Delivering a way to help young people use generative AI thoughtfully (the solution)

If you know of folk who’d like to support young people learning about AI, we’d love to hear from you

Over the past 6 months, we’ve co-created a means of helping 15 – 17 year olds to understand, examine and use generative AI in positive ways.

The National Citizen Service (NCS) and Centre for the Acceleration of Social Technology (CAST) asked dxw family-member Neontribe to answer the question “How might we support young people to become thoughtful makers and critical users of this technology?”

Young people are the ones who will be living with this technology and this is about empowering them with the ability to embrace the opportunity AI offers safely and thoughtfully. Jo Hutchinson of the NCS at the time said:

This is why the co-creation model with young people at the driving seat was so important to me, the tech needs to be understood and shaped by those who will be using it.

We’ve been collaborating with youth organisations Beats Bus, The Politics Project and Warrington Youth Zone, who’ve been sharing their expertise with us while also learning about the digital design process.

We began the project in November, starting a discovery phase, which Hannah Wallwork wrote about in a previous blog. That done, we focussed in on NCS’s question, defining and iterating a response, a process we describe here, and then started development as Charlie explained here. This post explains how we delivered the solution we’d developed.

From discovery to delivery

It’s no coincidence that this series of blog posts have talked about discovery, definition, development and delivery. 

This is the Design Council’s “Double Diamond” model, and it’s how we work. Explore the problem space, focus in on a well-understood problem, explore the ways in which that problem can be solved, then ship a solution.

The latter is a technical process, supported by usability testing, where we iterate on and implement the prototype we defined. In this case, it’s where we changed interface copy, and tweaked the journey we’d designed, in response to live testing of the actual code.

It’s now live at realchatai.org

Real Chat AI surrounds an AI tool with content that puts its use in context. Some of that is flat text – things to think about. More interestingly – some is itself generated by Generative AI, posting out ways in which the specific prompt you’re using might lead to biased, or inaccurate, responses. 

It also keeps the user one step away from the actual LLM. The site fires off requests to an API so the actual commercial service doesn’t get any data about the person using it – unless they deliberately give it. Which we advise not to.

We’re not done yet

So; discovery, definition, development and delivery. As others have pointed out, its main deficiency is that it assumes once you’re delivered, you’re done. I doubt that was ever true, but it’s particularly wrong in the digital world. 

A wise person once said “Send your website live many times or never.”. It’s from actual data on use that you’ll learn what really works, and what needs improving. 

So we’re delighted to continue our relationship with youth organisation Beats Bus. They work with young people, using hip-hop to educate and engage, building skills and communities.

Their boss, Steve, has demo’d RealChatAI to a number of different people. He’s using it cautiously for now, but he’s got some great plans for rolling it out in Hull and West Hull aiming first at areas of specific deprivation. He’s had some really enthusiastic support from other organisations using music, and creative arts in general, to engage with young people. 

Alongside that, we’re exploring business models for this kind of community-owned technology. Could smaller organisations who wish to use RealChatAI in their work, each pay in a bit – to create a collective pot we can use to pay the bills? Might larger organisations pay into a fund for making changes to how the tool works? Further down the line we’ll work together to figure out which of these approaches we want to pursue. 

Essentially, Steve is doing a cracking job of product ownership. He’s getting people excited about it and lining things up. Neontribe and CAST are here to support him with contacts, connections and coding new features when he feels they’re justified. Someone else has to decide what is happening with the tool and its development. That is Steve, and it’s a pleasure working with him.

Our big question now? How long will our hosting budget last. It’s more complicated than just accounting for the server space that runs the service. We have to pay for every API hit, for every prompt we send off to the LLM we’re using. We’ve a model that gives us 3 months, and we’ll see how accurate it is as the real chat builds. 

Recognising the fantastic young people involved 

We’d like to say one more thing, and underline a prime cause for the successful delivery of something we believe to be really relevant. The young people involved in this project.

They came from a variety of backgrounds and lived experiences. They each brought unique perspectives and ideas, as well as having fantastic energy and enthusiasm. We wanted to put our heads together to think about how we can appreciate their involvement in this project. We decided to:

What’s next?

We thought we were lined up for more funding this year, but sadly the organisation had to close its doors. I’ve spoken before about the sustainability of these kinds of projects – and we’re learning all the time. 

Our next task is to develop a “lesson plan” for using the tool in a structured setting, and give it design love. Alongside that, we will be applying for additional funding support with our fantastic youth organisations so that we can:

If we’re to actually keep the tool running past the summer, and improve it, we’ll need money. 

If you know of folk who’d like to support young people learning about AI, we’d love to hear from you. 

And so would Steve.