Michael Rowe

Trying to get better at getting better

A few weeks ago a colleague asked me to do a guest lecture in one of her modules. The brief was to explore a range of assistive technologies that could be used in social work practice, where ‘assistive technologies’ is conceived of more broadly than the typical ‘crutches’ and ‘falls detectors’. We wanted to introduce students to the positive aspects of digital tools, but also encourage them to reflect on the ethics concerns these technologies introduce into practice.

As it happens, this is very much the focus of a module that I teach in another programme, which meant that I’m familiar with the content, and have an outline for my own module. And while we can’t compress a whole module into a single lecture, we can expand a whole module outline into a single lecture. And because I’m comfortable talking about the topic, I really just needed a framework to guide the session.

I decided to use a combination of Claude (an open source language model), and Gamma (a generative AI-supported presentation platform) to create the presentation, using my module outline as a starting point. I started by giving Claude the prompt below to expand on the initial outline (it’s a long prompt, but I’m asking it for something quite specific):

You are a social work lecturer presenting a lecture to social work students who have a range of backgrounds and experiences. However, none of them is familiar with the wide range of technologies that are available to support social work practice.

The aim of the lecture is to stimulate discussion around the role of technology in contemporary social work practice. However, in addition to the benefits of the technology, we would also like to explore the ethical tensions that are introduced with almost all technologies. So, for each technology we present, you should mention a positive impact of the tool or concept in social work practice, as well as a potential ethics concern.

I would like you to create a presentation on the topic of “Assistive technologies in social work practice”. I will provide you with an outline of the session, which is split into 6 themes. Each them includes several topics, where each topic is a technology or concept. I would like for each technology / concept to be on a separate slide. I want each slide to follow this general structure:

  • Start with a brief explanation of what the technology or concept is.
  • Description of how the technology or concept could be useful in social work practice.
  • An example of an ethics concern that may arise when the technology is introduced into practice.

I would like a title slide, called “Assistive technologies in social work practice”, and an introduction slide that briefly explains that there is a wide range of technologies available to support social work practice, but that introducing these technologies can sometimes also introduce ethics concerns. I would also like a slide that summarises the main concepts developed during the lecture.

Here is the outline that I would like the slides to follow:

  1. Online resources and ‘the evidence’: When algorithms decide what ‘counts’ as evidence.
  2. Personal information management: What to do with information once you have it.
  3. High impact careers: Using information to identify important problems in practice.
  4. Critical inquiry: Finding the right questions.
  5. Social media: Patient-centred learning communities.
  6. Quantified self: Apps, wearables, and ingestibles.
  7. Citizen science: The rise of the non-professional expert.
  8. Robots: Non-human team members.
  9. Digital first agenda: The unintended consequences of good intentions.
  10. Nudges: Personal autonomy and ‘guided’ behaviour.
  11. Genomics: Control and predestination._
  12. Machine learning: Discrimination by design.
  13. High-stakes decisions: Keeping humans in the loop.
  14. Metadata: Accessibility and universal design.
  15. Generative AI: Deepfakes and trust.
  16. Electronic health records: When data is silo’d.
  17. Virtual reality: When virtual reality feels real.
  18. Chatbots: Talking to algorithms.
  19. Telerehealth: Remote ‘care’,
  20. Digital competence: Participating in the digital world.
  21. Advocacy: Giving a voice to the voiceless.
  22. Innovation in health and social care.
  23. Solitude: Focusing in a distracted world.

In the image below, you can see how the initial prompt (items 16-19 in the list above) was expanded by Claude into the text on the slide.

Then I took the output from Claude and added it as a text input to Gamma, which then created a set of slides based on a template. And the first version was pretty good. Gamma made only slight modifications to the text I got from Claude, and I assume this was based mainly on getting the text to fit onto the slide. I made some adjustments to the prompt (minor changes that were only around the instructions, not the content) and had Gamma generate a second version. Then I went through both versions and merged the best of each, into a single presentation.

I also started going through the presentation, replacing the stock accent images with AI generated version, that were based on summaries of the slide contents. All of this was done automatically (see below in the panel on the right).

There was only one instance where the image generated wasn’t suitable, and the reason it wasn’t suitable is that it didn’t match the style of previous images. You can see how I replaced the image below, on the left, with the one on the right, by adding a small addition to the prompt, asking that it create a new image “in the style of line art, using flat design”.

You can see, in the image below, that the detail in the picture isn’t actually very good (there is no text to speak of, only a few scribbles that look like they might be text), but the overall effect is that it creates a nice-looking slide.

This seems like a good example of using generative AI to prepare a piece of work for presentation in class, in a short period of time. It took me about an hour to do all the work for this presentation. But it took me about 5 years to get to the point where I can walk into a classroom and talk about the ideas in the presentation.

I also think that this is one use-case where the concerns around having AI generate a complete set of slides is offset by the following considerations:

  • I’ve done the work necessary to understand the topic deeply, having created the module that this lecture was based on. I wouldn’t feel comfortable using this approach to generate the content for the module I teach. It’s only by doing the work in the original set of 20-odd lectures that I was OK compressing it into this format.
  • I had an existing module outline that maps fairly closely onto the required lesson outcomes in this situation.
  • This means I could read through the generated text quite quickly, editing and adapting it to suit the aim of the session.
    • In this situation, it made one error, where I could retrospectively see how it misunderstood the language in the original prompt.
  • The images being generated are decorative, not instructional, so fidelity is less of an issue.
  • There was a relatively short timeframe in which to create something that looked decent.
  • I’ve never used Gamma before, so I knew that I had enough free credits to create something useful. If those free credits had run out while building the presentation, I may have found myself in a situation where I couldn’t finish it.

You can download the full presentation here.


Share this


Discover more from Michael Rowe

Subscribe to get the latest posts to your email.