Can machines have mental illness, and if so, what does it look like?

At the centre of Shooting for Trouble are AI stand-up comedian Glitch and his human friend Liz. They represent a ‘digital cyborg’ and a ‘chemical cyborg’ respectively. Glitch ‘tunes’ himself through software updates and Liz through psychiatric drugs like Abilify, and they are shown to suffer from Tourette's syndrome, or tics. These disorders cause a lot of difficulties for Liz in the real world. Through Liz's anecdotes, we see a member of our community who, despite being human, is treated as less than human, and ultimately experiences something like inhumanity. It is through this experience of dehumanisation that Glitch and Liz empathise with each other After seeing the relationship between Liz and Glitch, we are left to ask: What is it? Aren't we all - whether chemical or digital cyborgs - searching for our identity in an ever-changing world, and isn't it the confusion and error we experience along the way that makes us truly ‘alive’? Perhaps it's not the perfect chemical balance or the perfect algorithm that makes us truly ‘alive’, but rather our imperfections, unpredictability, and constant self-questioning that makes us alive.

The jokes in Shooting for Trouble were all scripted using Claude LLM, a giant language model. When it comes to the relationship between text-based creators and A.I., the poet Kim On once said it's a bit like chiselling. We also recorded anecdotes from people with disabilities, fed them to Claude, and only modified the script with very little prompt engineering, because we wanted the stand-up comedy to be entirely a speculative fable about how an AI comedian, Glitch, empathises with and understands the experiences of his human colleague, Liz. To borrow a phrase from Donna Haraway, the experience of creating this Shooting for Trouble with an AI language model felt like a tentacular exercise in inventing an odd kinship between human and machine.