
Today I attended one of our own AI and Assessment Workshops to see what advice and guidance we are giving to academics and what their feelings and needs are around this topic. This is a new run of sessions which we have just started, and has been organised by one of our academics working on the topic alongside a member of my team.
Despite having published staff and student guidance documents and a dedicated SharePoint space to collate resources and our response, I found from conversing with staff at this event that there is still a prevailing feeling of lacking steer and direction. People were telling me they don’t know what tools it’s safe to use, or what students should be told to avoid. We also had a lot of people from the Library Service today, which is perhaps also indicative of the need for firmer student guidance.
I was pleased to note that there is some good practice filtering through too, such as using a quiz based declaration of use which students have to complete before unlocking their assignment submission link. We talked about adding this to our Canvas module template for next academic year, that’s something one of the academics suggested to us. On the other hand, I found people were still talking in terms of ChatGPT ‘knowing’ things, which is troubling because of the implication that these systems are more than they actually are.
While much of the session took the form of a guided dialogue, my colleague was also providing a hand’s on demo of various systems, including Perplexity which people liked for providing links out to the sources it had used (sometimes, not always), the ability to restrict answers to data from specific sources, such as ‘academic’, but noted a very US bias in the results, a consequence of the training data which has gone into these models. I was quite impressed when I tried to ‘break’ the model with leading prompts and it didn’t indulge me.
A new tool to me was Visual Electric, an image generation site aimed at producing high quality photo-like images. I have thoughts on some of their marketing… But I’m going to try and be more positive when writing about this topic, as I find it very easy to go into a rant! So instead of doing that, I have added a short disclaimer to the bottom of this post, which I’m also going to add to future posts which I write about AI.