Tweaking my ‘Policy on Use of AI and Other Creative Tools’ (version 2.0)

In Spring and Fall 2023, my students had some successes in using generative AI such as ChatGPT for coursework (without crossing the line into a stated integrity violation). Some of these uses were:

  • Creating a prototype image of a “Quantified Toilet” in situ, as part of a privacy design project (based on a thought experiment at CHI 2014, but a real-life possibility too).
  • Writing quick research summaries for a shared Google Slides deck in my Collaborative and Social Computing graduate seminar.
  • Testing out whether commercially available Large Language Models (LLMs) can reliably and validly answer questions about dealing with security and privacy concerns.

But they also ran into a few obstacles. One, they did not know how to effectively prompt these models or to generate versions or new iterations of the first idea. This semester, I will provide them with more guidance.

Two, students did not know enough to be on alert for errors generated by the models. With images, that can be as simple as a missing flush valve on the toilet tank drawing. With text, the errors can be harder to notice if you are not knowledgeable about the topic. I had to correct students on a few occasions that the papers or author names that ChatGPT generated, based on an existing research paper, simply did not exist!

Georgia Tech’s Amy Bruckman has provided her draft of an AI policy that puts students on notice of potential harms. She notes that her courses are writing-heavy, so she discourages the use of genAI, noting that it has the strong potential to reduce their learning. Other potential harms noted in her policy: factual errors, bias, fake references, and poor style.

With this in mind, I have revised my Version 1.0 policy to use the following text (italics show emphasis in the syllabus given to students):

In this course, students are permitted to use tools such as Stable Diffusion, DALL-E, ChatGPT, Bard Gemini, and Bing Copilot. In general, permitted use of such tools is consistent with permitted use of non-AI assistants such as Grammarly, templating tools such as Canva, or images or text sourced from the internet or others’ files. No student may submit an assignment or work on an exam as their own that is entirely generated by means of an AI tool. If students use an AI tool or other creative tool to generate, draft, create, or compose any portion of any assignment, they must (a) credit the tool, and (b) identify what part of the work is from the AI tool and what is from themselves. Students are responsible for identifying and removing any factual errors, biases, and/or fake references that are introduced into their work through use of the AI tool.

I give a syllabus quiz at the start of every semester. I now have written a question to reinforce to students what they should retain about this policy. In future class sessions, I aim to follow up with a discussion in class about how to identify problems in generative AI output, and how to remedy these problems.

My tips for conducting an online Zoom class amid the Covid-19 pandemic

Phew, what a semester! I ran a section of our Programming Usable Interfaces course here at Carnegie Mellon University, and I mentored several student assistants and two research associates for our HCII Social Cybersecurity research project – all while taking a required course (Social Web, roughly a survey of Computer-Supported Collaborative Work and Social Computing) and an elective (Computer Science Pedagogy). Oh and finished all of this AMID A PANDEMIC, while WORKING FROM MY CRAMPED APARTMENT with TWO INCREDIBLY FUSSY CATS.

It has been a steep learning curve to work out how best to use Zoom and other tools when carrying out university work. I found the following practices helped our sessions to work best:

  • Be very explicit in what you want students to do. I wrote out a script where I would verbally tell students to post in the Chat window, raise their hands with the icon or just unmute in order to ask questions or make a comment.
  • Use breakout rooms to facilitate discussion and social connection. No one will be able to see the discussion prompt slide once in the room, so it is best to keep it general or re-post the prompt via message once students have joined the groups.
  • Accept that you only get half the attention as in an in-person class – typical user is multitasking with in-home activities and distractions (in bed or cooking or managing kids/pets or doing laundry) – this includes me, when I’m not the lecturer, so I have empathy for this user persona!
  • No one will share screens or audio if the group is too big to fit in gallery view. This unfortunately amplifies the already present distancing of the video screen interface in any such conversation.
  • Rather than sharing links for additional material in Zoom chat, create a Slack workspace or use Piazza threads. Both are persistent and searchable, and Slack allows for lightweight engagement such as emojis. However, I also would upload to Canvas the (edited) chat transcript with my encoded mp4 file at the end of lab, for students who were not able to attend synchronously. I do not think a Canvas discussion thread is going to be useful for this, because the UI seems primarily designed for required discussion posts on assigned readings, but you could post it as an Announcement, which will be front and center for students.
  • Streaming videos can still be a fun and useful break in lecture – I had a lot of success showing a video demonstration of the Bootcamp.js grid system amid a lecture on using it for web design – but keep in mind you need enough bandwidth to stream multimedia and also need to configure your sharing settings properly: Turn off the video background or even your in person video altogether if needed, and check on Share Computer Audio when starting the screen sharing.
  • Stick with one persistent Zoom link for each type of meeting. It was a lot easier for me to simply create one repeating calendar item with a persistent Zoom URL, rather than constantly have to hunt for whatever the new Zoom link was for that day’s class or meeting. I saved my personal Zoom link for office hours and other activities that were likely to have happened in-person in my office. Use the waiting room feature or enable the password as part of the URL itself, if you are concerned about securing the meeting to only those unlikely to cause trouble. For one-off events, use a registration form to collect and vet attendees, then send a separate email with the actual Zoom link to the approved attendees.

Let me know in the comments what other practices you found helpful – or share on social media!