What are the Guidelines for Human-AI Interaction?
The Guidelines are a set of 18 generally applicable best practices for designing human-interaction with AI-based products and features. The Guidelines are divided into four groups, based roughly on when they are the most relevant to interaction with an AI system: upon initial interaction, during interaction, when the AI system is inevitably wrong, and over time.
How do I use the Guidelines?
Use the Guidelines throughout your development process, when you have an idea for a new AI-based product or feature or are revising an existing one. Because implementing the guidelines typically impacts all aspects of an AI system, including what data to collect and how AI models should be trained, the earlier you consider them the better.
Bring multiple disciplines to the table during planning and requirements meetings to think through which Guidelines are most important to your product’s user experience and which implementation strategies to use. The HAX Workbook can help align and guide your team in early planning with the Guidelines and our HAX Design Library and design patterns can give you ideas for how to implement them.
Keep in mind the guidelines are not a checklist, and not all might apply in every scenario. Some scenarios might need additional, specialized guidance.
Learn about each Guideline and see available design patterns and examples in the HAX Design Library.

Why the Guidelines?
We created the Guidelines because AI is fundamentally changing how people interact with computing systems, and practitioners were asking for guidance, saying, “[This is] the most ambiguous space I’ve ever worked in, in my years of working in design … There aren’t any real rules and we don’t have a lot of tools.” To build a foundation for creating intuitive AI user experiences, we synthesized and validated more than 20 years of research and design thinking to create the Guidelines.
What teams say about the Guidelines
[This is] all about creating trustworthy assistive systems … what really matters here is the model for human interaction with a system that has been given tremendous authority over human life.
Senior UX designer
It took us four years to come up with 80% of [the Guidelines] … if you can institutionalize this into the design parts of products, you give people an opportunity to build a much better product in Gen One.
Program manager
It has really made a difference in my life as a designer and I want everyone to know about it and use it.
UX designer/Instructor
Ever since I first read them, I’ve found myself bringing them up quite often in specs and design reviews, and they’ve been fantastic as a framework for exactly what they were intended for. I even have a poster of the guidelines up in our hall ?. And perhaps the best thing is that every time I cite the Guidelines, there is someone new in the room who, after reading, then becomes a new evangelist.
UX researcher
Learn more about the Guidelines
- The Guidelines were introduced in a 2019 CHI paper that won a best paper honorable mention award. Read a summary of the paper or learn more about the process we used to create and validate them on the About page.
- To learn more about other ways teams are using the Guidelines in their creative design process, see our Medium post, AI Guidelines in the Creative Process: How we’re putting the Human-AI Guidelines into practice at Microsoft.
- To understand how the Guidelines impact the entire AI system, we offer considerations for machine learning and software engineering.
Printable resources
- Printable cards with the Guidelines: English (PDF) and Chinese (PDF)
- Printable poster with the Guidelines: PDF and jpg
Quick reference
Initially
- Guideline 1: Make clear what the system can do.
Help the user understand what the AI system is capable of doing. - Guideline 2: 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
- Guideline 3: Time services based on context.
Time when to act or interrupt based on the user’s current task and environment. - Guideline 4: Show contextually relevant information.
Display information relevant to the user’s current task and environment. - Guideline 5: Match relevant social norms.
Ensure the experience is delivered in a way that users would expect, given their social and cultural context. - Guideline 6: Mitigate social biases.
Ensure the AI system’s language and behaviors do not reinforce undesirable and unfair stereotypes and biases.
When wrong
- Guideline 7: Support efficient invocation.
Make it easy to invoke or request the AI system’s services when needed. - Guideline 8: Support efficient dismissal.
Make it easy to dismiss or ignore undesired AI system services. - Guideline 9: Support efficient correction.
Make it easy to edit, refine, or recover when the AI system is wrong. - Guideline 10: Scope services when in doubt.
Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about a user’s goals. - Guideline 11: 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
- Guideline 12: Remember recent interactions.
Maintain short-term memory and allow the user to make efficient reference to that memory. - Guideline 13: Learn from user behavior.
Personalize the user’s experience by learning from their actions over time. - Guideline 14: Update and adapt cautiously.
Limit disruptive changes when updating and adapting the AI system’s behavior. - Guideline 15: Encourage granular feedback.
Enable the user to provide feedback indicating their preferences during regular interaction with the AI system. - Guideline 16: Convey the consequences of user actions.
Immediately update or convey how user actions will impact future behaviors of the AI system. - Guideline 17: Provide global controls.
Allow the user to globally customize what the AI system monitors and how it behaves. - Guideline 18: Notify users about changes.
Inform the user when the AI system adds or updates its capabilities.


