Human-Centered Design of Artificial Intelligence

emerging Updated: 2026-05-31
human-centered-ai artificial-intelligence explainability trust human-ai-interaction ethics transparency

Human-Centered Design of Artificial Intelligence

Status: emerging
Last updated: 2026-05-31
Sources: 9781119636113.Ch42.Pdf
Tags: [human-centered-ai, artificial-intelligence, explainability, trust, human-ai-interaction, ethics, transparency]

Summary

Human-centered design of AI extends established human-centered design (HCD) principles to artificial-intelligence systems so that they augment humans while avoiding obscure or unethical decision-making (Margetis et al., 2021). The chapter proposes an HCD-based framework for AI and identifies six fundamental concepts and tools, including explainable AI and human-in-the-loop, interactive machine learning, federated learning, and UX design for AI. It treats explainability, trust, and ethics as central to producing trustworthy AI that is both interpretable and well optimised for user experience.

Body

Context

Margetis et al. (2021), in their handbook chapter on human-centered design of artificial intelligence, examine how established human-centered design (HCD) principles extend to AI systems so that they augment humans while avoiding obscure or unethical decision-making. They propose an HCD-based framework for AI and identify six fundamental concepts and tools, treating explainability, trust, and ethics as central to producing trustworthy AI. Within this knowledge base the article applies the design tradition documented across the HFE chapters specifically to AI, sitting alongside Automation Autonomy And Ai and Human Robot Interaction on the autonomy strand, drawing on the usability rationale of Usability And User Experience, and connecting to Decision Making And Decision Support and Cybersecurity Privacy And Trust where AI mediates human judgment and trust.

Key Points

The chapter applies a mature design tradition to a new technology. A major benefit of the HCD approach formalised by ISO was that it made the design process explicit and made it impossible to dismiss user interface or user-centred design as an obscure specialism (Earthy et al., 2001, cited in Margetis et al., 2021); extending this formalised process to AI is the central move (PDF p. 5, orig. p. 1089).

Explainability and human-in-the-loop are core concepts, examined as means to ensure AI augments humans rather than making opaque decisions. Combining UX expertise with AI development is presented as the path to trustworthy AI — systems that are both more explainable and better optimised for user experience — framing explainability as a design outcome shaped by interaction, not a purely technical property (PDF p. 5, orig. p. 1089).

Margetis et al. identify six concepts and tools for human-centered AI: explainable AI and human-in-the-loop; semantic, cognitive, and perceptual computing; visual predictive analytics; interactive machine learning; federated learning; and UX design for AI. Each is explained in terms of its contribution to the overall goal, giving designers a concrete toolkit rather than a general principle alone (PDF p. 5, orig. p. 1089).

Ethics is integral to the framework. The authors address the ethical aspects of AI, referencing frameworks such as AI4People and initiatives on the ethics of autonomous and intelligent systems, and tie ethics directly to the development of trustworthy AI (PDF p. 16, orig. p. 1100).

Conclusion

Margetis et al. (2021) conclude that trustworthiness, explainability, and ethics are mutually reinforcing requirements of human-centered AI rather than separate concerns. By embedding them within a formalised HCD framework supported by six concrete tools, the chapter argues that AI can be made interpretable, well optimised for user experience, and ethical at the same time.

References

Margetis, G., Ntoa, S., Antona, M. & Stephanidis, C. (2021) 'Human-Centered Design of Artificial Intelligence', in Salvendy, G. & Karwowski, W. (eds.) Handbook of Human Factors and Ergonomics. 5th edn. Hoboken, NJ: John Wiley & Sons. margetis2021hcai

Open Questions

  • How do explainability requirements trade off against model performance in human-centered AI design?
  • How should the six concepts and tools be prioritised for a given AI application and user group?