Making The Play Definition Generator
I wanted to write a 'deep dive' recording the process of designing, making and deploying the Play Definition Generator (PDG) in a way that felt productive and useful for future research. So, I'm attempting to record not only the process, but the feeling of making the thing, the experience of play itself; not just a how-to, but a consideration of if and how I was playing as I made it.
I'm going to experiment with using these quote blocks as a kind of introspective 'aside'. So when you're reading this, you're reading more inner monologue than usual. Perhaps it won't work - but I thought I'd give it a go. Let me know how you feel about this style of writing in the comments below.
The Idea
It started, as most good ideas do, in the shower. This is where me having a cold is particularly relevant, as I took a very long un-environmentally friendly hot shower in a near-desperate attempt to feel somewhat human.
I was thinking about the question we're constantly butting up against designing the methodology for the *Medicine Without Doctors: Reimagining Care and Voice through Play *project: what is play? This is a question I've had to dance around several times now, both in writing my PhD and in some of the papers and conference presentations I've given about it. It's also something I think about every time I read a paper about play, because almost all of them have a similar opening gambit where they accept play is mercurial and hard-to-define, talk about the range of disciplines it touches and either do or don't adopt or create a definition of their own. And I thought it would be funny to create something that took all of this and playfully poked fun at the very notion of definitions of play.
I don't always run with ideas I have - there's too many, or they are often quite diffuse. This was a simple, clear idea though; perhaps the simplicity of it was what drove me to try it in the first place. That, and that the idea of it made me laugh - that's a really important driver for me, I think.
I knew I wanted a simple website that had a click button that generated 'definitions' that ranged from heavily theoretical (i.e. using lots of long words) to the silly and playful (e.g. "play is...dogs in hats"). This split was important, as I felt like it covered both ends of a spectrum that has De Koven - or perhaps, just children - at one end, and the likes of Schiller and Marx at the other.
*FWIW, I'm pretty sure I know where on that spectrum I sit. *
I figured that in order for that to work, you'd need to split sentences down into grammatical chunks and have a system for changing the words around. You'd also need some basic almost template structures for the thing to work around, just to make it work. From a technical standpoint, I thought it would be best to code it as HTML so that it could be easily deployed on a website (or played locally). These were all guesses based on my experience as a designer. Despite having a relatively clear imagine in my head of what I wanted, I didn't know how to achieve it.
*Reflecting on my process in general, when I like an idea I tend to get excited about it and put quite a lot of effort into the early prototype stage. It's the bit I have the most fun in, the bit where I very easily slip into play - I know a thing I want to see realised, even if it doesn't go anywhere from there. I chuckled to myself about the idea and that it was pretty daft, and a bit provocative, but that it might be a fun conversation starter around play. I was excited about the idea of playing with the idea behind the research concepts as much as building something playful. I was getting a real joyful feeling, and wanted to run with that excitement and spontaneity - the fact that this wasn't what I had planned to do with my ever-mounting workload wasn't really relevant. In a week where I felt a bit rubbish, this made me feel quite good. *
The Prompt
I used an AI prompt generation tool to help me write a prompt that ChatGPT might functionally be able to use. I wish I'd saved the original prompt, but as I learn while writing my PhD, making and researching yourself simultaneously is tricky at best, and at worst takes all of the fun out of the situation. You can see some semblance of my direction in the screengrab below.
This then spat out a very long a wordy prompt which I adjusted to include some specifics about the project.
This spat out a set of design summary and a whole heap of code to get the generator going. What I hadn't appreciated, was that I would probably want to see and clarify what ChatGPT was doing behind the scenes before getting it to write a massive chunk of code. What it initially spat out was a surprisingly functional, but grammatically idiotic, generator. There was a lot of useful stuff in there - including the ability to add new words, which I took out but might put back in - but overall it didn't feel as natural as I wanted it to. Also, the design of the thing wasn't at all near what I wanted - unsurprising, seeing as I hadn't given it any visual style prompts.
Iteration
Unhindered, I asked ChatGPT to undertake a wide-ranging review of the literature, and provide a list of pre-existing disciplines and definitions of play, including Piaget, Huizinga etc. and to base sentence structure, and the theoretically-leaning content on these definitions. I told it not to produce any more code until we had developed the theoretical, conceptual and stylistic keyframes of the project.
Prompt #2
Telling it not to write code before we had discussed the details allowed for a much clearer back-and-forth about the back-of-house aspects of the generator. We developed lists of keywords drawn from academic definitions and whimsical sources, built on 12 or so theoretical templates, which would generate a small sample of definitions that ranged from the whimsical and poetic to the heavily theoretical.
ChatGPT then created a "stylistic palette" based on some of disciplines/styles of writing and created a series of "voices" - tonally different combinations of style and template. After a bit of iterative back-and forth, we built on this logic to create a typology that would serve as effectively the database and code that would combine more or less serious/whimsical vocabulary and syntax depending on the voice or discipline that randomly came up.
Early stylistic palette
A suggestion that arose from ChatGPT was the ability to affect the level of whimsy, 'noise' (seriousness) and length of the definition via sliders. This change became a key part of the design of the PDG - the ability to interact, and play with the types of definition you were receiving, or generating. Being able to tweak sliders made it feel as though you were generating the definition yourself; in essence, it made the PDG feel more like a toy.
PDG v2
However, it didn't look like a toy - it looked like quite a dry database website. Further, the grammar and language was all over the place; sentences didn't work, grammar was interrupted (plural nouns not matching singular sentence objects etc.) What was interesting was just how much these minor things stepped me out of the play - the generator only really works when you buy into the conceit that it is offering you real definitions, regardless of how silly they are. "Play is a bubbled mirror of materiality" is easier to play along with than "play is a bubblemateriality that mirror".
This is interesting to me - why? Something about the suspension of disbelief, etc - it takes you out of the magic circle, into the real world, because of how dumb it is when it doesn't work.
I iterated on both the code and the design; the design was tweaked to be far more specific, in line with the Swiss school of design; bold typography on a simple background. As a side note - the code for this was rendered by ChatGPT as "play-is-simple.html" which made me laugh considering the context of the whole thing.
*However, what's really interesting about this is that chance moments like this were accompanied by moments of humour, which encouraged me to keep going on the project - I was making myself laugh, which was pleasureable, so I kept going. *
The code produced the first viable iteration of the PDG; a simple, bold typographic generator that created randomised definitions.
It's telling that I spent some time playing with the sliders and clicking generate just to see what would happen - I was enjoying playing not just with a toy, but with a toy that I had built. There was a sense of satisfaction built into making your own thing that I recognised from making animated short films. On reflection, this isn't a feeling I get quite so strongly when making printed or painted work - there is something about the interactive quality that adds an extra layer of satisfaction and enjoyment. I felt myself chuckling along whenever a 'good' (i.e. funny, poignant) *definition was produced. *
Version 1.1
From there, iteration was rapid. I focused on making it feel more “real”—fixing syntax errors, tightening templates. The sliders still didn’t have enough impact, which was frustrating; playing with a toy that doesn’t respond properly breaks the illusion. After some final tweaks, I uploaded the PDG to neocities.org.
*There was a definite joy in seeing the site live on the internet - being able to share it with people. Even though this is only the first iteration of the idea, I rushed to social media to share it - I was excited, happy. *
There were plenty of tech hiccups—at one point, the HTML file corrupted entirely—but the frustration didn’t kill the play. In fact, solving those problems was oddly satisfying; the challenge itself became part of the pleasure.
It felt like the technical issues that got in the way themselves contributed to the positive and pleasureable feelings associated with making the piece; *the challenge was something to be relished, that wanted solving, rather than something that made me want to walk away from the thing. If anything, times of challenge brought an increase in focus, and an enhanced sense of satisfaction at having solved the problem. *
Conclusions
There is still work to be done on the Playful Definition Generator. I want to expand the language base, making outputs more random and surprising. I’d also like to revisit co-creation and word-suggestion features. But for now, v1.1 has done its job—it got the idea across.
That’s an oversimplification, really. The PDG doesn’t just represent play; it enacts it. It uses humour, irony, surprise, and chance to generate joy. And I had real fun making it.
Through this process, I learned a lot about working with ChatGPT as a design partner—about iteration, prompt-writing, and pacing the collaboration. But maybe the biggest learning was that I enjoy poking fun at play research itself.
I know it’s serious—really, I do. Play is important, political, powerful. But it’s also absurd, and sometimes I think we lose that. Play research, even the accessible kind, can feel heavy; maybe that’s an unavoidable irony of seriously studying play. Still, I wonder if toys like the PDG can inject some joy back into the theory.
Final thought - this is, perhaps, an example of Playful Research; despite the knowledge that it was for a research project, it put play first . The joy of making—the curiosity and silliness—drove it. The learning came second.