Guidelines for Human-AI Interaction
Eighteen guidelines for designing interaction with AI systems, across four moments: initially, during interaction, when the system is wrong and over time.
What problem it addresses
It gives the team a concrete checklist for reviewing an AI product before launch.
Components
Initially
Explain what the system can do and how reliably.
During interaction
Show what matters at the right moment and in the right context.
When it’s wrong
Make it easy to dismiss, correct or undo.
Over time
Learn from what the person does and give notice of changes.
How it works
The guidelines are grouped into four moments of the interaction.
At each moment, the team checks whether its product follows them.
Whatever falls short is fixed before launch.
Applications in design
Use it like this
As a review checklist for an AI product, as a team.
Avoid
Applying them only at the end, when nothing can be changed.
Don’t fall for it
Test them with real situations, not just the ideal demo.
Connections
Used in
Authors and works
- 2019
Saleema Amershi et al.
Compiled them from earlier research and tested them with 49 design practitioners on 20 AI products.
Sources
Original sources
- Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., Suh, J., Iqbal, S., Bennett, P. N., Inkpen, K., Teevan, J., Kikin-Gil, R., & Horvitz, E. (2019). Guidelines for human-AI interaction. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (pp. 1–13). ACM. View source: Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., Suh, J., Iqbal, S., Bennett, P. N., Inkpen, K., Teevan, J., Kikin-Gil, R., & Horvitz, E. (2019)
Last reviewed: