Ethical use of artificial intelligence in industrial-organizational psychology research and practice
Citation. Landers, R. N., & Nakamoto, S. (2025). Ethical use of artificial intelligence in industrial-organizational psychology research and practice. Practice Innovations. Advance online publication. https://doi.org/10.1037/pri0000310
Summary. Landers and Nakamoto argue that AI is transforming I-O psychology practice across recruitment, selection, training, monitoring, workforce relations, DEI, and work redesign, but that the ethical problems are extensions of older I-O tensions around validity, worker rights, organizational control, and scientific responsibility. They recommend grounding AI decisions in the APA Ethical Principles and domain-specific frameworks, then using I-O expertise proactively rather than waiting to study harms after deployment. The article uses case studies on AI interviewing, training chatbots, and worker replacement to show that technical performance is ethically insufficient without transparency, accountability, validation, worker dignity, and attention to who bears the costs of AI-mediated change.
Key claims
- AI creates ethical challenges across the full employee lifecycle: recruitment chatbots and search tools, AI-powered matching and assessment, automated job descriptions, customized training and onboarding, performance monitoring and feedback, HR chatbots, employee-communication tracking, and fairness audits.
- Over-eager AI adoption often happens before validation, safeguards, or regulation, so I-O practitioners need to evaluate specific work domains and affected worker groups rather than treating AI adoption as a generic efficiency upgrade.
- AI systems can intensify old selection ethics problems because predictors of future performance are contestable, seemingly objective models can absorb biased training data, and asynchronous or AI-scored interviews raise disclosure, consistency, cultural context, and model-development questions.
- AI monitoring and behavior-modification systems can support safety and performance but also threaten privacy, autonomy, psychological safety, and worker well-being when they make employee life more visible to employers without clear boundaries or worker control.
- Differential access to paid, powerful, English-dominant AI systems may advantage workers and organizations with more money, infrastructure, language fit, internet quality, and technical expertise, producing unequal AI benefits across socioeconomic, demographic, and national lines.
- AI-enabled job displacement and reskilling revive long-standing ethical conflicts between organizational efficiency and worker rights; the article cites estimates that generative AI could perform tasks occupying 60%–70% of employee time and survey evidence that 38% of U.S. workers worry AI could make some or all of their duties obsolete.
- The APA Code’s principles of beneficence and nonmaleficence, fidelity and responsibility, integrity, justice, and respect for rights and dignity remain relevant to AI: I-O practitioners should forecast harm, disclose AI use and limits honestly, reduce bias, protect privacy and autonomy, and remind organizations that humans remain accountable for AI decisions.
- Domain-specific frameworks are necessary but insufficient by themselves because AI-specific checklists can become obsolete quickly, while first principles are too abstract to apply without contextual guidance.
- In AI training-chatbot deployments, the boundary of accountability must be drawn explicitly: if workers are told that the chatbot is authoritative, errors belong to the system and organization; if workers are trained to vet chatbot information, performance accountability shifts differently.
- I-O practitioners should neither reject AI categorically nor rush blindly into adoption; they have an ethical responsibility to stay engaged, ask uncomfortable questions, and vet AI systems before worker harm occurs.
Evidence & limitations
- Conceptual/practice article, not an empirical study; the evidence comes from ethical frameworks, historical I-O examples, regulatory and professional guidance, and three illustrative case studies.
- The source is valuable for ethical framing and practice guidance, especially where I-O expertise intersects with AI deployment decisions, but it does not estimate causal effects of any specific AI system.
- Because the article is broad across I-O domains, topic synthesis should avoid treating it as evidence for performance gains and instead use it as a boundary-setting source on validation, accountability, worker dignity, and governance.