Human-AI Collaboration
More information
Ross Dawson recomments 5 research papers:
⭐ Human diversity fuels collective creativity that large language models cannot simulate or sustain - Mengchen Dong, Ph.D. et al
A study shows that AI-simulated diversity always remains below the contributory diversity of human groups. "Human diversity remains a valuable creative resource that current AI cannot simulate or sustain; the design of human-AI collaborative workflows determines whether it survives."
⭐ The tragedy of the cognitive commons: collective intelligence beyond AI-induced knowledge collapse - maher kallel et al
Acemoglu et al recently described the conditions for "cognitive collapse". This paper identifies five structual critiques to frame this as a classic systems problem, the tragedy of the cognitive commons. Understanding this allows us to identify how to address it, by improved aggregation of validated human knowledge.
⭐ Complementarity in Human–AI Collaboration: Concept, Sources, and Evidence - Patrick Hemmer et al
To achieve the complementary potential of Humans + AI we need to distinguish between information asymmetry, where each possess different information, and capability asymmetry, where they process information differently. Empirically, the unique context available to humans provides superior performance.
⭐ Generative Artificial Intelligence in Scientific Research: Individual Benefits, Collective Risks, and a Framework for Responsible Research with AI - Fulvio Castellacci et al
There is a gap between private and social returns in science, due to information asymmetry, negative externalities on a shared knowledge base, and depletion of research capacity. The authors propose a governance framework Responsible Research with AI (RRAI), based on four principles: disclosure, differentiation, narrative, and proportionality.
⭐ Robust Human-AI Complementarity under Uncertainty - Yewon Byun et al
The biggest barrier to effective human–AI collaboration is often not AI quality itself, but uncertainty about when to trust it. Humans and AI can be complementary when they make different or negatively correlated errors, due to different information or reasoning processes. This can occur in complex or novel domains where humans may have context not available to systems.