Heer, J. (2019). Agency plus automation: Designing artificial intelligence into interactive systems. Proceedings of the National Academy of Sciences, 116(6), 1844–1850

Lovely paper suggesting a key design principle for designing for IA (IA vs AI): shared representations which can be created by either human or machine. This allows the machine to suggest possible domain actions which can be fluidly adapted/adopted by the user, and for the user’s actions to be understood in the same terms by the ML system.

Contains discussion of a particularly thoughtful mixed-initiative translation interface:

Q. What key principle does Heer (2019) identify for designing ML-supported IA-style interfaces?
A. Shared representations for domain actions which can be created by either human or machine, fluidly adapted by the former, understood by the latter.