Stephen Childs was hired by the University of Calgary to develop an individual-level predictive model mapping students' decisions to attend the University. In his experience, the higher education sector was slow to use all the data it has available, but this is now changing.
As interest in making use of organizational data grows, staff must consider how these models will be used, and any problems that could arise. When individual predictions become the basis for decisions, how do we ensure our algorithms don't make existing problems worse? A framework for handling these issues now will let organizations handle these issues in a way that is consistent with their values.
Given the culture of today's institutions, and the success of predictive analytics in other fields, there is no doubt that these tools will be used. These techniques can improve student success and the competitiveness of educational organizations, but the benefits should not be gained at the expense of individuals within the system. This talk will propose a set of best practices for using institutional data for predictive modelling to address equity, privacy and other concerns. We must start thinking of this now, before other practices become entrenched.