Beyond AI Scales: Fine-Tuning Assessments for GenAI Use

Since Leon Furze originally introduced (as far as I know) the idea, many different “AI scales” have been proposed.

While they are certainly a step in the right direction and provided a straightforward, early solution to issues surrounding AI misuse, their limitations call (in my opinion) for a more refined approach.

I outlined these limitations last year in this post.

To address them (including the insufficiencies of the version I had developed myself), the most promising approach, it seems, is to fine-tune assessments for GenAI use.

Here, rather than setting an overall allowed level of AI use (which is highly problematic, as this post explained), teachers:

-Break down any assessment into the targeted learning objectives (rather than limiting themselves to the traditional steps of the writing process)

-Clarify the appropriate level of AI use for each of these dimensions separately

-Provide practical guidance, including a recommended toolbox and good AI literacy practices

This shift is very similar to the transition from traditional letter grading to standards-based teaching and learning. And it has similar benefits for both teachers and students.

I laid out the general idea last year here.

And further explained how this approach helps integrate AI automation and the progressive development of autonomous skills here.

In case anyone is interested as the new school year begins ;)