More On Adobe Express Image Bias, Ethics, Values and AI Slop

Screen grab from Emily's Substack article

Opened up my Substack account yesterday and saw a post from Emily Cherkin, who according to her own profile "is a an author, teacher, speaker, writer, activist, and parent." I've followed Emily for a while and her posts are focused on the use, or should I say, abuse of digital technologies in the classroom, again to quote from her profile, her mission is "to save children, education & democracy by dismantling the EdTech industry through grassroots change." 

The post in question was called "P*rn, Gambling, Genitals, and AI Slop: A Day in the Life of a School-Issued Laptop" which looks at real life examples of what learners can access on school issues laptops / Chromebook. I recommend you have a read to see the very concerning examples of what young learners were able to view. These were shared with Emily and were mostly examples from US users. However, not all of those examples were US, because to my pleasant surprise, Emily included examples I had emailed to her! Why did I do this? Well, a couple of weeks ago, Emily posted a piece on Substack about a questionable image that was AI generated through the Book Creator app. I found the post interesting because back in January and February I had written two posts outlining my concerns about learners generating AI images, not with Book Creator, but with Adobe Express, which is available and being used by learners through the Hwb platform.  I was particularly concerned with the obvious north American bias in AI image output. Simple prompts like "a police officer in the street" would generate clearly north American police officers. Changing the prompt to "a UK police officer in the street" would generate the same type of image. A "football player" prompt would generate an American football type image. I also found it quite humorous but also quite concerning, that prompts such as "a secondary school headteacher" would produce images of people who looked like they had only just come out of 6th form! What came across clearly from all the examples I created was that everyone was 30 years old or under and that they looked like they were movie stars. Handsome men and beautiful women only. So, I emailed some of those examples to Emily. I also wanted to try out Adobe Express again, as it had been nine months since I generated those examples, just in case Adobe had updated the AI image generation at it was now better at producing non-Americanised images. But I needn't have worried, as I was getting very similar outputs as previous. Therefore I emailed my images through to Emily and they appeared in her article.

The Welsh Government were clear in their response to my letter to the Education Minister, that learners in Wales don't have access to generative AI chatbots, such as Google Gemini and Microsoft Co-Pilot, but they still have access to this AI image generation product through Adobe Express. They obviously think that there is no possible issue with our young learners using this product. Perhaps it's just me, but I still don't understand this. Just focusing on AI image generation alone, in my opinion there are far too many underlying ethical issues that should at least stop teachers in their tracks and think twice before introducing this product to primary school learners or at the very least, use it as a starting point to talk about the wider generative AI issues. Opportunities to discuss "AI slop" and the homogenisation of images, as an example, just look at the huge amount of adverts put out on social media by schools advertising events, all generated by AI, each looking exactly the same. No individualisation in these posts, it's pretty impossible to tell one school from another from these outputs. Many businesses are being 'called out' for using AI to generate posters and some are now explicitly stating that they will not use AI products to advertise their services. Not so with schools, at the moment. 

"People are really getting AI fatigue already and I think it comes across as being quite inauthentic. It raises questions around trust that people have in brands, how relatable a brand is, if there's nothing to differentiate it, nothing that's really unique about it, I think people are so overwhelmed with messages they'll just pass by and it even has a sort of repellent effect for an increasing number of people." Why everyone from your local restaurants to the school fair is using the same ads

There are obvious opportunities to discuss ethical questions around whether it is right that all photos, paintings, illustrations, videos created by artists have been stolen by big tech to feed into their LLMs? Learners need to understand how bias, such as the examples I've highlighted, happen. I recently saw a post on Instagram where the person had a great idea to turn the art ethical questions in value questions that might relate to individuals. Let's face facts, people only tend to get upset about something if it impacts things they feel are important in their own lives. For example, some people will argue that these products 'empower' users who can't draw, paint, create music, etc. So let's ask the children critical, ethical, philosophical, value questions such as; if you were a professional illustrator of children's books, how would you feel if your work was stolen to train this product and then produces images similar to your own? Now, some may not value the 'arts' so let's compare it to things that some might value:
Would you feel the same way if I used a taxi to complete a marathon that you were running in?
If I used a fork lift truck to lift weights in the gym?
If you competed in a music contest and I entered and pressed auto-play on my piano?
If I used a jet ski to compete in a swimming gala against you?
If I used a helicopter to get to the top of a mountain and then told people I had climbed it?

I do have concerns that even though Estyn calling for the need of "critical evaluation" and "ethical understanding", I just get a gut feeling that this will become a very narrow lens of ethics, limited to not using generative AI to 'cheat' and being honest in their use of generative AI products to help them with their work. While I'm not going to disagree with this need, it is in my opinion a very narrow focus, completely missing on the wider social and environmental impact of these products. I would also like to see some reference, that learners have right not to use generative AI. That appears to be completely absent from any AI documents or discussions relating to education in Wales.  I hope that I'm wrong and that the DCF, whenever it is finally updated, will include this wider look at learner generative AI ethics.