An AI prompt can expose a blind spot in college data
A study used four real college cases to test how people read data with help from a guided AI chat. The prompts led them to notice local factors behind a first reading. That is a useful pause before a decision. It does not prove fairer outcomes or apply to K-12 schools without more study.
What the source reports
Francielle Marques and colleagues studied how people read college data in four real cases. They used a guide called FACTRIA to group possible sources of bias. The groups include steps in data analysis, the college setting, courses and student groups. A chat tool used that guide to prompt questions. The qualitative study reports that people noticed factors they had missed in their first reading. That is an observation about how they thought through the cases, not a measured improvement in decisions. The abstract does not show that the tool cut gaps between groups or made an unfair rule fair. These cases come from higher education. They do not show how the same method would work in K-12 schools.
Original source
- Title
- Responsible Institutional Analytics: Interpreting Bias with AI Support
- Author
- Francielle Marques, Ariel Ortiz-Beltrán, Ishari Amarasinghe, Davinia Hernández-Leo
- Publication
- arXiv
- Date
- Wednesday, October 7, 2026