Sociologist Patricia Hill Collins and Columbia Law School Professor Kimberlé Crenshaw developed the notion of intersectionality to describe the ways in which multiple, multi-faceted identities associated with race, ethnicity, gender, class, sexuality and ability (among others) compound marginalisation (and privilege) in various contexts [1]. In contexts where an individual’s identities are in alignment with the normative or typical profile, all of the goods and services created in that context are likely to be accessible and useful for that individual. When the opposite is true, and an individual’s identities are not in alignment with the norm, one can expect that the goods and services created for the normative profile may not be as accessible or useful for that individual [2]. What are the dangers of serving the normative profile in education? How do we get past it? What are some of the limitations of those solutions and what should we be designing for the future? In this workshop, we are asking participants to consider the implications of their work for the goals and objectives of equality, diversity and inclusion in education.
For example, several recent studies evaluate the impact of delivering learning analytics based interventions with students, which shows a promising progress beyond pure development of predictive models [4,5,6]. However, what we don’t know is how these models impact different subgroups and whether they decrease or highlight existing gaps [3]. This is important because divergences in the learning and behavioural patterns of minority groups (students of a minority ethnicity, for example, or students with disabilities) may not be given the same relevance than those of the majority group. One of the first studies analysed the impact of using Predictive Learning Analytics by teachers on students coming from different socio-ecocomic backgrounds and students from different ethnicities [10].
Another emerging topic in the Machine Learning community is investigation of fairness of predictive models. Several papers from the Learning Analytics community have been published in the last two years [7, 9], describing methodologies for making models fair. But the number of fairness metrics is huge and sometimes contradictory, and various papers select different metrics to evaluate fairness. On the other hand, no research has discussed implications of deploying models that are fair on decreasing the existing gaps or on equity. For example, what is the impact of removing the ethnicity/race factor from the predictive models used to target at-risk students on the completion rates, overall score or student satisfaction ?
*All submissions should be made through Easychair
Jiten Patel is an author and award winning Inclusion and Diversity strategist and practitioner, helping to maximise the value of diversity through conscious inclusion strategies designed to be integrated with:
His book, Demystifying Diversity (co-authored by Gamiel Yafai), was described by People Management as an important contribution to the field.