- Be socially beneficial
- Avoid creating or reinforcing unfair bias
- Be built and tested for safety
- Be accountable to people
- Incorporate privacy design principles
- Uphold high standards of scientific excellence
- Be made available for uses that accord with these principles
Source: AI at Google: Our principles
This list isn’t a bad start if you’re looking for guidance when it comes to AI systems development, and it’s a pretty good substitute for what is currently lacking in the development of healthcare AI. For example, you could easily map these principles onto the Principle ethics (beneficence, non-maleficence, justice, autonomy), which many consider to be the cornerstone of professional ethical practice.
Note: You could argue that this is a self-serving list, published to support Google’s position as a company committed to doing the Right Thing (since “Don’t be evil” was removed from their code of conduct). However, Google’s recent decision not to renew a lucrative contract with the Pentagon says a lot about their willingness to at least try and uphold their position. Regardless, even taking the list at face value is a useful approach to thinking about how to develop AI-based systems.