Late last year I was having a coffee with narrative designer, historian and researcher Holly Nielsen. I always enjoy talking to Holly about games because she has a great supply of anecdotes both about her work in today's games industry, and also about her research on the games industry of a hundred years ago. We often talk about AI, and the arguments people use to criticise it (which we largely agree with), and she said something that's kind of stuck in my head since: people often point out the errors AI make in things like writing or art, but that didn't seem to her like a very good basis on which to criticise it.
She's right. AI may be better than you one day. It might already be, right now. If you want to organise, criticise or simply understand AI better, you need to come at it from a different angle.
I really want to highlight Mike's point here about what we ask of players. This was actually something I was talking about back in 2019 when looking at what constitutes interactive drama and the problems much of that work runs into.
These systems, LLM based characters included, forget that they are asking usually a novice person to do very involved improv and roleplay basically completely by themselves. I've characterized this as a blank page problem. If you let someone do anything, without training and support they will default to being themselves and it is very difficult to maintain dramatic tension, suspension of disbelief, and even just the fantasy of being someone else, if you are stuck only saying what you can come up with by yourself.
By removing many of the affordances of chatacter, we are putting the onus on players to support the fantasy we are giving them. This is a recipe for fatigue and homogeneity across an experience.
I'm a writer, one who can sit down and make pretty good first drafts of stories even. That does not mean I am good at doing improv. Most people aren't. In a TTRPG group, there's at least the human connection and collective building of the fantasy even if no one is performing brilliant dialogue. Remove the people from that side of the equation and like Mike said, you are left being awkward in your living room or at your desk.
Traditionally game design is about delivering some kind of fantasy to players and, importantly, supporting them in their enactment of that fantasy. By letting players do anything in that fantasy, some of the support for enacting it is removed, and engaging with the fantasy becomes harder.
I also want to point out that the Holodeck, that magical device that has captured research and industry's imagination for decades in fact breaks when it is given out of context problems. As smart as the characters in the programs are, they have a limited set of things they can do. And something else to keep in mind, everyone in the holodeck is a professional actor with professionally written dialogue. Me improving in the exact same scenario as Patrick Stewart will not give me the same experience as being able to act and speak the most effective lines I can the way Patrick Stewart does
