PAIR AI Design Principles

Google's People + AI Research (PAIR) design principles for creating human-centered AI products that balance user autonomy, safety, and helpfulness while adapting to user feedback and real-world contexts.

  1. User Autonomy

    Design for the appropriate level of user autonomy in AI-supported workflows, considering different user tasks, expertise, and the effort required to steer the AI system.

  2. Data & Model Alignment

    Align datasets and models with people’s real-world interactions with AI systems, accounting for unexpected tasks and contexts that may emerge as the product scales.

  3. Evolving Safety

    Treat safety as an evolving endeavor with multifaceted strategies that adapt to changing technology, user contexts, and complexity over time.

  4. Adapt with Feedback

    Create bidirectional feedback loops where AI learns from users to personalize experiences, and users adapt their behaviors in response to AI outcomes.

  5. Helpful AI

    Create AI experiences that enhance aspects of work and play that people enjoy, focusing on seamless integration and inspiration rather than just efficiency gains.

Edit this page

Know of a set of design principles that should be here? Contribute an example