Searching for the AI bridge builders

We need to democratise access to AI but the language we use to talk about it is a barrier.

There is a lot of fear around AI advances and this is perpetuated when only the ‘big tech’ and the academics have access to the tools, the theory and the conversations. For me this is a major theme in AI ethics right now. We can’t have conversations about ‘the black box around generative AI’ as if everyone understands the concept. Similarly ‘language models’, ‘the dynamics of knowledge production’, or ‘neural networks’. I suspect that I have already lost a large chunk of my friends and family and we are only on the first paragraph.

We talk a lot about bias in the data; there’s a great advert doing the rounds on social media at the moment where an AI was prompted to draw Barbie dolls from around the world. Some of the results are quite a shocking reflection on our own stereotypes and cultural tropes with German Barbie depicted in a Nazi uniform and African Barbie carrying a gun. AI may have created the images but we have supplied the data. It is an accessible depiction of bias, we need more accessible depictions of AI concepts.

As academics, researchers and professionals, what we don't see so easily is the bias innate in our own use of language around AI. It is the same in all industries, in all academic circles across all disciplines, we are so used to discussing with each other that we become stuck in our bubble of understanding, of acronyms and concepts. What we need is a giant pin, and we need more AI Pioneers to bridge the gap between theory and practice. More people willing to stop and ask questions. More translators of AI speak. More people who are comfortable in both worlds, who do not feel alienated by the academic circles and equally do not alienate practitioners, who, lets face it, are the real experts here. It is the practitioners who will be finding innovative ways to teach with and about the tools, and as with all previous ed-tech advances, it is the practitioners who will work out how to ‘hack’ the systems to fit their contexts. It is also the trainers who will be on the ground working with learners with poor digital literacy, trying to engage and enthuse them to not be automated out of a job.

I’d like to think that my work and that the projects Pontydysgu are involved with fit the gap nicely, providing introductory materials and creative ways to use AI tools, but I was reminded by a group of trainers I ran a workshop with recently of the need to slow down, take things back to basics. 

When I first started out in edtech I was the trainer-in-training, in one session billed as a ‘hands-on practical introduction to e-learning’ the instructor showed us how learners’ work could be exhibited on a website - it was new and exciting, the dawn of web2.0, everyone in the room was eager to learn how. But we were then left with the bamboozling task of “now build a website.”

In my workshop, I heard the words “now use that to build a bot” escape my mouth and realised that the student had truly become the master. 

We need to remember to put the scaffolding into place so as not to lose people over the edge, and that includes explaining ourselves clearly or at least signposting people who can. To quote Einstein, “If you can't explain it simply, you don't understand it well enough” If you are one of those people, a gap-bridger, a mediator, an educator and also an AI enthusiast I warmly invite you to join the AI Pioneers network. Use the contact form on our website to get in touch, join in the conversation on Mastodon (like Twitter but without the megalomania) or find us via LinkedIn.


AI and the digital divide

How do we make AI accessible? If you're already excluded from the system how can you possibly catch up when technology is changing exponentially?
I will be moderating a breakout session on making AI familiar and fun to foster learner motivation at this online workshop for trainers - Creating An AI Literacy Training Model.
These free sessions are run by Digital Collective, a non-profit organisation that aims to tackle the increasing digital divide, that is not only creating an alarming distance to the labour market for individuals in vulnerable situations, but also causing social exclusion as society moves into the digital age.
10:30CET Thursday 26th October 2023

Pioneers in Practice – Word Clouds

In this series, I'm re-visiting our best loved teaching and training resources from the past 15 years of projects and updating them to reflect the changing world of ed-tech. Here are some ideas for using Word Clouds, originally contributed by Nic Daniels, one of the old Taccle2 team in Wales, and updated here by me, Angela Karadog to incorporate AI tools. Whilst this particular piece has been an interesting research journey, it's proof that adding AI doesn't necessarily improve an already useful tool.

We love word cloud software! It’s so simple to use and the possibilities are endless, for any topic at any level. We’ve outlined how to use it as a fun and quick lesson warm-up activity. It's also a great tool for learners to use too.

Using either type or cut and paste the focus text in the large white box. This traditionally includes topic vocabulary, spelling lists, poems or text extracts from books, to name but a few sources.

By now you’re probably used to making word clouds for spelling and vocabulary games e.g. put adjectives you’d like to revise in the box and press go to create a word cloud using the words you’ve provided. There's now also a cool updated version which allows you to fit the words into a shape. like I've done with this article in the illustration above.

So what about AI? Check out this Word Cloud generator from speakAI which uses natural language processing so you can also add a transcript, audio or video data as well as unstructured text, poetry or prose and the AI will then ‘Analyze’ the text to produce a visually appealing cloud with interactive words and, in the paid version, multiple ways to visualise your data. This tool from Shulex has done away with the need for a source text and will have ChatGPT generate the words for your cloud from a single key word - sort of an AI thesaurus in pastel. 


Teaching WITH AI

Here’s Nic's suggested activity with some minor tweaks.

Display the Wordle on the interactive whiteboard for a set time (30 seconds is usually enough) and the learners must make a record of as many words as they can in the allotted time.

If using edWordle Be sure to select easy to read orientations and fonts. 

Lastly, if you have pupils who prefer coloured screens as opposed to the harsh white, you can change the background in the ‘bg color’ box by clicking the FFFFFF

What do I need?

For a whole class activity, an interactive whiteboard or projector is pretty essential. You can however create Wordles for pupils to use individually or in pairs on a device or printed.

Timing device – I used an egg timer! But a clock, watch or stop-clock would do the same job.

This is one online resource that takes less time to create than doing similar activities non-electronically. Not only is it quick, but it’s endlessly adaptive! To create something similar on a poster or a conventional writing whiteboard/blackboard would take at least an hour. This is done and ready to use in under 5 minutes!

Beforehand, you may like to create a Word Cloud and ask learners which colour scheme, font and layout makes it easier for them to recognise the words. This is especially important if you have learners who read books using a coloured overlay.

The more words you use, the more complex the Word Cloud, so for younger or less able learners you may choose only 10 words. You can use the same word as many times as you want, this can also simplify the activity.

For us, the biggest attractions of using a Word Cloud is that:

  • it’s fun!
  • it really kick starts the lesson, ensuring learner engagement for the beginning.
  • you can save your Word Cloud and use it again and again.
  • it’s endlessly adaptive

Teaching ABOUT AI

The AI backed Word Cloud software picks out words and sentiments in the text that it has learned are useful to us whereas the non-AI software is only looking for word count and returning the most popular words. The non-AI clouds will automatically contain numbers and mis-spellings unless you specifically tell it to exclude them.

With that in mind, lets compare and contrast the same text fed into both types of cloud generator.

Use the activity as a basis for discussion;

  1. What rules might each word cloud tool be following?
  2. As humans, what are we hoping to see when we create a word cloud?
  3. Does AI have a role to play in analysing texts?

As an example I pasted the exact same text, the first 20 pieces of 'Advice to a Wife' from project gutenberg into some word cloud generators. The results did surprise me, I was genuinely expecting the AI to do something clever. Personally I prefer the non-AI result, but if I were analysing my web content to improve the SEO or adapting my campaign to give it a more positive spin, the sentiment analysis tools would better come in to play.


Pioneers in Practice – teaching with and about AI

water colour painting of a middle aged white female teacher with dark-pink curly shoulder-length hair wearing a green v-neck dress and red rimmed glasses working at a mac computer

Over the past 15 years or so, Pontydysgu has created hundreds of free digital Open Educational Resources for teachers, trainers and educators to use, re-use and adapt to their own needs. Given the recent advances in AI tools, generative AI, natural language processing etc, I thought it would be pertinent to revisit our old, well loved resources and give them an AI inclusive update, I'll likely be including a few new ones too. Over the coming weeks expect a return to the chalk-face with scenarios, practical ideas, hints and tips in the Ange's Scribbles corner of the Pontydysgu blog. I'll also be exploring new pedagogies and old learning theories with AI in mind.

As always if you have a great idea for a contribution, text, video or podcast, get in touch.

Featured image generated by

Prompt: water colour painting of a middle aged white female teacher with dark-pink curly shoulder-length hair wearing a green v-neck dress and red rimmed glasses working at a mac computer