Can a person fix the AI mistake in time?
Microsoft researchers studied guidelines for how people use AI tools. In 2019, 49 practitioners applied them to 20 products. A useful test now is whether a person can spot, stop, and fix a mistake in a harmless draft.
What the source reports
A Microsoft Research team published guidelines for designing how people use AI. In a 2019 study, 49 design practitioners applied the guidelines to 20 products. One lesson for a tool today is to make errors visible and give people a way to correct them. Try that idea on a public, low-risk draft. Put a harmless mistake in the input and see whether a person can catch it, stop the run, and recover the draft. This is a design exercise, not a result from a modern, long-running AI system. The older study does not prove that a new tool is safe or that a school gained anything from it.
Original source
- Title
- Guidelines for Human-AI Interaction
- Author
- Saleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, Eric Horvitz
- Date
- 2019