Can AI errors help users? Upol Ehsan and coauthors propose that Seamful XAI design can be leveraged to boost users' understanding and agency. "Acknowledging AI is imperfect does not entail that we give up and do not work on problems [61]. It simply means we are realistic about what we can solve and opportunistic about the rest."
Read the full manuscript on arXiv.
You might also like
- communityThe apocalypse that wasn’t