In 1968, psychologists John Darley and Bibb Latané conducted a series of experiments that became some of the most famous in social psychology. A participant sat in a room and heard someone in the adjacent room become ill. There was only one key variable: whether the person believed they were the sole witness, or knew that others were hearing what was happening too. The result turned out to be robust and counterintuitive: the more people were presumably observing the situation, the less likely a specific participant was to intervene — and the longer they delayed before taking any action.
There are two reasons for the effect, and in organizations, both operate at full strength. The first is the diffusion of responsibility: when there are many witnesses, the personal sense of "this is my job" weakens because it seems that someone else will react, perhaps someone more competent or more authorized. The second is the social proof of inaction: a person looks around at others, sees that no one is reacting, and concludes that the situation is probably not as serious as it first seemed. Moreover, each of the observers at that moment reaches the exact same conclusion while looking at one another.
In companies, this explains one of the most frustrating categories of crises — those that were predictable. A manager's toxic behavior noticed by dozens of people. A project whose inevitable failure was anticipated by all participants. A departing key employee whose signals were clear for months. Post-factum, it almost always turns out that the information was present in the organization — it was simply never voiced anywhere because each individual assumed that it was either outside their area of responsibility or that someone had already reported it.
The peculiarity of the organizational version of the effect is that structural amplifiers are added to basic psychology: hierarchy makes speaking up risky, blurred zones of responsibility turn "not my job" into a legitimate explanation, and the lack of an explicit channel for signaling makes inaction convenient as well.
What helps:
assigning an explicit owner for areas where responsibility would otherwise diffuse among many — the effect weakens sharply when a person knows the task is specifically theirs;
creating low-risk channels for signaling, including anonymous ones, so that speaking up does not require personal courage;
asking specific individuals directly instead of throwing an open question to the room — "does anyone see any risks?" almost always gathers silence, whereas an addressed question gets an answer;
establishing in the culture that reporting a problem that turns out not to be confirmed is not a mistake — otherwise, the cost of a false alarm remains higher than the cost of silence.