Key Takeaways

• Current ethics oversight relies too heavily on review boards: The paper identifies that existing ethical oversight systems place "full responsibility for ethical reflection in the hands of review boards and committees," suggesting this approach alone is insufficient for ensuring responsible AI development.

• Training researchers in ethics is a practical alternative: Rather than depending solely on external review processes, the authors propose that AI researchers themselves should receive structured ethics training to develop their own capacity for ethical reflection and decision-making.

• Ethics training can be integrated into doctoral programs: The case study demonstrates that ethics education can be systematically incorporated into AI researcher training programs, suggesting this approach is feasible and implementable at the educational level.

• Researchers develop measurable ethical capabilities through training: The paper presents evidence from second-year doctoral students showing that structured ethics training produces "newly found capabilities as ethically responsible researchers," indicating that ethical reasoning skills can be taught and developed.

• A distributed responsibility model may be more effective: By shifting some ethical responsibility from committees to individual researchers through education, organizations can create a culture where ethical considerations are embedded throughout the research process rather than treated as a final checkpoint.