AI-based research mentors: Plausible scenarios and ethical issues

Guided by Brey’s Anticipatory Technology Ethics, we examined AI-based research mentors (AIRMs) through technology foresight as well as identification and evaluation of ethical issues. Scenario planning was employed to inform foresight, yielding four plausible future scenarios: 1) AIRMs are used solely for guidance, 2) AIRMs are used for guidance and monitoring, 3) AIRMs are banned, and 4) AIRMs are used solely for monitoring. Resnik’s twelve principles informed the identification of ethical issues within these scenarios. Our analysis revealed that certain principles – openness, education, legality, and mutual respect – were violated in all scenarios. Others were contravened to varying degrees across the scenarios; for example, freedom was only violated in scenarios where AIRMs were used for monitoring. Furthermore, the guidance scenario showed that AIRM’s responses could be manipulated to justify poor practice (“AIRMing”). In our evaluation, we weighed ethical issues against the benefits and found that the guidance-only scenario was the least problematic. While this scenario has benefits, such as providing expert guidance on research, ethical issues arise with regard to honesty, openness, credit, education, legality, and mutual respect. Therefore, policy must be developed to ensure that AIRMs are used solely for guidance while mitigating these issues.