Human Factors in Design of A3: Automation, Autonomy, and Artificial Intelligence¶

Status: emerging
Last updated: 2026-05-31
Sources: 9781119636113.Ch52.Pdf
Tags: [automation, autonomy, artificial-intelligence, human-automation-interaction, trust, design-frameworks, a3]

Summary¶

Automation, autonomy, and artificial intelligence — abbreviated A3 — are technologies that extend human ability by contributing self-produced, non-human effort (Sawyer et al., 2021). The chapter treats them together as a set of computational tools that learn from data and act in reasonable, even human-like ways, and examines how humans interact with A3 systems. It introduces concepts such as human-A3 teams of teams, function allocation across levels, tandem failure and mutual reinforcement, and machine ethics.

Body¶

Context¶

Sawyer et al. (2021), in their handbook chapter on the design of A3 — automation, autonomy, and artificial intelligence — examine these technologies together as computational tools that learn from data and act in reasonable, even human-like ways, and study how humans interact with them. They introduce human-A3 teams of teams, function allocation across levels, tandem failure and mutual reinforcement, and machine ethics. Within this knowledge base the article is the synthesising node for the machine-partner strand of human-centered design: it generalises the AI-specific treatment in Human Centered Design Of Ai, the monitoring focus of Supervisory Control Of Automation, and the embodied case of Human Robot Interaction, and connects them to the error-compounding concerns of Human Error And Reliability and the applied automation of Aviation Human Factors.

Key Points¶

A3 groups three related technologies as extensions of human ability that contribute self-produced, non-human effort, encompassing computational tools that learn from data and systems that act in reasonable, even human-like manners. Sawyer et al. trace the pursuit of such computing to at least the 1950s, when Simon predicted machines capable of doing any work a person can do, a vision now associated with Artificial General Intelligence (PDF p. 1, orig. p. 1385). Treating the three together reflects their convergence in modern systems, where A3 can share the load to extend human ability (PDF p. 2, orig. p. 1386).

The chapter centres on human interaction with A3 systems. Sawyer et al. develop the idea of human-A3 teams of teams, in which humans and A3 systems collaborate, and address function allocation and levels of autonomy between humans and A3 (PDF pp. 3–5, orig. pp. 1387–1389). This extends classic levels-of-automation thinking to systems where the machine partner is increasingly capable and autonomous, raising new questions about who does what and when.

Failure modes receive specific attention. Sawyer et al. introduce tandem failure and mutual reinforcement as phenomena in human-A3 systems, where human and machine errors can compound rather than compensate for each other (PDF pp. 17–18, orig. pp. 1401–1402). They invoke a "nocere" framing — the principle of not causing harm — applied to A3, signalling that these systems can amplify as well as mitigate risk (PDF p. 19, orig. p. 1403).

Design strategies, ethics, and future challenges complete the chapter, covering security design and design strategies and frameworks (PDF pp. 15–16, orig. pp. 1399–1400), and A3 and machine ethics and future challenges in A3 design (PDF pp. 20–21, orig. pp. 1404–1405).

Conclusion¶

Sawyer et al. (2021) conclude that safe A3 design is a balance between the augmenting and the failure-amplifying potential of these technologies. By tying design frameworks to ethics, security, and trust, they position A3 design alongside the human-centered design of AI and the concerns of supervisory control and human-robot interaction, where preventing tandem failure and allocating function appropriately become central as machine capability grows.

References¶

Sawyer, B.D., Miller, D.B., Canham, M. & Karwowski, W. (2021) 'Human Factors and Ergonomics in Design of A3: Automation, Autonomy, and Artificial Intelligence', in Salvendy, G. & Karwowski, W. (eds.) Handbook of Human Factors and Ergonomics. 5th edn. Hoboken, NJ: John Wiley & Sons. sawyer2021a3

Open Questions¶

How can design prevent tandem failure, where human and A3 errors reinforce rather than compensate for each other?
How should function allocation between humans and A3 adapt as machine capability and autonomy increase?