computing systems will thus evolve from their current status of support systems to active collaborators in complex human activities. The theoretical foundations of AAC will constitute a solid base for the development of a coherent long term approach to the design and realisation of a new breed of self-organizing, adaptive, and proactive ICT systems. Actor-agent communities are the solution; the only viable answer to contemporary and future complex demands on human and artificial systems’ performance.

For more information please contact Kees Nieuwenhuis