Just a quick note to highlight this new report on the use of 'dark patterns' in chatbot design. This is something I've touched on before (and have a paper relating to, in the deep abyss of academic procedure) but it's really good to see these things being highlighted from people with a far deeper focus on the purely design side than I have.
There's a lot of insidious, anthropomorphisation behind chatbots in general, never mind specific faux-social LLMs and very little of it is that obscure or hidden. As this report shows there's a lot of 'tool's being used which are entirely familiar from social media and digital marketing. We're not in any position to be surprised by what's happening, although we should absolutely be disappointed in the lack of lessons being learnt from past experience.
The paper does come to the usual point of guidelines, guardrails and legislative measures which... well, your mileage may vary but I'm not all that convinced that tech companies are willing to modify their own behaviours. Certainly not while there's money in not bothering (and, for some, looming IPOs on the horizon). Still, there's plenty here that's well worth checking out.
- Dylan