Dr Lucy Rycroft-Smith

I see, thanks.

Noted, thank you. What would you like to see more of?

An image of the front page of the article with a weird distorted phot of some flowers.

I have a new article out in the Chartered College of Teaching journal Impact, called ‘Embedding AI in education: A critique’. Here I make five main points, which I will helpfully plop in this thread for those without access or time to read it.

A photo of an outside tap next to a salvia bush. If you look carefully there are two or weavers in their webs.

My partner just using the kitchen tap to water the plants because they didn’t want to disturb these beautiful orbweavers in front of the outdoor tap. Oh my heart ♥️

Ooh, tell me more!

Embedding AI in education: A critique : My College LUCY RYCROFT-SMITH, FACULTY OF EDUCATION, UNIVERSITY OF CAMBRIDGE, UK It has been difficult to escape conversations about AI (artificial intelligence) in education in the last year. We have had an exp...

The article is here if you would like to/are able to read it: my.chartered.college/research-hub...

Huge thanks to @danmcquillan.bsky.social @darrenmacey.bsky.social @alexhanna.bsky.social @emilymbender.bsky.social and @katecrawford.bsky.social who have really helped shaped my thinking on this.

Without the ability to discern – to sift info and reject or choose it based on background knowledge and asking people actually doing the job of teaching – AI models are giving us a reductionist, datafied & careless view of the world, the very opposite of what education is supposed to do.

Finally, 5. AI risks dangerous reductionism. McQuillan (2022, p.29): ‘Far too often, the ‘ground truth’ that machine learning practitioners cite as validation for their models is determined without going anywhere near the ground and without asking anyone at ground level what their truth might be.’

E: It is important to keep up with new technologies in general; we don’t want to ‘miss the boat’. ❌ very poor argument where children and/or large scale effects are involved, and education is both. Instead we need better regulation and much greater time-to-use with proper research.

D: ‘AI tools will be able to do administrative, boring, repetitive work or automate decision-making, freeing teachers for more creative or ‘human’ work’ ❌ This claim is not well supported by the evidence.

C: ‘ AI tools mean that teachers will be able to offer learners educational experiences that are more personalised, flexible, inclusive and engaging’ 🕳️ The evidence here is mixed & opaque, with large claims without substantial evidence, a low standard that we would not and do not accept elsewhere.

B: ‘It produces better-quality resources, explanations, tasks and assessments than we currently have’ ❌ Not currently supported by the evidence.

4. I pose the question- is it still worth it? and investigate the evidence behind the main claims as to why it might be.
A: It saves teachers time by doing tasks for them. ❌ Not currently supported by the evidence.

(The article itself is written in much more academic language than this, I assure you.) 😁

‘This means that it is most likely that AI is a bubble that can and will burst, because AI systems are extractive.’ This is not labour-saving. It is labour-transferring - to places and people who are already marginalised and vulnerable. It is exploitation - a plain ol’ pointy pyramid scheme.

3. Both the technology and philosophy of AI therefore have hidden costs that have been ‘kept a secret’, such as MASSIVE water and energy consumption, and embedding ideas about ‘intelligence’ that depend entirely on eugenics. Both morally flawed and simply not good science.

I argue that this is something of an issue (and how are we not having this argument loudly already) in education, where we claim to value truth’, as well as nuances of language, depth of understanding, and understanding bias - all things LLMs just don’t. This clash of values seems problematic.

2. Large language models are not (and never will be) accurate. ‘The so-called ‘hallucinations’ (i.e. making statements that are lies) that happen are really the model working as intended, which is to produce something that looks and sounds plausible, rather than pursue ‘truth’ (McQuillan, 2022).’

This means, of course, that ‘right’ is not (and never) neutral. AI tech embeds a kind of utilitarianism (maximising something ‘good’) which involves valuing some things over others, and often those things are invisible to the user. Just ‘plop!’ A lovely decision, backed by maths.

1. AI means lots of different things to different people, but it is not just a set of technologies- it’s a philosophy (way of seeing the world). ‘AI often involves efforts to make computers ‘do the right thing’ in a given scenario. ‘

An image of the front page of the article with a weird distorted phot of some flowers.

I have a new article out in the Chartered College of Teaching journal Impact, called ‘Embedding AI in education: A critique’. Here I make five main points, which I will helpfully plop in this thread for those without access or time to read it.

A bowl of hand-picked tomatoes, one of which has a tumescent little growth

Well this little guy seemed *pleased* at the prospect of being made into soup…

Ooh @darrenmacey.bsky.social one for you!

Mais oui 😁

‘Honeypisquared and I’ 😬

We are alternating so we can cover both!

#Teachers, we would love to hear from you. How are we doing? What could we do better? Which books would you like us to review next? What accessibility features can we do better at?

For the best experience accessing Amazon Music, we recommend using the latest version of Firefox, Chrome, Safari or Edge.

And on (booo) Amazon Music: music.amazon.co.uk/podcasts/55f...

Research Party Education Podcast · Updated Monthly · Lucy and Darren review

Here we are on Apple Music: podcasts.apple.com/us/podcast/r...