Unpacking Human-AI interactions: From interaction primitives to a design space

Interaction primitives provide a common language for AI design: The paper develops a standardized vocabulary for describing how humans and AI systems communicate, which helps designers understand and compare different AI interaction approaches—from explainable AI to human-in-the-loop systems—in a consistent framework.

Existing AI interactions can be mapped to reusable patterns: By analyzing current practices, the researchers identified common "interaction patterns" that appear across different AI systems, suggesting that many seemingly unique designs actually follow similar underlying communication structures that can be documented and reused.

The framework spans multiple responsible AI approaches: The design space explicitly incorporates key responsible AI concepts including human-in-the-loop processes, explainable AI, collaborative learning, and hybrid intelligence—indicating that thoughtful interaction design is central to implementing these principles in practice.

Design patterns can expand possibilities for human-AI collaboration: By making interaction structures explicit and modular, the framework opens up new possibilities for how humans and AI systems can work together, potentially enabling designers to create more innovative and appropriate interaction models for different contexts and user needs.

Standardization supports better AI system evaluation: A shared vocabulary and set of design patterns allows stakeholders to more easily compare how different AI systems handle human interaction, which is essential for assessing whether systems are truly designed with user needs and responsible practices in mind.