Human calibration determines value in ai assisted decision review

Knowing What You Know: A Trait Theory of Reviewer Value in Human-AI Cognitive Systems

Artificial IntelligenceComputational Engineering, Finance, and ScienceHuman-Computer InteractionMachine Learning

Summary

This paper looks at how humans and AI work together to make decisions, especially who adds value when both review the same item. It finds that the key factor is how well a reviewer’s confidence matches their actual accuracy, called metacognitive calibration. The authors show that just looking at behavior isn’t enough to tell experts from confident novices until correctness outcomes are considered. They also find that humans and the AI model share the same kind of self-awareness about their judgments, so the system should trust whoever is better calibrated for each case. Finally, they explain it’s impossible to identify the exact value of a single human judgment in a group decision context.

What this means in practice

  • For accounting and auditing teams: Optimize task assignment by prioritizing reviewers with proven calibration to focus human effort on likely AI errors in finance checks.
  • For online content moderation teams: Improve reviewer selection strategies by measuring and deferring to well-calibrated moderators rather than relying on agreement rates or activity alone.

Tested on one dataset.

Authors

Priyanka Sinha, Sugam Pandey

Abstract

When a capable model and a human reviewer cooperate on the same decision, which human still adds value? We argue this is a question about a single cognitive trait, metacognitive calibration: the alignment between a reviewer's behavioral confidence and the actual correctness of their judgments. We develop a trait theory of reviewer value to make it precise. The theory makes five claims: value is governed by calibration rather than by activity, speed, agreement, or tenure; calibration leaves observable traces in behavioral telemetry; those traces are non-diagnostic on their own, because the fast, low-revision signature of a calibrated expert is indistinguishable from that of a confident novice until it is gated by outcomes; human calibration and a model's self-knowledge are the same construct, so the joint system should defer to whichever agent is better calibrated on each item; and individual calibration is causally identifiable only above the level of the single judgment. We ground each claim in a six-month production window of an AI-assisted accounting system (541621 reviewer events, 483 reviewers), and find that a purely behavioral expertise profile inverts the ground truth, scoring an overconfident cohort (5.2 percent correction survival) higher than genuine experts (99.2 percent); that reviewer quality is flat across tenure, so the trait is not learned on the job; and that the same calibration logic that separates good from bad reviewers, applied to the model, concentrates 92.5 percent of human effort on genuine errors. We close with the boundary the theory must respect: under majority-vote aggregation, individual causal value is unidentifiable.