Here's a closer look at the research that appeared in July in AERA Open
<h2 id="what-the-researchers-found-about-ai -algorithms-ability-to-predict-student-success-3″>What researchers discovered about ai algorithms' ability to predict student success
“Our study was motivated by the increasing use of machine learning (ML) and artificial intelligence (ai ) in college and university operations, including to improve student success outcomes,” says Denisa Gándara, co-author of the study and an adjunct professor in the College of Education at the University of Texas at Austin. “Colleges and universities leverage ML to predict student success outcomes, such as persistence and graduation rates. They do so for a variety of reasons, including to make admissions decisions or to target interventions.”
Gándara adds that she and her co-authors were inspired to undertake this study because research in other fields, such as health care and criminal justice, had previously revealed a bias against socially marginalized groups.
“In higher education, these biases could exacerbate existing social inequalities and influence crucial decisions such as admissions and the allocation of student support services,” she says.
The models that Gándara and his colleagues studied incorrectly predicted that a student would not graduate 12% of the time if the student was white and 6% of the time if the student was Asian. If the student was Hispanic, the models were wrong 21% of the time, and if the student was black, they were wrong 19% of the time.
A similar pattern was followed for predicting success. The models incorrectly predicted success for white and Asian students 65% and 73% of the time, respectively, compared with just 33% of the time for black students and 28% for Hispanic students.
How was the research conducted?
For the study, researchers tested four predictive machine learning tools that are widely used in higher education. The tools were trained on 10 years of data from the U.S. Department of Education's National Center for Education Statistics, including 15,244 students.
“We used a large, nationally representative data set that includes variables that are routinely used to predict college student success,” says Gándara. These include demographic and academic performance variables.
The researchers used 80% of this data to train the models and then tested each model's ability to make predictions with the remaining 20% of the data. The process likely meant that the models were receiving better training data than what many universities use.
“Schools and universities typically use smaller administrative data sets that include data on their own students,” says Gándara. “There is wide variability in the quality and quantity of data used at the institutional level.”
What is the impact of this research?
Given how entrenched ai tools are becoming in education, the implications of this research are significant. “Predictive models are not limited to universities, they are also widely used in primary and secondary schools,” says Gándara. “If predictive models are used in college admissions decisions, students from racial minorities could be unfairly denied admission if the models predict lower success rates based on their racial categorization.”
Beyond admissions, these inaccuracies can create other problems. “In primary and secondary or higher education, there is a risk that students from racially minority groups are steered into less challenging educational paths,” says Gándara.
Furthermore, inaccurate predictions about success could lead schools or educators to provide increased resources to students who may not need those interventions.
“The evidence of algorithmic bias and its various implications underscores the importance of educating end users, including admissions officers, academic advisors, and faculty members or professors, about the potential for bias,” Gándara said. “Knowledge of algorithmic bias, its direction, and the groups affected can help users contextualize predictions and make more informed decisions.”
NEWSLETTER
