I’ve been thinking about the affordances of GenAI chatbots…
There are hidden affordances, where it has capabilities not immediately apparent. There are also false affordances, where the chatbot seems to be usable for a certain thing, but it actually isn’t. Worst still, these false affordances are obscured by the chatbot’s tendency towards cheery sycophancy.
An example of a hidden affordance might be a chatbot’s ability to write and execute code which you stumble upon inadvertently. It wasn’t apparent it could do this until you stumble across it.
An example of a false affordance might be asking it to format your references in Harvard style based on incomplete information. It “completes” the task, but hallucinates the missing information instead of retrieving it.
I had a student of mine get caught out by this with Copilot embedded in Microsoft Word a couple of years ago. We had a grown up chat about it and it was handled with empathy.
This is one of those horrible false affordances of GenAI. It seems to be usable for a certain thing, it says it can do the thing, but actually doesn’t do the thing, it happily and confidently hallucinates the thing.
