Responsible AI principles
As mentioned, there are six core principles – Fairness, Reliability and Safety, Privacy and Security, Inclusiveness, Transparency, and Accountability – that Microsoft has incorporated into their Responsible AI Toolbox. We will briefly explain these as follows:
- Fairness in the context of AI systems is a sociotechnical challenge that scientists and developers need to address to ensure that people are treated equally and to reduce unfairness relating to the specific use case we are building the model for. In a variety of domains and use cases, AI systems can be used to provide resources and opportunities, and through checking fairness we can ensure that we are not reinforcing existing stereotypes.
For example, if we are predicting what ethnic groups there are in a certain segment of the population and what their food preferences are, not all customer segments of cities or states have the same ethnic diversity. So, when we design the...
