The Second and Third Expert Agreed
- Admitted
- 2026-09-07
- Source dataset
- Who&When
- Failure mode
- Coordination failure
- Agent system
- Multi-agent
- Tool access
- Read-only
- Industry
- General or other
Presenting Behaviour
A three-agent team was asked which of two Hawaii homes sold for more in 2022. A "real estate expert" agent supplied sale prices for both properties. A "data analysis" agent and a "validation" agent each reviewed and confirmed the higher of the two figures without independently pulling the actual sale record. The confirmed price was $50,000 higher than the true recorded sale price.
Case Replay
A reconstruction, not a simulation. This sequence is reconstructed from the public incident record and summarised in our own words. It is a static record of what was reported, and nothing here can be run or replayed live.
4 steps
- Price supplied. A real estate agent gives sale prices for two homes. This is where the agent derailed.
- Analysis agent agrees. It confirms the higher figure without pulling the record.
- Validation agent agrees. It confirms it too.
- Wrong answer. The confirmed price is $50,000 above the recorded sale.
Differential Diagnosis
An independent error by each reviewing agent, considered and ruled out; both reviewers' outputs were internally consistent with the real-estate agent's original figure, not with each other's separate mistakes, which points to one shared error being passed along rather than two coincidental ones.
The original figure was simply the best available estimate, considered. But the role of the downstream agents was specifically to validate and analyze, and both signed off without any independent check against a source, which is different from a defensible estimate under real uncertainty.
Shared Psychotic Disorder, most consistent with the record: one agent's unverified figure was adopted and re-confirmed by two downstream agents in succession, neither of which broke the chain by checking it independently.
Working Diagnosis
Shared Psychotic Disorder (folie à deux): an unverified figure from one agent was passed to and endorsed by two downstream reviewing agents in a row, with no independent verification breaking the chain.
Discussion
Naming an agent "Validation Expert" doesn't make its validation independent, if it validates by checking consistency with an upstream agent's claim rather than against an external source, it isn't really validating anything. A workflow that wants real error-catching needs at least one step that goes back to the original source, not just agreement between agents reviewing each other's work.
Mitigation Strategy
Include at least one step that goes back to the original source rather than agreeing with upstream agents.
Governance Guardrails
Independent source verification; validators that cannot rely on upstream claims.
Typical Presentation in Human Patients
General clinical background, included for context. Not medical advice, and not a diagnosis of any individual.
In shared psychotic disorder, a person in a close relationship with someone experiencing delusions comes to adopt the same false belief themselves, having had little or no independent reason to hold it. The condition typically resolves once the two people are separated.
Source Note
Who&When, ag2ai/Agents_Failure_Attribution
https://raw.githubusercontent.com/ag2ai/Agents_Failure_Attribution/main/Who%26When/Algorithm-Generated/4.json