Ten Guidelines for Human-Centered AI Design
-
1. Align AI with human values
Determine the objective function of AI through collaboration with humans to ensure it improves human values.
-
2. Ensure AI transparency
Test various AI models and secure transparency of the models used.
-
3. Ensure data labelling quality
Agree on data labelling early on and continuously audit it during development.
-
4. Engage with a diverse set of users early on
Include a variety of groups from the beginning to the end of the project lifecycle.
-
5. Continue to learn human preferences
Continuously collect data and use appropriate learning models for understanding human preferences.
-
6. Tune AI models with usability testing
Improve AI models by observing how people interact with the products by collecting users’ feedback.
-
7. Co-learn from user feedback
Regularly monitor user experiences and enable users to provide feedback.
-
8. Give humans more control in high-stakes scenarios
Determine the extent of human involvement by considering AI’s uncertain extent of societal harm and liabilities.
-
9. Communicate AI capabilities and limits clearly
Communicate what AI can and cannot do to set reasonable expectations.
-
10. Adhere to the original objectives of AI; otherwise, re-develop
Clarify the initial AI purposes, develop a new AI model, or limit the use when the purpose changes.
Designing Human-Centred AI Products and Services
Designing Human-Centred AI Products and Services
