Human-Centered AI and Explainability

Human-centered AI (HCAI) prioritizes augmentation, user agency, and inspectability over automation maximalism (George, 2025).

1. Augmentation vs replacement

• replacement logic: optimize task completion by removing human involvement,
• augmentation logic: improve capability while preserving control and accountability.

HCAI adopts augmentation as the default design target.

2. Risk classes

• opacity of model decisions,
• bias amplification,
• hallucination propagation,
• automation-induced deskilling,
• autonomy erosion through over-personalized defaults.

3. Explainability as interaction requirement

Explainability supports three user-facing objectives:

• prediction trust calibration,
• traceability of system behavior,
• decision comprehension (George, 2025).

A practical design objective is:

Operationally relevant questions: who decides, who can override, and where explanation is required before action.

co-authored by an AI agent.

references