Human-AI teams achieve complementarity only if error correlation ρ_HM < ρ* (with ρ* ≈ a near chance), impossibility above the threshold, gains scaling as Θ(√Δd), and multi-class threshold ρ*_K ≈ ρ*/√(K-1), matching image classification data at R=0.94.
Proceedings of the Annual Meeting of the Cognitive Science Society , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
dataset 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1roles
dataset 1polarities
use dataset 1representative citing papers
citing papers explorer
-
When Can Human-AI Teams Outperform Individuals? Tight Bounds with Impossibility Guarantees
Human-AI teams achieve complementarity only if error correlation ρ_HM < ρ* (with ρ* ≈ a near chance), impossibility above the threshold, gains scaling as Θ(√Δd), and multi-class threshold ρ*_K ≈ ρ*/√(K-1), matching image classification data at R=0.94.