Dropout-risk-and-grade-prediction
“AI can fundamentally violate the principle of equal access. Universities in the U.S. are using deterministic algorithmic systems to recommend applicants they should admit. These are often custom-built to meet the school’s preferences, and have a host of issues that can lead to discrimination, including use of historical data of previously admitted students to inform the model. Since many elite universities have historically been attended by wealthy white males, any model that uses these data risks perpetuating past trends.105 Such systems will likely employ ML in the future, which would make bias harder to detect. This could result in universities discriminating under the guise of objectivity. Looking forward: If AI is used to track and predict student student performance in such a way that limits the eligibility to study certain subjects or have access to certain educational opportunities, the right to education will be put at risk. Given the growth of research into early childhood predictors of success, it is likely that such a system could be used to restrict the opportunities of students at increasingly younger ages, resulting in significant discrimination, with students coming from underprivileged backgrounds ultimately being denied opportunities because people from that background tend to have more negative outcomes. Such a system would ignore the students that overcome adversity to achieve academic and professional success, and would entrench existing educational inequalities.” ([Access Now, 2018, p. 28](zotero://select/groups/4907410/items/KIU6F9PR)) ([pdf](zotero://open-pdf/groups/4907410/items/598C5P37?page=28&annotation=DR727J3N))