Can I Resuscitate the University Term Paper?

The experiment

Like most university instructors, I have been at a loss over how to keep my writing assignments in the hands of human writers. This has pushed me away from the traditional term paper as a course assessment, which is unfortunate: research-driven writing is one of the core skills we are supposed to teach.

This page describes an experiment in developing a term paper assessment that can reward responsible work while lowering opportunities for abuse.

I call the approach consistency-weighted writing. The main idea is that students are invited to write co-piloted, AI-assisted term papers — but they're only fully rewarded if they can maintain the style and substance of those papers in a live setting.

There are two parts to a consistency-weighted assignment. Part A is the term paper itself, written with AI as a co-pilot. Students declare all AI use and include all prompts. Part B is an in-class essay that overlaps with, and extends, the content of Part A. Each part is scored separately, but the final grade is a weighted average based on consistency. If the student maintains more of their voice and content across the two formats, 80% of the final grade comes from the higher-scoring part. If they don't, 80% comes from the lower-scoring one. Students are ultimately encouraged not to "out kick their coverage" — that is, to write in Part A only at a level they can maintain on their own. I think this might strike a balance between discouraging AI abuse and promoting effective AI-assisted writing.

I am assigning consistency-weighted writing in Fall 2026 in two classes at the University of Hong Kong: one a mixed graduate/undergraduate research seminar, the other the tutorial sessions of a large lecture. In November 2026 I reported on early results of this experiment for TALIC — read more here.