Managing knowledge navigation in design with mixed-initiative dialogue

In generative design systems, the interaction between the designer and the generative formalism is typically transactional. The system responds to a single query or utterance with one or more results, which when sent back to the user define the transaction. A neat conceptual separation exists between user tasks and system tasks, rendering the modelling of joint responsibility for generation difficult. A mixed-initiative model of interaction is necessary to support forms of dialogue where both the designer and the formalism may share responsibility over the same task. Mixed-initiative provides a robust basis for combining design interaction with automated or intelligent systems. In such systems, the role of both user and formalism are integrated into an incremental-based model, where the user and the formalism can share and input data through a common shared resource, on a common shared task. Generative design systems provide formal representations that aid designers in exploring spaces of designs. We develop a formal representation of design spaces and unification-based algorithms for generation in the specification and implementation of design space explorers, computer programs for constructing, navigating and exploring spaces of design problems using feature structure theory. They adopt the unification view of incremental generation and provide a representation for the constraints on a design space, relate those constraints to the generation of partial designs and supports intermediate representations of design states. Extending this work with exploration systems, we report on the development of a graphical notation to support mixed-initiative dialogue in generative design systems. We demonstrate the use of this notation in design space exploration and discuss its implications for interaction with generative design systems.