CourseHero

AI & Exam Preparation Research Project

A user research and product strategy project analyzing how USC students use AI for exam prep and identifying opportunities for Course Hero to build professor-specific, AI-powered study tools.

Explore CourseHero

My Role

Product Researcher & Product Strategist
(I owned the problem framing, research execution, synthesis, and product recommendations end-to-end.)

Problem

AI study tools are becoming a daily part of how students prepare for exams — but students don’t fully trust them.
Current tools often fail in accuracy (especially for STEM), don’t match professor-specific exam formats, and overwhelm students with irrelevant or generic explanations. As a result, AI creates friction instead of confidence during exam prep.

Course Hero had an opportunity to understand how students actually use AI, where trust breaks down, and how AI could meaningfully improve exam preparation rather than just provide answers.

Overview

I led an end-to-end research and product strategy project analyzing how USC students use AI for exam preparation, where existing tools fall short, and how Course Hero could leverage AI to deliver more personalized, reliable, and exam-aligned study support.

Using mixed-methods research (surveys + in-depth interviews), I translated real student behaviors and frustrations into concrete product opportunities — framing AI not as a shortcut, but as an adaptive study companion.

What I Did

  • Designed and conducted a mixed-methods research study (survey + interviews)

  • Collected quantitative insights from 39 USC students across majors and class years

  • Conducted 5 in-depth interviews to uncover deeper study behaviors and trust issues

  • Synthesized qualitative and quantitative findings into clear user pain points

  • Translated insights into actionable AI product concepts tailored to Course Hero

Key Insights

  • Students use AI as an interactive tutor, not a replacement for studying

  • Trust breaks down in STEM subjects due to accuracy errors

  • Generic AI answers fail because they don’t align with professor-specific exam styles

  • Students want exam-level practice, not surface explanations

  • Over-explaining and lack of citations reduce confidence in AI responses

  • AI is most valuable when it saves time and reinforces conceptual understanding

Product Opportunities Identified

  • Professor-Specific AI Study Mode trained on uploaded course materials

  • Adaptive practice exams that mirror real exam difficulty and format

  • Flexible AI response modes:

    • Quick Exam Mode (concise, test-relevant answers)

    • Deep Study Mode (step-by-step conceptual breakdowns)

  • Citation-backed AI answers grounded in lecture slides, notes, and textbooks

  • AI-powered study planner driven by syllabi, deadlines, and exam dates