Generative, not creative; how AI highlights the value of researchers, writers, Phoebe Bridgers and a focussing on immediate experience

I wrote my last article about the loss of nuance, about this era of social media and data that ‘encourages us to evaluate media in a direct, literal way.’ It can be hard to pitch nuanced content against this creatively parched context, but the nuanced content that pushes through can hit much harder; hence Greta Gerwig managing to make 1.4 billion dollars with a film about feminism, existentialism and a toy named Barbie. Shortly after writing that article, I attended an ‘AI Hackathon’ event, which made me think about this oncoming era of Generative AI, another technology that may also increase the scarce value of nuance.

The AI hackathon was organised by the AQR (Association for Qualitative Research) in London. For those of you who don’t know, qualitative researchers are like quantitative researchers, but more focussed on words and less focussed on numbers. We run focus groups, in-depth interviews, sometimes we’ll just sit in a public place and observe people. It’s a scary job to have in 2023 with the arrival of generative AI; suddenly computers are a lot better at words than they used to be!

At my previous job I worked in a qual team focussed specifically on media and tech, and a consistent theme in our research was the concern that AI might replace jobs, or make them harder. Fast forward to the hackathon, where we heard from a transformation director named Alasdair Ramage working in the health and wellbeing industries. Ramage told us that this process has already begun for qualitative researchers; big brands’ AI tools are using online data to create clusters and personas, and thus forecast habits and customise adverts. On the more positive side, he discussed how powerful AI could be when combined with regulated data; companies could predict illnesses before they happen and make pro-active health tips. The audience of researchers sipped our coffees and nodded enthusiastically.

After the talks, we split into teams and got our hands on some AI tools. First, we played with a generative AI tool, which was trained to predict the most likely response to a given question (like ChatGPT and Bard). My team asked it to develop an essay about the value of qualitative research, and got back a lot of very basic stuff, but gradually prodded it with more developed prompts until we had some decent content. We joked that this kind of stuff that could fuel a last minute powerpoint, if the writer was too hungover to come up with anything alone. Then we played with CoLoop, an ‘AI copilot for qualitative research’ whose developpers were here with us in the room. CoLoop can analyse interview transcripts and then answer questions about them; we could ask if the older interviewees had reported a certain view, or ask it to summarise responses to a certain question. Hangover or no hangover, this could save researchers some serious time in navigating notes and confirming intuitions.

Once we were done with the tools, we had a general discussion about AI and its role in qual. Reflecting on ChatGPT and CoLoop, I suggested that these tools could be helpful for zoning in on novel, unexpected insights. Say you’re researching the impact of Covid on enjoyment of superhero films, ChatGPT could predict the more obvious, generic answers (eg: a lying government makes us value responsibility and honesty in superheroes) and – after you interview a bunch of real people on the subject - CoLoop could summarise all the actual answers to these questions. ChatGPT’s predictions would probably come up, be it because people agree with them or because they can easily imagine agreeing with them, and doubtlessly there would also be some other, less obvious answers. A researcher could quickly cross reference ChatGPT’s predictions and CoLoop summaries, eliminating the former from the latter, and theoretically be left with a page of pure nuance, just the unexpected stuff.

Alasdair Ramage had talked a lot about the philosophy of AI and what makes it unique, and my suggestion prompted him to compare it to qualitative research. He said that ChatGPT is designed to replicate norms and avoid nuance, hence why it can write something in the style of P Diddy or Adele, it is designed to predict the most obvious thing that someone might say. Ramage said that he hadn’t spent much time working with qualitative researchers, but he got the sense that the long and short of our job is finding nuance, the exact opposite goal.

I think there is something poignant in this, not just as a pointer to the limitations of AI, but also as a reminder of what makes a really good writer. ChatGPT develops content by churning through noise and discussion and Google results, with no heed for accuracy (the phenomenon of AI answering a question with a falsity is popularly described as a ‘hallucination’). The best writers and researchers do the opposite; they find moments of truth outside of social constructs and popular narratives.

One of my favourite writers is a singer/songwriter named Phoebe Bridgers. In her amazing 2020 single Garden Song, Bridgers sings ‘the doctor put her hand over my liver, she told me my resentment’s getting smaller.’ The lyric works brilliantly as a metaphor for emotional development, but it’s not actually a metaphor. In a 2020 Playboy interview, she revealed ‘that was a nutritionist in Los Angeles who literally did that to me’, before commenting that ‘my imagination is not as creative as my reality.’ The best writers are awake to those strange glitches in the matrix, the glimpses of profound depth that happen when someone speaks their truth. Some, very rare writers can stretch open those peekholes and expose whole situations, real experiences that make us question our assumptions.

Of course this isn’t always how creativity or research works, there’s a lot of value to be found in plugging oneself into the Matrix and analysing prevalent themes, or letting all of that noise subliminally guide one’s words. Bridgers’ approach, the approach of tuning out all that background static and documenting the literal reality infront of us, it can seem like a lazy way to create content, one might even wonder if it’s even a creative act at all. But analysing pure, unfiltered reality is slow, tiresome work. Reality can be boring, it’s not easy find these special, inspirational moments, and even when something unexpected does happen, it takes a creative mind to know whether this strange moment is worth focussing on; Bridgers could have dismissed her nutritionist’s comment as hippy crap. She didn’t, she knew that there was something in this moment, something real.

Before I wrap up this article, I want to take one more left turn. I’ve recently been enjoying an amazing 2008 book named Don’t Sleep, There Are Snakes by Daniel Everett, an ex-Christian missionary who spent decades living in the Amazon with the Pirah? tribe, deciphering their unique language. This tribe feels profoundly seperate to us in geography and culture; they have no creation myths or history and no cultural practice of sharing fiction. They love to share stories, but these stories are always based on immediate experiences; whether this be the experience of fighting a tiger, or of exploring another spiritual plain through dreams (the Pirah? classify dreams as real experiences). They keep few possessions despite the various challenges that surround them; ‘don’t sleep, there are snakes’, is a popular phrase said before bedtime, playfully joking about the variety of predators. Their sleep rarely lasts for more than 3 hours at a time, there is always noise in the settlement, but the Pirah? consistently amazed Everett with their positivity, good humour and energy.

I wonder if the Pirah?s’ focus on immediate experience is connected to their resilience; there must be some reason that these tribes people can be so full of energy despite sleeping so little. Their culture is direct and truthful, it is not fogged up with so many layers of abstract social constructs. This does not mean that their life is more simple than ours, it's just that they focus on the complexity of the immediate world, they source meaning from the abundance of life all around and inside them. And in this sense, they are the perfect comparison to generative AI, the ultimate abstract construct, seperate from the context of life.

AI can access an enormous amount of human data and culture and history, but it cannot access the present. It cannot look at the world as it is now, it cannot experience the essential human experience of consciousness. I wonder if this new technology will make it easier for people to appreciate the power of our oldest, most inherent human capabilities, the abilities to detect truth and flavour in life.