Researchers Propose a New Model for Human and AI Decision-Making Collaboration

Artificial intelligence is increasingly being harnessed to support doctors, lawyers, and other experts, but collaboration often does not make the team better than the best individual human. A new article argues that the problem is not only in the imperfect accuracy of AI but also in the way human-AI collaboration is generally understood.

Current large language models are typically integrated into systems that provide advice or alternative solutions. In practice, however, experts find themselves balancing between two extremes: endless loops of verification and blind trust. According to researchers, the promised complementarity—where humans and AI would compensate for each other's weaknesses—does not materialize this way.

The authors propose a solution called Collaborative Causal Sensemaking. The idea is to design decision support systems specifically as thinking partners, not just answer machines. Such systems would participate in the mental work of decision-making: they would maintain and share with humans continuously evolving images of the situation's cause-and-effect relationships, goals, and constraints.

The article suggests that collaboration between humans and AI should be seen as a joint cognitive process. In it, both humans and AI build, test, and correct a shared understanding of the changing situation. The goal is for AI not just to provide ready-made answers but to help the expert understand why a solution is justified and what consequences it might have.

The research primarily serves as an opener for a research program and framework: it invites the design of decision-making AI agents from the ground up as collaborators who share the responsibility of cognitive work with humans.

Source: Collaborative Causal Sensemaking: Closing the Complementarity Gap in Human-AI Decision Support, ArXiv (AI).