Towards an Argumentative Foundation for Evaluative AI: Towards Explainable and Contestable Decision Support

Summary:This position paper advocates for (computational) argumentation as a foundational paradigm for Evaluative AI (EAI), which supports human decision-making by presenting competing hypotheses along with evidence for and against each. The authors argue that this approach provides a formal, computable basis for explainable and contestable AI systems, paving the way for a long-term research agenda towards distributed and human-centered EAI systems.

---

Core Breakthroughs

• Evaluative AI (EAI) Framework: Proposes a computational argumentation-based framework for EAI, which supports decision-making by presenting competing hypotheses and their evidence rather than single recommendations.
• Explainability and Contestability: The framework emphasizes the explainability of AI systems, allowing users to understand the reasoning process and contest the outcomes.
• Distributed and Human-Centered AI Systems: Lays the theoretical groundwork for future distributed, human-centered EAI systems.

Technical Highlights

1. Argumentation Paradigm: Utilizes computational argumentation as the foundation for EAI, enabling AI to handle complex decision-making scenarios and provide explainable reasoning.
2. Formal Basis: Provides a formal, computable basis for EAI, ensuring the reliability and consistency of AI systems.
3. Long-term Research Agenda: Sets a long-term research agenda for EAI, covering key areas such as distributed systems and human-computer collaboration.

Industry Impact

• Revolutionizing Decision Support Systems: The EAI framework has the potential to revolutionize existing decision support systems, particularly in fields like healthcare, finance, and law.
• AI Ethics and Governance: By enhancing the explainability and contestability of AI systems, the research offers new perspectives on AI ethics and governance.
• Deepening Human-Computer Collaboration: Provides a more reliable technical foundation for future human-computer collaboration, promoting the application of AI in complex tasks.

Developer Recommendations

• Focus on Argumentation Techniques: Developers should explore the application of computational argumentation in AI, investigating its potential in various domains.
• Build Explainable AI Systems: When designing AI systems, consider how to improve their explainability and contestability to enhance user trust.
• Engage in Long-term Research: Encourage developers to participate in the long-term research agenda for EAI, contributing to the advancement of AI technology.

Conclusion

This paper provides a new paradigm for Evaluative AI, emphasizing the importance of explainability and contestability, and points the way for the future development of AI systems.

Source: arXiv:2608.07473

— END —

Tags: #Evaluative AI #Computational Argumentation #Human-Computer Collaboration