The Teaching Hub Virtual Assistant is available now.

See below for how to use it, an FAQ and a feedback form.

How to use the Virtual Assistant

You can see the virtual assistant at the bottom left of the window. Click on it and ask a question or request guidance. The assistant will provide a concise answer and suggest links to relevant articles and pages on the hub for more information. Examples it could help you with include:

  • How do I record in Panopto?
  • How do I create an assignment in Moodle?
  • Are my students allowed to use generative AI?
  • What is the Bath Blend Baseline?
  • What does fair allocation in Moodle mean?
  • What can I do to make courses more accessible?
  • I need guidance on how to provide high quality feedback
  • What tools are available to help improve accessibility for my students?
  • What is the CASE model?

Virtual Assistant FAQ

Additional information is added as the assistant is developed.

The assistant is designed to provide clear, concise answers to questions and direct you to the relevant guidance articles and pages on the Teaching Hub.

The assistant is specifically trained on Teaching Hub content. It cannot assist with ‘general’ questions about the University of Bath – Please visit the main University of Bath website for general enquiries.

The Teaching Hub comprises over 260 guidance articles and 60 pages of information at the start of September 2024. It has been trained on all this information. As our content is regularly updated, it is ‘re-trained’ on a weekly basis to ensure new and updated pages and articles are included.

This issue is less prevalent than it used to be, but it is possible the assistant will ‘invent’ information. This is most common if you persist with a specific line of questioning. As the assistant is prompted to suggest links to relevant hub pages, it may ‘make them up’ if specific requests for links are made. Every conversation is logged to help us identify when this occurs and refine the responses, but it is not possible to completely eradicate this issue at this time.

All interactions with the assistant ARE STORED in full. This helps us identify when the assistant is providing poor responses or a user expresses concern about a response. The IP address and full conversations are recorded for all non logged-in users The IP address, logged in username and full conversations are recorded for all logged-in users. By interacting with the assistant you accept this information is stored.

The assistant has ‘fair use’ limits in place. If you are using our site as a guest (I.e. You are not logged into the Hub using your Bath credentials) a strict fair use limit is imposed to avoid abuse of the system. When you are logged into your University of Bath account, the assistant will allow a significant number of interactions every 24 hours. These limits may change at any time without notice.

Please use the assistant feedback form – It’s on the page you’re reading now!

The OpenAI model ‘GPT 4o‘ is used to build the response shown to the user.

Every article and page is converted into a vector embedding, which is stored in a vector database

When a question is asked, the ‘most relevant’ embeddings are selected and used to form the response.

It is re-trained once a week on the information available on the Teaching Hub. As our guidance pages are updated on a rolling basis throughout the year, there can be a delay between systems and processes being updated and the assistant being ‘aware’ of this.

Events information is not within scope at this time. We will look into adding this information at a later date.

We may update the models and prompts used by the assistant over time, based on feedback and analysis of previous responses. This is expected behaviour during the trial period.

Feedback form

This form is for feedback about the virtual assistant only. Include your contact information if you would like us to contact you. Feedback will be reviewed on a weekly basis during the trial period. Every conversation with the assistant is stored so we will already be aware of ‘obvious’ mistakes.