“Discoverability” here means the ability for the user to learn the systems abilities. Historically, you could read a menu or list of app icons.
A chat box is empty - it tells the user nothing about the systems abilities, just its interface. A smart speaker is even worse - there’s minimal UX hints. Asking a chat box “what can you do” is not likely to be exhaustive, and and will likely require a series of “can you do X” queries.
We need to get to the point the very least likely response is "I am sorry Dave I can't do that".
Obviously, I am not going to ask Alexa to take out the rubbish; but any reasonable request needs to have a >>50% chance of working; rather than a <<1% chance.
Ask people how they feel about Alexa's "follow up" suggestions ;)
It's not easy to do discoverability in general. Especially not through an intentionally limited modality. People study this stuff, many businesses and product researchers have spent years workshopping ideas. If it was a quick and easy idea like "just tell the user" then it wouldn't be a challenge in 2024.
The problem with "show messages... suggest followups" is that you can't teach the user about new features that are unrelated to their current interaction because it feels like advertising, and its distracting to their current task at hand.
“Discoverability” here means the ability for the user to learn the systems abilities. Historically, you could read a menu or list of app icons.
A chat box is empty - it tells the user nothing about the systems abilities, just its interface. A smart speaker is even worse - there’s minimal UX hints. Asking a chat box “what can you do” is not likely to be exhaustive, and and will likely require a series of “can you do X” queries.