Musings

Goodbye Canvas

For decades, certainly longer than I've been in education, there have been debates about who owns the work students create. There have been legal fights over universities keeping IP rights to work created by its employees:

Under the Copyright Act’s “work-made-for-hire” doctrine, a work is the property of an employer (i.e., the university) under two circumstances: (1) when the work is specially ordered or commissioned for use, including contributions to collective works, if the parties agree in writing that the work will be considered a work-made-for-hire; or (2) when the work is prepared by an employee within the scope of his or her employment. (source)

But what about students? When they create something in my class, does their work belong to them or to the school? Research universities will often have graduate students sign contracts to cover patentable work created by the students while at the university, but what about undergrad and high school students? There doesn't seem to be a lot of focus on IP control there, and I think that's been a disservice to the students.

Turnitin, a popular plagiarism checker used by schools around the world keeps student work in their database for an extended period of time. They collect a fee from schools and use student work to pad their database at the same time.

Online tool use exploded during the pandemic, but a 2022 paper found that upwards of 96% of those tools shared student data with third parties, including tools used by minors.

And now we have AI (LLMs) worming their way into schools. It's no secret that AI companies have been using stolen IP to train their models for years. There have been some legal fights against them, but most of the decision makers seem uninterested in providing a solution (we'll see how the courts rule).

The newest and most egregious AI-ification is coming from a new Instructure and OpenAI partnership. Instructure owns Canvas, one of the most pervasive and profitable LMSs around.

Learning management system provider Instructure logged $128.8 million in revenue in the first quarter of 2023, up 13.6% from the year before, the company announced this week. CEO Steve Daly estimated that the company’s LMS, Canvas, is now used by around 36% of North American higher education institutions. “We win a very high percentage of new deals as customers recognize the value we offer,” he told analysts on a call late Monday. (source)

The press release is full of jargon ("...establishes Instructure's future-ready, open ecosystem with agentic support as the AI landscape continues to evolve"), but at the end of the day it seems that Canvas will take student work and use it to train OpenAI's models. The multi-page release has no concrete examples of how this could be used in the classroom or even why it would be beneficial. Nor does it address IP rights or privacy concerns, especially with regards to minor users. Like any bad fortune teller, the promises are intentionally vague and the hype is high.

Ultimately I don't know what the solution will be. I know that because of this partnership I will not use Canvas in my room next year. That won't cause my state to drop their license with Instructure, but I can at least protect the work my students produce.

It's time to update my syllabus.