Where is AI leading us and do we want to go there?

Generative AI is everywhere - you'll no doubt see it in search results, phone apps and even in some of our University databases such as Ebook Central, where the 'Research Assistant' will offer to summarise, explain and much more.  We, of course, encourage you to use generative AI wisely and within the University guidelines of such use for academic work.  When people rely on AI to do their thinking for them, this has also been shown to stunt their ability to think and create.^1,2 

A lack of reliability and transparency

It's particularly important to note that generative AI engines still 'hallucinate' and may give incomplete, incorrect, or misguided answers.  Not that AI can actually think or hallucinate as you or I might - the output of generative AI is simply words and images compiled by brute force statistical analysis that analyses what you asked for and predicts what you will find to be an acceptable output, regardless of whether it is right, true, fair, representative, or even based in reality.  The outputs are often useful and, within the limits of the training the AI received, may be accurate, but there is increasing concern from those involved in the AI industry and elsewhere that we no longer understand how AI is reaching its conclusions: the more complex the decisions entrusted to AI, the less clarity we have over precisely how it reached those judgements.  Even if we asked AI to explain its 'reasoning' it could not explain precisely how it arrived at the conclusions it has because the statistical models that power AI are so very different to human thought as to be unintelligable to us.³  

Critically appraising AI

It's also worth considering where the AI responses are coming from:

• Who shapes the algorithms and output? 
• Who benefits from generative AI? 
• Whose values and histories and voices were included in the materials on which AI was trained, which comprise its entire 'worldview'?
• Whose values, histories and voices were systemically excluded from the training data? 
• Does AI give pre-existing social biases in its training data a false veneer of objective scientific rationalism?
• What happens to a society when the power to make significant decisions is handed over to an AI whose reasoning is fundamentally unknowable and whose training materials are known to be systematically biased?
• Are the current models biased or leading us to places which no one has asked for?

Fuel for the mind

To help you explore these kinds of questions, we've created a new reading list on Decolonising Generative AI.  This small collection of books, articles, and web pages highlights critical and global perspectives on AI.  We hope this offers you a richer and more inclusive take on how these technologies are built and used.  We certainly hope that it helps show how generative AI is neither neutral nor created in a political vacuum: it is created by people, shaped by power, and the current direction of travel simultaneously threatens to intensify and normalise the inequalities and discrimination found in the societies from which its training data is drawn.

Upwards not backwards

Hopefully, seeing the limitations and dangers inherent in the indiscriminate use of AI will inform how we understand and use it, and how it subsequently impacts people differently around the world.  This does not mean rejecting new technology but to challenge the unthinking acceptance of all that is new as good and find ways to make the AI we are developing inclusive, transparent, and accountable.

If you have any suggestions for additional books, articles or web pages, please let us know.

References

1. Georgescu, R. I., Bodislav, D. A., & Andrei, I. V. (2025). The acceleration of education: professors, AI platforms and the transformation of learning and human growth. Theoretical & Applied Economics, 32(4), 143–154. https://research.ebsco.com/linkprocessor/plink?id=ffb6f8a8-49ad-3db6-a473-a3b088b5e59d

2. Zhai, C., Wibowo, S., & Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on students’ cognitive abilities: A systematic review. Smart Learning Environments, 11(1), 1–37. https://doi.org/10.1186/s40561-024-00316-7

3. McQuillan, D. (2022). Resisting AI : An anti-fascist approach to artificial intelligence. Bristol University Press. https://ebookcentral.proquest.com/lib/portsmouth-ebooks/detail.action?docID=7042128.