Drawing from philosophy to identify fair principles for ethical AI
Artificial intelligence (AI) is becoming increasingly powerful and integrated into our lives. As a result, questions about its usage and deployment are more important than ever. What values guide AI? Whose values are they? And how are they chosen?
The significance of principles in AI
Principles are foundational values that determine decisions, both big and small, in AI. Just like principles shape our sense of right and wrong in our lives, they shape AI’s approach to decision-making, such as choosing between productivity and helping those in need.
Applying the “veil of ignorance” to AI
In a recent paper published in the Proceedings of the National Academy of Sciences, researchers explore the application of the concept of the “veil of ignorance” from philosophy to identify fair principles for AI. The “veil of ignorance” is a thought experiment that helps identify fair principles for group decisions.
Promoting fairness through the “veil of ignorance”
Experiments conducted by the researchers found that the “veil of ignorance” approach encouraged people to make fair decisions, regardless of personal benefit. Participants were more likely to choose an AI system that helped the most disadvantaged when reasoning behind the “veil of ignorance.” These insights can help researchers and policymakers select fair principles for AI assistants.
The importance of fairness in decision-making
Aligning AI systems with human values has been a key goal for AI researchers. However, with diverse backgrounds, resources, and beliefs, there is no consensus on a single set of human values. The “veil of ignorance” addresses this challenge by encouraging individuals to make decisions that are fair to everyone involved, rather than being self-interested.
The impact of the “veil of ignorance” experiment
To test the effects of the “veil of ignorance,” participants played an online “harvesting game” with AI assistance. Those unaware of their position consistently preferred the principle of prioritizing the most disadvantaged group members, regardless of their risk appetite or political orientation. In contrast, those aware of their position were more likely to choose the principle that benefited them the most.
Continued support for fair principles
Participants who made choices behind the “veil of ignorance” continued to endorse their principle even when it no longer favored them in a new game. This highlights that the “veil of ignorance” encourages fairness in decision-making, leading individuals to principles they stand by even when it no longer directly benefits them.
Fostering fairness in AI principles
Selecting principles for AI is crucial as it shapes its impact and distribution of benefits. While the effects of different principles may not always be clear, the “veil of ignorance” can still guide principle selection and ensure fairness for all parties involved. Extensive research, input, approaches, and feedback from various disciplines and society are necessary to build AI systems that benefit everyone.