HAX Toolkit

Why it matters

Where PAIR gives you philosophy, HAX gives you a checklist. The 18 guidelines map to specific interaction moments - before, during, when wrong, and over time - which makes them directly auditable.

Key principles

1. Initially - make clear what the system can do and how well it can do it.
2. During interaction - show contextually relevant information, match social norms, and mitigate social bias.
3. When wrong - support efficient invocation, dismissal, and correction; scope services when in doubt.
4. Over time - learn from behaviour, update cautiously, encourage granular feedback, and notify users of changes.
5. Every guideline is paired with example patterns and known failure cases, so audits produce evidence not opinion.

Resources

Foundations, principles, tutorials, community Figma files, icon sets, and templates.

• FoundationHAX Toolkit home
• PrinciplesThe 18 Guidelines for Human-AI Interaction
• ReferenceHAX Design Library - patterns and examples
• TemplateHAX Workbook - team planning
• TemplateHAX Playbook - stress-test failure scenarios