Learning while optimizing a design: a situated agent-based design interaction tool

This paper presents an approach that enables a design tool to learn through its use. This approach uses a situated agent that wraps around a design tool and constructs concepts from interactions between the agent, the design problem and the use of the tool. The concept underpinning this research is founded on the ideas of “situatedness”. Situatedness involves the context and the observer’s experiences, as well as the interactions between them. A situated agent uses a constructive memory system to learn concepts while interacting. When applied to design optimization fields, such a situated agent-based design interaction tool demonstrates adaptive behaviours in reflexing, reacting and reflecting to various design optimization scenarios based on the incrementally learned knowledge structures and their intentional descriptions – the conceptual knowledge. Design tools no longer simply respond to their use based on a priori knowledge. Experiments show that a situated agent-based design optimization tool can learn and develop adaptive behaviours to support design decision-making. A prototype system is presented.