Get 20% off on all plans of User Evaluation – the AI for Research. Use the code LIMITED20. Sign up today!20% off User Evaluation, the AI for research: code LIMITED20
AI UX patterns

HAX guidelines

Updated UX Pickle editorial team

Definition

Microsoft's 18 Guidelines for Human-AI Interaction: design rules for AI features, grouped by when they apply, from first use to long-term use.

The HAX guidelines are 18 design guidelines for AI features, published by Saleema Amershi and colleagues at Microsoft in the CHI 2019 paper Guidelines for Human-AI Interaction. Microsoft maintains them, with design patterns and examples, in the HAX Toolkit.

Where they come from

The authors consolidated more than 150 AI-related design recommendations, then tested them in several rounds, including a user study with 49 design practitioners who tested them against 20 popular AI-infused products. The paper groups them by when they apply as "Initially" (G1-G2), "During interaction" (G3-G6), "When wrong" (G7-G11) and "Over time" (G12-G18).

The 18 guidelines

Wording below is from Microsoft's HAX Design Library.

Initially

  • G1. Make clear what the system can do. Help the user understand what the AI system is capable of doing.
  • G2. Make clear how well the system can do what it can do. Help the user understand how often the AI system may make mistakes.

During interaction

  • G3. Time services based on context. Time when to act or interrupt based on the user's current task and environment.
  • G4. Show contextually relevant information. Display information relevant to the user's current task and environment.
  • G5. Match relevant social norms. Ensure the experience is delivered in a way that users would expect, given their social and cultural context.
  • G6. Mitigate social biases. Ensure the AI system's language and behaviors do not reinforce undesirable and unfair stereotypes and biases.

When wrong

  • G7. Support efficient invocation. Make it easy to invoke or request the AI system's services when needed.
  • G8. Support efficient dismissal. Make it easy to dismiss or ignore undesired AI system services.
  • G9. Support efficient correction. Make it easy to edit, refine, or recover when the AI system is wrong.
  • G10. Scope services when in doubt. Engage in disambiguation or gracefully degrade the AI system's services when uncertain about a user's goals.
  • G11. Make clear why the system did what it did. Enable the user to access an explanation of why the AI system behaved as it did.

Over time

  • G12. Remember recent interactions. Maintain short-term memory and allow the user to make efficient references to that memory.
  • G13. Learn from user behavior. Personalize the user's experience by learning from their actions over time.
  • G14. Update and adapt cautiously. Limit disruptive changes when updating and adapting the AI system's behaviors.
  • G15. Encourage granular feedback. Enable the user to provide feedback indicating their preferences during regular interaction with the AI system.
  • G16. Convey the consequences of user actions. Immediately update or convey how user actions will impact future behaviors of the AI system.
  • G17. Provide global controls. Allow the user to globally customize what the AI system monitors and how it behaves.
  • G18. Notify users about changes. Inform the user when the AI system adds or updates its capabilities.

How to apply them

  • Do use them as a heuristic review, as the authors did: go guideline by guideline and note where an AI feature applies or violates each one.
  • Do map them to concrete UI. In a meeting-notes summariser, G2 could be a note that summaries can miss action items, G9 an inline edit on each bullet.

Common mistakes

  • Treating G11 as a demand for full technical explanations. The user needs access to an explanation, which can be short and partial (see explainability).
  • Applying G13 without G16 and G17, so people see the product change but can't tell why or switch it off (see personalization controls).
  • ExplainabilityHow well people can understand why an AI system produced an output. Often paired with interpretability: what that output means for their task.
  • AI feedback loopsWays users tell an AI product what worked or failed, such as thumbs up/down, edits and corrections, and how the product shows that input had an effect.
  • Personalization controlsSettings that let people see, adjust, pause or reset what an AI has learned about them and how it uses that to tailor results.
  • AI onboardingHow an AI product introduces itself: what it can and can't do, how to get good results, and how it will change, taught in short steps when needed.
  • Calibrated trustWhen how much people trust an AI system matches how reliable it actually is, avoiding both over-reliance and under-reliance.

Sources

  1. Amershi, S. et al. (2019). Guidelines for Human-AI Interaction. CHI 2019, Glasgow
  2. Microsoft HAX Toolkit: HAX Design Library (the 18 guidelines)
  3. Microsoft HAX Toolkit: Guidelines for Human-AI Interaction

Browse the full AI UX glossary