Skip to content

PAN Lab levers

Lever

Train verification

The organization trains its staff to check the AI system's output against the source. They learn to look for errors in the reasoning, not only in the wording. Checking becomes a skill people are taught instead of one they are assumed to have.

What it is

People catch the errors they know how to see. The randomized study cited below found that people caught surface errors far more often than errors that needed conceptual judgment. How hard they checked, how much they trusted AI to begin with, and how visible the error was made the difference. Payment and time spent did not. Verification training teaches the harder half: tracing a claim back to its source and asking whether the reasoning holds.

What it pushes on in the Lab

In the Lab, this lever raises the checking and correcting that the people using the system can do. It also holds down operator deference drift, because people defer less to output they can check themselves.

increasedPeople or agents using it
cappedOperator deference drift

You can also pull this lever at a strong tier, which costs more. At the strong tier, each effect below the Lab's strongest setting pushes harder.

In the modes that offer aiming, you can aim this lever at particular parts and pathways of a network. Otherwise it applies to the whole network.

Its pattern in the Practice Library

The Practice Library describes the pattern behind this lever:Verification training

The pressures it answers

No pressure page lists this lever among the levers that answer it.

Where you can pull it

No example network in the Lab offers this lever. Every network the Lab builds at random offers it.

The evidence behind its effects

The Lab cites these claims from the evidence registry for this lever's effects.

In a randomized study (N=2,784) with objective ground truth, humans accepted incorrect AI suggestions about a third of the time, and their rate of catching AI errors was governed by verification effort, prior trust in AI, and error legibility — surface errors were caught ~82% of the time versus ~31% for errors requiring conceptual judgment — not by financial incentives or time spent.[†]

beck2026AcademicSave

Beck, J., Eckman, S., Kern, C., & Kreuter, F. (2026). Bias in the Loop: How Humans Evaluate AI-Generated Suggestions. Harvard Data Science Review, 8(2). https://hdsr.mitpress.mit.edu/pub/nrcn4h7d/release/1

https://hdsr.mitpress.mit.edu/pub/nrcn4h7d/release/1

Appears in: PAN framework development

Topics: algorithmic-fairness, human-ai-interaction

A validated collaborative-AI metacognition scale (planning, monitoring, evaluation of one's own reliance) predicted collaboration benefits incrementally beyond general metacognition — verification-skill training, not generic AI knowledge, is the calibrated counter to over-reliance.[2]

sidra2025AcademicSave

Sidra, S., & Mason, C. (2026). Generative AI in Human-AI Collaboration: Validation of the Collaborative AI Literacy and Collaborative AI Metacognition Scales for Effective Use. International Journal of Human–Computer Interaction, 42(7), 5084–5108. https://doi.org/10.1080/10447318.2025.2543997

doi.org/10.1080/10447318.2025.2543997

Appears in: Evidence reverification (2026); Paramerge authored research

Topics: ai-safety, human-ai-interaction

bucinca2021AcademicSave

Bucinca, Z., Malaya, M. B., & Gajos, K. Z. (2021). To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW1), Article 188. https://doi.org/10.1145/3449287

doi.org/10.1145/3449287

Appears in: Evidence reverification (2026)

Topics: human-ai-interaction

Pull this lever in the PAN Lab and watch which way it pushes the network.

Open the PAN Lab