Students Choose Human Counselors Over Algorithms in College Applications, but Not Always

Hemanshu Das, Sofoklis Goulas, Faidra Monachou

Keywords: recommendations; college applications; algorithm aversion; AI in education; experiment

TL;DR: Students often prefer human over algorithmic college recommendations due to perceived intent and goal alignment, with implications for designing hybrid, resource-efficient advising systems.

Abstract: This study examines student preferences for human versus algorithmic recommendations in college applications. Conducted across 14 public high schools in Greece, the experiment reveals that students exhibit aversion to algorithmic recommendations when the recommendation basis is more objective but not when it is most subjective. We find that student perceptions of the recommender's intent strongly drive this aversion, consistently across scenarios and statistical approaches; perceptions of alignment with personal goals, ability, and comprehension also play significant roles. The results further reveal substantial heterogeneity in recommendation adoption rates across several dimensions, including gender, academic performance, adherence to the norm of ``prestige chasing,'' and school urbanicity. Free-text student responses suggest that students seek guidance and information about alternative study options from human counselors but turn to algorithms for recommendations based on grades and admissions chances. Using an optimization approach, we demonstrate how a planner can navigate the heterogeneity in recommendation adoption rates and optimally prioritize the assignment of human versus algorithmic recommenders, under varying social preferences and limited capacity of human counselors. We find that a targeting policy relying on few readily available student and school features can approximate the first-best, personalized targeting policy effectively. These insights underscore the importance of understanding student preferences in designing effective and equitable recommendation systems, and highlight the potential of hybrid approaches that integrate human guidance with algorithmic tools.

Submission Number: 12