Chapter 16 - Responsible AI, Ethics, and Calibrated Reliance
Last updated 2026-07-13. Review the responsible-AI primary sources behind Chapter 16, then use the reading paths and operator obligations checklist to connect them to the architectural layers.
Companion activity
Use the Responsible AI page to map one obligation to the architectural layer that can actually implement, log, or review it.
What this chapter argues
The architecture is a tool, and tools carry obligations. Chapter 16 maps responsible-AI duties onto the layers that can implement them rather than treating governance as a separate overlay. It distinguishes generated uncertainty signals from measured calibration: an AI system saying it is unsure can help a person pause, but only evaluation against outcomes shows whether confidence tracks correctness. Over-reliance and under-reliance are twin failure modes. The architecture's job is to support informed human judgment through evidence, meaningful review, appeal, and proportionate control - not to replace that judgment.
Key sources
Chapter 16 cites in the print book: D13, D14, D15, D16, D17, D18, D19, D20, D39, D40, D41, D44, D45, D47, D48, D49, D67, D79, D80, D81. Entries below are the sources this page draws on; some are related literature cited elsewhere in the book rather than in this chapter.
- D13 - Microsoft. (2022). Microsoft Responsible AI Standard v2: General Requirements. Microsoft RAI v2 PDF
- D14 - Microsoft Learn. (2024). What is responsible AI?. learn.microsoft.com/azure/machine-learning/concept-responsible-ai
- D15 - NIST. (2023). AI Risk Management Framework (AI RMF 1.0). nist.gov/itl/ai-risk-management-framework
- D16 - European Commission. (2024). Regulatory framework for artificial intelligence (EU AI Act overview). digital-strategy.ec.europa.eu
- D17 - Sweller, J. (1988). Cognitive load during problem solving. doi.org/10.1207/s15516709cog1202_4
- D18 - Bainbridge, L. (1983). Ironies of automation. doi.org/10.1016/0005-1098(83)90046-8
- D19 - Parasuraman, R., and Riley, V. (1997). Humans and Automation: Use, Misuse, Disuse, Abuse. doi.org/10.1518/001872097778543886
- D20 - Lee, J. D., and See, K. A. (2004). Trust in automation: Designing for appropriate reliance. doi.org/10.1518/hfes.46.1.50_30392
Primary-source links
The chapter's load-bearing primary sources, organized by what they ground:
- Industry RAI frameworks - Microsoft RAI v2, Microsoft RAI overview.
- Government frameworks - NIST AI RMF 1.0, EU AI Act overview.
- Cognitive-load and trust foundations - Sweller (1988), Lee & See (2004).
- Operator-burden foundations - Bainbridge (1983), Parasuraman & Riley (1997).
Chapter contents
- The metaphor's limits, fully unpacked (extension of Chapter 4's acknowledgement).
- Responsible AI principles (fairness, reliability/safety, privacy, inclusiveness, transparency, accountability) mapped to architectural layers.
- Calibrated reliance and the twin failure modes (over-reliance, under-reliance).
- Cross-disciplinary touchpoints (psychology, philosophy, cognitive science, STS).
- What the architecture owes by existing. What this book does not solve.
Responsible AI resources
The Responsible AI page hosts the full cross-disciplinary reading paths (organized by discipline) and the operator obligations checklist mapped to which layer carries each obligation. That page is the practitioner-side companion to the chapter's principles-side argument.