Human-AI collaboration is not very collaborative yet: A taxonomy of interaction patterns in AI-assisted decision making from a systematic review (2310.19778v3)

Abstract: Leveraging AI in decision support systems has disproportionately focused on technological advancements, often overlooking the alignment between algorithmic outputs and human expectations. A human-centered perspective attempts to alleviate this concern by designing AI solutions for seamless integration with existing processes. Determining what information AI should provide to aid humans is vital, a concept underscored by explainable AI's efforts to justify AI predictions. However, how the information is presented, e.g., the sequence of recommendations and solicitation of interpretations, is equally crucial as complex interactions may emerge between humans and AI. While empirical studies have evaluated human-AI dynamics across domains, a common vocabulary for human-AI interaction protocols is lacking. To promote more deliberate consideration of interaction designs, we introduce a taxonomy of interaction patterns that delineate various modes of human-AI interactivity. We summarize the results of a systematic review of AI-assisted decision making literature and identify trends and opportunities in existing interactions across application domains from 105 articles. We find that current interactions are dominated by simplistic collaboration paradigms, leading to little support for truly interactive functionality. Our taxonomy offers a tool to understand interactivity with AI in decision-making and foster interaction designs for achieving clear communication, trustworthiness, and collaboration.

• The paper introduces a taxonomy categorizing human-AI interactions from simple recommendations to interactive decision processes.
• The paper systematically reviews existing literature to reveal that basic AI suggestions can limit user engagement and decision efficacy.
• The paper advocates for enhanced explainability and interactive modalities to transform AI tools into active decision partners.

Overview of Interaction Patterns in AI-Assisted Decision Making

The paper "Human-AI collaboration is not very collaborative yet: A taxonomy of interaction patterns in AI-assisted decision making from a systematic review" (2310.19778) extensively explores the dynamics and challenges associated with human-AI interaction, specifically in the context of decision making. As AI systems increasingly integrate with human environments, it is vital to understand and refine interaction patterns to enhance decision accuracy and user satisfaction. This paper provides a foundational taxonomy aimed at standardizing and improving interaction methodologies between humans and AI systems.

Significance of Interaction Pattern Design

Human-AI collaboration promises to leverage AI's computational capabilities to augment human decision-making processes. However, the interaction pattern—defined by how AI's suggestions are presented and integrated—plays a critical role in this partnership. This paper identifies a lack of standard terminology and structured understanding across current literature, hindering effective communication and the advancement of innovative solutions. Explainable AI emerges as a crucial component, providing clarity and insight into AI-driven recommendations. The paper underscores that merely generating AI suggestions is insufficient; the presentation modality substantially influences user acceptance and integration into decision-making processes.

Taxonomy of Human-AI Interaction Patterns

The paper conducts a systematic review of existing literature to derive a taxonomy of interaction patterns within AI-assisted decision-making. The taxonomy categorizes patterns based on the complexity and interactivity levels in human-AI exchanges. Presently, interactions are often simplistic and unidirectional, where AI systems primarily serve as recommendation generators without deeply engaging users. This limits the potential for genuine collaboration and reduces the effectiveness of AI interventions in dynamic decision environments.

The taxonomy seeks to classify interactions as follows:

1. Simple Recommendation Patterns: These include basic advisory roles where AI systems propose solutions without further engagement.
2. Interactive Decision Patterns: These involve more complex exchanges requiring continuous interaction and feedback loops between humans and AI systems.
3. Collaborative Patterns: These promote shared responsibility in decision-making, ensuring that both human and AI insights contribute meaningfully to the outcomes.

Implications and Future Directions

The proposed taxonomy functions as a structured framework for refining human-AI collaborations in decision-making domains. By promoting a shared vocabulary and clear classification, it aims to enhance interdisciplinary communication, fostering innovation in AI system design. As AI's role in decision support evolves, the paper advocates for integrating richer, interactive features within AI interfaces to transform passive assistance into active collaboration. This approach could substantially improve user engagement and decision accuracy, emphasizing interactivity and participatory integration from a human-centered perspective.

Future work may extend this taxonomy to encompass emerging AI technologies and interaction modalities, such as voice-activated systems and immersive environments, to accommodate different domains and contexts. Exploring the intersection of AI with cognitive science and user-centered design principles will likely yield systems that better align with human decision-making processes, increasing acceptance and effectiveness.

Conclusion

The paper provides a critical analysis of the current landscape of human-AI collaboration, presenting a taxonomy to standardize interaction patterns within AI-assisted decision-making. By addressing the existing gaps in terminology and interaction complexity, the taxonomy offers a platform for future research and practical advancements in AI system design. This initiative advocates the integration of interactivity and user engagement as essential elements for enhancing the efficacy of AI in supporting human decisions. Researchers and developers are encouraged to adopt and extend this taxonomy, promoting collaborative innovations that advance AI's beneficial integration into decision-making processes.