Princeton study finds human judgment essential for college advising
Research from Princeton Engineering indicates that artificial intelligence cannot fully replace college advisers because human expertise drives interventions in Georgia State University meetings.
Study analyzes advising actions
Lydia Liu, assistant professor of computer science at Princeton Engineering, led the investigation into algorithmic advising limits. Over a six-year period, the group examined 5,269 sessions that included 8,385 advising actions at Georgia State University. Their paper presented last year at the Equity and Access in Algorithms conference shows advisers used unique information in two out of three cases. These specific actions included rearranging schedules for health issues or offering encouragement to struggling math students.
Grades improved slightly with AI
Students enrolled in the AI-assisted program achieved grades 0.16 points higher on average than the control group. Graduation rates were also 7 percent greater for those receiving algorithmic support compared to peers without it. The randomized controlled trial confirmed that while predictive analytics boosted success, human interaction remained critical for complex student needs. Liu noted these tools mirror systems used to predict hospital sepsis or criminal recidivism risks today.
Human relationships matter most
Liu emphasized that advising relies on relational connections where students feel truly seen by their mentors. She warned that automation removing this human element would be sad for the educational experience overall. The research concludes that human intervention determines the downstream impact of any algorithmic prediction made by software. As AI expands into health and justice fields, assessing required human expertise remains critical for effective implementation.
Reported by one outlet
Only one outlet has published this. Nothing here has been checked against a second report, so read it as that outlet's account and follow the link below for the original.
Reported by
1 independent outlet. Headline as published. Links open the original report.