Last week, Vanderbilt University decided to disable Turnitin’s AI detection tool. Congratulations are in order!

To date, there’s little evidence as to the effectiveness and appropriateness of such tools (also see: their unintended consequences). Equally importantly, Vanderbilt’s decision lends credence and support to recommendations that numerous working groups put forward to their institutions, and paves the way for others to feel confident in taking similar actions. Earlier this year for example, I led a generative AI taskforce at Royal Roads University. The relevant recommendation we put forward in early June is this:

Recommendation 5: Investigate, and potentially revise, assessment practices.

We recommend that faculty examine their current assessment practices and question them through the lens of AI tools. For instance, faculty could try their discussion prompts or reflection questions with AI tools to explore the role and limits of this technology. Regardless of the outcome of such efforts we recommend that faculty do not rely on AI detection tools to determine whether learners used AI in their work. A service that asserts to detect AI generated text does not provide transparency on how that assertion is made and encourages a culture of suspicion and mistrust. Emerging research also highlights challenges with reliably detecting AI-generated text (Sadasivan et al., 2023). Instead, we recommend that faculty engage with learners in conversations at the beginning of the course as to the appropriate and ethical use of AI. We further encourage faculty to continue their efforts towards experiential and authentic learning– including work integrated learning, live cases, active learning opportunities, field trips, service learning, iterative writing assignments, project-based learning, and others. These are not necessarily failsafe approaches to deter cheating, and it may even be possible to leverage AI in support of experiential learning. Ultimately, we recommend that faculty question their assignments at a time and age when generative AI is widely available.