Problem Solving Through Human-AI Preference-Based Cooperation

2024-08-14

arXiv.org

AI Summary

Generative AI faces three key bottlenecks in expert domains (e.g., software development): inability to track the state evolution of complex solutions, oversimplified preference expression, and weak dynamic adaptability during human-AI interaction. To address these, this paper proposes HAICo2—a human-preference-driven collaborative construction framework—and formally defines the paradigm of “human-preference-guided AI collaborative construction” for the first time. Methodologically, HAICo2 integrates multidimensional preference modeling, interactive state tracking, interpretable feedback mechanisms, and formalization of the collaborative process. It significantly advances AI’s capability to maintain complex solution artifacts, respond to diverse human preferences, and adapt in real time to evolving user intent. Furthermore, HAICo2 systematically identifies and organizes core open challenges in trustworthy human-AI co-construction, thereby establishing foundational theoretical and methodological groundwork for building reliable, evolvable, expert-level human-AI collaboration systems.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Knowledge Representation and Reasoning: Preferences
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

While there is a widespread belief that artificial general intelligence (AGI) -- or even superhuman AI -- is imminent, complex problems in expert domains are far from being solved. We argue that such problems require human-AI cooperation and that the current state of the art in generative AI is unable to play the role of a reliable partner due to a multitude of shortcomings, including difficulty to keep track of a complex solution artifact (e.g., a software program), limited support for versatile human preference expression and lack of adapting to human preference in an interactive setting. To address these challenges, we propose HAICo2, a novel human-AI co-construction framework. We take first steps towards a formalization of HAICo2 and discuss the difficult open research problems that it faces.

Problem

• Addressing shortcomings in AI for expert domain problem-solving
• Enhancing human-AI cooperation through preference-based interaction
• Developing a framework for collaborative human-AI solution construction

Innovation

• Human-AI co-construction framework HAICo2
• Formalization of interactive preference adaptation
• Versatile human preference expression support