AI's source value 

AI as a Source

AI’s source value – the validity of using an item generated by AI as a scholarly source – is currently quite low. The reason for this is because it tends to “hallucinate”, or – in plainer language –make things up. 

It is unlikely that the AI is going to come up with something as a source that is of any greater use to a student than the corresponding Wikipedia, Stack Overflow, or generally ‘googleable’ source. This will change as the technology improves, but as of March 2023, the source value for facts and many other calculations produced by generative AI is quite low. We do not recommend students referencing AI as a source for facts.

The use of AI to find source documents is patchy at best and highly dependent on the model. Some models (e.g., Bing incorporating ChatGPT) have access to the internet in the same way that a more familiar search engine will. Some models only have access to data they trained on. Some models will just make things up! 


Referencing

The use of generative AI in academic contexts is very new. There is not a clear, settled way to reference generative AI yet in the spheres of higher education. The Office of the BSSS will be consulting with disciplinary experts about referencing to help inform our advice to schools later in the year as the initial waves made by the technology smooth out into calmer waters.

For now, different universities, disciplines and referencing styles offer different recommendations. 

The recommendations from most citation systems and universities are currently to cite generative AI as a non-recoverable source. This is because AI generations generally cannot be precisely duplicated by someone else using the software, making the source inaccessible to a reader.

Reflect 

In your workbook: 

Reflect on the use of AI as a source in your discipline and context. This might include ideas such as: