David Brokenshire
.
Research Interests

I have several major areas of interest currently. My main interests are in Artificial Intelligence, particularly Machine Learning. I prefer to do applied work and I do so in Education and Urban Planning, my main areas of interest outside of computing. I am currently researching the application of graphical causal models to educational theory for use in educational technology.

.
.
Graphical Causal Models of Self-Regulated Learning

My M.Sc. thesis topic was Discovering Causal Models of Self-Regulated learning. I investigated the ability to represent Self-Regulated learning theory with an acyclic graphical causal model, and the degree to which such a model could be discovered from purely observational data given current causal structure discovery algorithms. I also evaluated the quantity of data necessary for reliably discovering such a model, the likely error rate, and a method for collecting this data from existing studies using meta-analysis. For further details seem my thesis (pdf).

.
Interoperable User Modelling

This project has been ongoing for the past 2 years. Inspired by the huge amount of information gathered in user modelling initiatives over the last 30 years, and their complete lack of sharing said information I have tried to come up with a technique for briding the gap between disparate models so that information my flow between them and not be lost forever. I believe this is valuable because of reduced expense collecting information, possible solution to the cold start problem of initializing models, and improved reasoning possibilities due to dramatically larger data sets. For more information please see my Publications

.
SPARC ePortfolios

SPARC ePortfolios, developed by a group of student researchers (including myself) under the direction, funding, and guidance of Dr. Vive Kumar. SPARC helps individuals to create electronic portfolios in an open ended format that allows them to be tailored well to particular uses. SPARC tracks artifacts and users and allows for marking, sharing, and feedback on portfolios. SPARC will publish portfolios to the web based on a users preferences and designs. We use SPARC as a platform for research into electronic portfolios and their educational impact as well as a platform in which we can evaluating AI techniques including user models, ontologies, and mixed initiative intelligence.

.
SCORMIT!

The SCORMIT! project is an advanced web spider for use in semi-automatically repackaging legacy eLearning content into SCORM packages which can then be delivered through any SCORM compliant Learning Management System (LMS). I was the initial programmer on the SCORMIT! project which was then handed off to Pat Lougheed