A Collage AI Case Study

Improving Student Outcomes

Improving Student Outcomes

Improving Student Outcomes

in Introductory Physics

in Introductory Physics

in Introductory Physics

This Spring, Collage AI was implemented in an undergraduate physics course at Harvard, one that is part of a year-long sequence designed to introduce undergraduates to physics, with a primary focus on electricity, magnetism, circuits, waves, and optics. We collaborated with the instructor of this course and measured their students’ learning outcomes after using Collage.

Harvard University

~80 students

The Story

Context

This particular physics course is designed as a flipped, project-based, and team-based course. In a flipped model, students engage with new material outside of class, freeing class time for active problem-solving, hands-on work, and peer collaboration.

The instructor’s role shifts from lecturing to guiding, but that only works if students arrive prepared.

Instructional Challenges

Previously, this course was taught through two separate platforms: a social reading platform that allowed for peers to collaborate and a web-based platform that allowed students to answer questions collaboratively and prepare for class activities.

The instructor’s biggest challenge in their previous course iterations is one that is common across higher education: how can you provide real-time, personalized feedback for every student so they know whether or not they are understanding the concepts as they work through material.

Additionally, students had to navigate between separate platforms, and the instructor had to upload everything to both and cross reference between them, adding cost in time and effort.

The result was a course where the pedagogy was strong, but the infrastructure struggled to support it. Students were not getting the feedback they needed to arrive at class ready to engage.

The Intervention

Come this Spring, Collage was the sole educational platform used.

Collage aided with creating instructional content, chunking material, providing adaptive and personalized feedback to students, and designing summative assessments that tested the course’s learning objectives.

Students were provided with real-time, personalized feedback and an instructor customizable AI tutor as they worked through material. Students had a clearer understanding of what mastering a concept looks like while meaningfully engaging rather than passively reading.

Collage is feature-rich, which allowed the instructor to build and adapt course material and meet content needs within one platform, eliminating the need to manage multiple systems. A student analytics dashboard gave the instructor visibility into how students were engaging with content and where they were struggling before class began.

After Collage

Student Achievements

62%

increase in mean normalized gain of knowledge from the Spring 2025 iteration to the Spring 2026 iteration

94%

of students found Collage helpful for their learning relative to other educational platforms on a mid-semester feedback survey

Mean normalized gain

Spring 2025

0.26

Spring 2026 — Semester with Collage AI

0.42

Comparing the cohort of approximately 80 students to the previous year’s cohort, the mean normalized gain on the pre- and post-assessment increased from 0.26 to 0.42, a 62% improvement. The difference is statistically significant (p = 0.004). Because the course structure, instructor, and assessment instrument remained consistent across both years, the introduction of Collage is the most plausible explanation for the gain.

The S25 mean normalized gain of .26 is close to commonly reported comparison averages for introductory electricity and magnetism courses. The normalized gain of 0.42 is substantially stronger relative to published research, exceeding several reported active engagement results.

The Benefits

01

Streamlining

Students and the instructor no longer had to navigate between multiple platforms at once, saving time and energy.

02

Immediate Feedback

Students received immediate feedback, resolving an instructor pain point and meaningfully contributing to academic outcomes.

03

Time Saved

The immediate feedback aided students’ concept mastery and freed up time for the instructor to re-delegate their time.

04

More Focused & Modular

Asynchronous learning became more focused and organized, as opposed to previous formats which included lengthy readings that required key information be extracted

05

Preservation of Rigor & Format

The class continued the rigor of its format without having any of the confusion that came with 2 platforms.

06

Course Expansion

Collage’s features aided the instructor in reworking the course format to serve their students in better ways.

Takeaway and Next Steps

Implementation of Collage in an introductory physics course improved student academic outcomes and allowed the instructor to reinvest their time into project based learning and student support. While this case study was specific to physics, we are now working to implement Collage and measure outcomes across multiple disciplines and class structures.

See how Collage AI works in practice.

Explore the platform or join a short walkthrough to see how it supports teaching and learning outcomes.

Contact

A Public Benefit Corporation dedicated to simplifying teaching and personalizing education at scale with AI

A Public Benefit Corporation dedicated to simplifying teaching and personalizing education at scale with AI

© 2026 Collage AI, Inc.

© 2026 Collage AI, Inc.

20 Holyoke St, Cambridge, MA 02138

20 Holyoke St, Cambridge, MA 02138