Communication channels between users and developers can help to discuss issues, dilemmas or emergent concerns regarding ethical use of machine learning. Users may, for example, wish to raise concerns about (perceived) bias or inquire about how an machine learning system reached a decision.
In order to facilitate communication, increase transparency and obtain feedback for your application, it is recommended to provide safe channels for users to raise concerns.
These channels can be as simple as mailing lists, blogs (e.g., Disqus) or phone numbers. Make sure to include options for anonymisation in order to protect the users’ privacy and be as inclusive as possible.
- Inform Users on Machine Learning Usage
- Explain Results and Decisions to Users
- Test for Social Bias in Training Data
- Provide Audit Trails
- Ethics Guidelines for Trustworthy AI
- Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims