Researchers Advocate Applying Research Ethics Principles to AI Training Data
• If AI systems are trained on biased data, they can make biased judgments that impact hiring, loans, benefits, etc.
• Researchers suggest applying existing ethical principles for human subjects research to AI training data.
• Core principles are respect for persons, beneficence, and justice from the 1979 Belmont Report.
• These principles could ensure transparency, consent, and fairness in AI training data.
• Authors advocate thoughtfulness and ethical data practices by private sector, not more regulation.