Using Generative AI in behavioural research

Recently Professor Susan Michie, Dr Janna Hastings and Professor Robert West hosted a BR-UK webinar discussing the pluses and minuses of using Artificial Intelligence to improve behavioural research. Read on to learn what the team did to test whether AI could help improve our efficacy.

BR-UK is running a series of three webinars on use of Artificial Intelligence (AI) in behavioural research. 

The first of these was attended by more than 300 people and focused on how any of us, with little practice or training, can use the new type of AI called ‘Generative AI’ and ‘Large Language Models’ to improve our efficiency and effectiveness as behavioural researchers. The second webinar in the series takes place on March 12th at 11.30am (register here) will be on what we are calling ‘Analytical AI’ and the third will be on May 12th at 1pm the responsible use of AI (register here). 

During the first webinar, it was suggested that using generative AI to convert the PowerPoint slides and transcript from the webinar and review the outputs would be an interesting test the ability of AI to improve our efficacy.  At BR-UK we routinely record our webinars so that people have an opportunity to catch up with them if they missed them.  We use the Zoom AI transcription to create transcripts that can be shared on our YouTube Channel afterward. 

For the first session we went a bit further and used Generative AI to create a briefing document from the transcript, and a blog from the briefing document! As a bit of fun we also used the briefing document to generate a podcast in which a male and female AI-generated commentator discuss the webinar. If you don’t have time to watch the whole webinar – and even if you do – have a look at the briefing document and blog to get the key points that were covered – and listen to the podcast to see what the current state of the art can do to try to spice up the presentation of what can be quite dry information.  The links are given below. 

  • NotebookLM is an AI tool from Google.  After uploading one or more documents, Notebook LM can summarise them and answer questions about them. So we put the transcript of the 90-minute session into NotebookLM and got it to provide a briefing (one of its inbuilt tools). Here is what it came up with. 
  • DeepSeek is an open-source AI tool that performs similarly to ChatGPT and similar Generative AI tools but is more open in terms of the way it works but has the advantage of using a ‘distiller’ which distils the model that is trained on the world’s data with hundreds of billions of parameters into a much smaller one with only about 7 billion parameters (!) that performs almost as well but uses much less by way of electric power and hence is better for the environment.  We used Deepseek to turn the briefing from NotebookLM into a 1000-word blog, which can be found here.
  • We then went back to NotebookLM and used the blog generated by Deepseek to create a podcast – what NotebookLM calls a ‘Deep Dive’ – in which a male- and female-voiced AI person discuss the webinar. When you first hear it, you will be amazed but, as with a lot of generative AI, on further exposure and reflection you will see that the algorithm behind it is rather formulaic. In this case, the male voice seems to do most of the talking and while the female voice interjects appreciatively. Here is a link to it – so see what you think. We will be returning to the issue of bias in AI in the third webinar in the series.

And finally, it’s worth saying that, while summarising and making scientific material is definitely a useful function of Generative AI, the bulk of this webinar was devoted to how we can use Generative AI in our research, and particularly in generating potentially useful components for behaviour change interventions using the scientific literature. Watch the full webinar and/or check out the summaries for more information. 

Learn more about our future webinars and how to register

This blog was written by Professor Robert West and Professor Susan Michie with input from Sancha Martin (BR-UK Hub Manager) and the outputs were generated by AI as outlined above.

This article was published on