HCAI Design Reflection Tool

A classroom heuristic for discussing whether an AI design supports human capabilities, agency, understanding and dependable interaction. It is not a validated psychometric scale or an AI-ethics compliance test.

1. Describe the scenario

1 — mostly human action3 — shared5 — mostly automated

First discuss the scenario as a group. Then score the design. A high score means that the scenario/design gives stronger evidence for that HCAI dimension; it does not mean that every HCAI system should maximise every dimension.

2. HCAI profile

3. Design choices

Based on your assessment, propose two concrete design changes. Try to connect each choice to one or more HCAI dimensions. Focus on what you would actually change in the AI system, the human-AI interaction, the workflow, or the context of use.

3. Results: your HCAI profile

4. Human control × AI automation

This is deliberately kept separate from the radar score. Shneiderman argues that human control and computer automation are distinct dimensions: high automation is not automatically “more HCAI”, and the appropriate position depends on the task and context.

High human control
Low automation
Useful where mastery, exploration or independent human action is central.
High human control
High automation
Often desirable for complex tasks where automation assists while people retain meaningful control.
Low human control
Low automation
Limited assistance; may be appropriate for simple devices, but offers little augmentation.
Low human control
High automation
Can be appropriate for rapid, predictable actions, but requires strong reliability, safeguards and monitoring.
AI automation: low ← → high   |   vertical position: human control low → high