Repeated warnings of crises that never arrive teach an organization to shrug, so when a real threat comes through the same channel, nobody wants to look gullible by acting on it.
An example
For four years the treasury group at a regional water company had warned the executive team about counterparty exposure on its interest-rate hedges. Each warning came as a red-flag memo predicting losses, and each time the losses failed to appear. By the fifth memo, the chief financial officer opened the meeting by saying that last time had been nothing. So when the treasurer's team flagged that the bank on the other side of most of the hedges had been downgraded twice in a month, the warning was noted and filed. The treasurer changed the route. She asked the company's outside auditor to review the exposure, and gave the executive team a probability range, a stated loss threshold, and a list of her group's past warnings with what had happened to each. That week the committee approved moving half the hedges to a second bank.
Why it works
Each unrealized warning makes doing nothing look easier, because the sender's record supplies the cover story. Mobilizing again looks worse, since past alarms produced only meeting hours, and costlier, because anyone who acts risks looking gullible. Fatigue builds faster when warnings are vivid, frequent, and binary, when no one tracks base rates, and when the sender's standing gets confused with the signal's accuracy. It eases when the sender states uncertainty honestly, an outside event re-anchors the channel, or someone else carries the next warning.
The thinking behind it
George A. Akerlof Akerlof's The Market for "Lemons": Quality Uncertainty and the Market Mechanism (1970) shows that when buyers cannot tell good goods from bad, they pay only for average quality. An executive team that cannot tell a calibrated warning from another false alarm discounts every warning from the channel. Probabilities, thresholds, and a published track record make the quality of a warning visible, so the good one can be told apart.
Carl I. Hovland Hovland's Communication and Persuasion: Psychological Studies of Opinion Change (1953) reports that the same message is judged differently depending on how expert and trustworthy its deliverer seems. A sender with a record of unrealized alarms has lost trustworthiness, and the content suffers with it. Routing the warning through an auditor or another channel whose credibility is intact lets the message be heard on its merits.
Daniel Kahneman Tversky and Kahneman's Judgment under Uncertainty: Heuristics and Biases (1974) describes availability, which makes the examples nearest to hand the measure of probability. Recent false alarms are the instances that come to mind when a new warning arrives, so the threat is judged by them. An outside signal, such as a downgrade or a regulator's notice, supplies a fresh instance pointing the other way.
Warning signs
Warnings arrive as binary predictions, with no probabilities or thresholds. Nobody tracks the sender's hit rate across cycles. Each new alert meets the reply that last time was nothing. Senior people who propose acting are teased for falling for it again. The underlying risk has not changed, but attention to it has dropped with every warning that came to nothing.
How to defend yourself
Separate the signal from the sender. The base rate of the threat did not change because earlier warnings were wrong. If you suspect someone is issuing alarms to wear out a channel, ask that every warning carry a probability and a review date. The record then becomes visible to everyone, including whoever has been crying wolf.
- Route a serious warning through a different sender or channel whose credibility is intact.
- Anchor it to an outside signal: a regulator, a customer, a market event, an auditor.
- If you are the sender, state probabilities and thresholds, publish your track record, and stand down warnings publicly when they pass.
How a leader shuts it down
Ask warning channels to state probabilities and thresholds, and keep a shared record of predictions and outcomes. Judge a channel's accuracy apart from the standing of the person who runs it. Keep a second channel, such as internal audit or an outside adviser, for warnings the first one has worn out.