Research Interests

I am interested in developing methods that can be used to reduce the workload or the cognitive load of the user. In particular I am interested in two areas:

1. Building models (of user behaviours or of domain knowledge) that can then provide information to enable intelligent interactions,
2. Using/adapting AI techniques to improve system performance (e.g. runtime and decision/recommendation accuracy).

As a postdoctoral researcher at the University of British Columbia (Canada), I focused on using a dynamic decision network to achieve real-time modelling the user's affective state, thus enabling an intelligent agent to communicate more effectively.

During my PhD, sponsored by BTExact Technologies (UK), I focused on reducing search space complexity and working with domain knowledge when I began to look at using machine learning techniques for scheduling prediction. The problem to solve was framed as an Inductive Logic Programming problem (i.e. automatically constructing Prolog rules to explain the given data set, and ranking the solution according to a given heuristic), but the search space of possible solutions was so large that it took several hours to find a solution. The contribution of my PhD was to exponentially reduce the search space (and therefore the processing time) by utilizing a divide-and-conquer technique that was based on the domain-specific independence assumptions.

Unfortunately, due to my sponsoring company's ownership of the intellectual rights to the solutions produced and their wish to patent the work, I was unable to publish anything until the patents had been applied for. This situation has now been resolved (with the patent applications complete).