Ethical Oversight of Predictive Analytics in Higher Education

A Case Study

Abstract

The use of machine learning (ML) or artificial intelligence (AI) algorithms to make decisions about persons has brought new opportunities and new challenges.  While these tools can increase efficiency or lead to new insights, algorithms deployed in human systems do not necessarily reflect the full complexity, nuances, and ethical frameworks of those systems.  In many cases, the underlying training data may be limited, or including sensitive information, eliciting privacy concerns.  Furthermore, deployment of these models and conveying the results to the stakeholders involved presents challenges and requires appropriate transparency.  In recent years, the use of AI/ML-based predictive analytics in higher education has boomed.  Institutions using these systems should establish appropriate oversight for their use and deployment.  Effective oversight includes not only dissemination of model performance and accuracy, but also how to implement the tool and communicate to stakeholders about results.  Oversight of this type of work should not be left to developers or a few technical people, but should be a concerted effort across a diverse committee with a range of expertise and perspectives.  In this paper we present a case study as a model for effective ethical oversight that addresses many of these concerns.