S3ntiment

Almost every organisation knows the pattern. Someone resigns, and in the exit interview it all comes out at once: the manager, the workload, the decision nobody supported and nobody said anything about. Not because that person didn't know it earlier. The employee survey three months before came back comfortably positive.

Which is confusing, because a survey looks like exactly the moment where that shouldn't happen. You fill it in alone, on your own screen, with nobody reading over your shoulder. No meeting, no manager, no room full of colleagues. If there's anywhere someone can write down what they actually think, surely it's there.

In practice it works differently. Social pressure doesn't sit in the room you're in; it sits in the question of what happens to your answer. Who reads this, how small is my department, will my score stand out if I'm the only one marking things low, would anyone recognise the way I write. People make that assessment in about two seconds and usually without putting it into words, and the result is an answer a little more careful than the thought behind it.

So the most honest data only becomes available once someone has nothing left to lose. That isn't chance and it isn't an incident, and it means every survey before it measured something other than what it thought it was measuring. Not the opinion, but the judgement of what was safe to write down.

This isn't a communication problem

The usual response is a culture programme. More psychological safety, a more open conversation, a manager who listens better. That helps, but it doesn't solve this, because the problem isn't people who know something and don't dare to say it.

It's people who don't know they aren't the only one. A hundred employees who privately think the same thing, each assuming they're the exception. Nobody suppressed anything. No messenger was shouted down. There was simply never anyone in a position to see that those hundred opinions added up to a majority.

That isn't a communication problem. It's a measurement problem.

Five disciplines, one conclusion

The phenomenon has been described from different angles for nearly a century, each time by researchers not building on one another's work.

In 1931 Daniel Katz and Floyd Allport documented what they called pluralistic ignorance: almost every member of a group privately disagrees with the assumed group norm, yet goes along with it, believing themselves to be the only dissenter. Nobody believes it, but everybody thinks everybody else does. The organisational consequence is that policies stay in place long after support for them has gone.

The sociologist Erving Goffman described in 1959 how people adjust their behaviour to whoever is watching, distinguishing front stage from back stage. What an organisation gets to see is the front stage, and a survey is no exception to that.

The communication scientist Elisabeth Noelle-Neumann published her theory of the spiral of silence in 1974. People have a largely unconscious fear of social isolation and continuously monitor their surroundings: who says what, who gets agreement, who gets ignored. Anyone who senses their opinion deviates keeps quiet. It becomes a spiral because silent minorities are invisible: the more people stay quiet, the stronger the apparent consensus, and the more others follow.

The political anthropologist James Scott added a sharper distinction in 1990. What people say in the presence of power he called the public transcript, what they say among themselves the hidden transcript. Both are sincere. But only the second contains what people really think about how things are going, and organisations by definition have no access to it.

Morrison and Milliken brought this together in 2000 under the term organizational silence, showing how it leads to decisions almost nobody internally supports but which go ahead anyway, simply because silence gets read as agreement.

Five traditions, one conclusion: people in groups don't say what they think, and not because they're dishonest. What looks like a mandate may be nothing of the kind.

Digital made it worse

When social media arrived, the expectation was that it would break this open: a lower threshold, a bigger audience, more voices. The Pew Research Center studied that in 2014 among 1,801 Americans, using the Snowden revelations as the topic, an issue on which public opinion was reasonably divided. People turned out to be more reticent online than offline. Social media users were even less inclined to share their views in ordinary conversation than people who weren't on those platforms at all.

The explanation lies in the architecture. Digital platforms are built on observation: every interaction is tracked, analysed, fed back. People sense that and adjust. A system built on visibility produces visibility data.

That applies to any survey system where an employee sees even the possibility that an answer might be traced back, to their manager, to HR, or to the vendor of the tool.

The difference between promise and proof

Most organisations answer this with an assurance. Your answers are anonymous. Results are only shared in aggregate. Nobody sees individual responses.

That doesn't work, and it has been demonstrated. Stanley Warner developed the randomized response technique in 1965: respondents were given either a sensitive or a neutral question at random, so the researcher saw the answer but not which question it belonged to. Reported rates of sensitive behaviour came out structurally and substantially higher than in ordinary anonymous surveys. Not slightly higher, but of a different order.

Ben Jann confirmed this in organisational settings: the difference between promised and proven anonymity has a demonstrable effect on the quality of the data. People know the difference and act on it.

A promise of privacy measures what people say. Enforceable privacy measures what people think.

Why it has to be architecture

The spiral of silence needs a perceived audience. As long as an employee sees the possibility that someone is watching, the mechanism engages and the front stage appears. That mechanism doesn't respond to reassurance, because the reassurance comes from the same party that might be watching.

Which is why better questions don't help, a friendlier tone doesn't help, and a culture programme only goes so far: the measurement is itself part of the social context shaping the answer.

What does work is a system in which looking is technically impossible, including for the organisation running the survey. Not as policy, not as a promise. Only then does the perceived audience disappear, and only then does the spiral have nothing to feed on.

Sources

Katz & Allport, Students' Attitudes (1931) Goffman, The Presentation of Self in Everyday Life (1959) Warner, Randomized Response (1965) Noelle-Neumann, The Spiral of Silence (1974) Scott, Domination and the Arts of Resistance (1990) Morrison & Milliken, Organizational Silence (2000) Pew Research Center, Social Media and the Spiral of Silence (2014)