What do you believe — and what

are you willing to tell me?

Surveys do not provide direct access to what people think. They collect answers produced by receivers responding to particular questions, asked in particular ways, under particular conditions. The answer may reflect what a person privately believes. But it may also reflect what they think the question means, what they remember, what they believe is socially acceptable, how they expect their answer will be used—and whether they feel safe revealing it. The Noticing Loop offers a way to examine what happens between holding an opinion and reporting one.

An answer is a signal

When someone answers a question, they transmit a signal from which the questioner tries to construct a representation. But the answer is not the belief itself. A respondent must first: • notice and interpret the question; • decide what information is relevant; • construct or retrieve an answer; • judge whether that answer is good enough; • decide what they are willing to reveal; and • translate their representation into one of the available response options. The survey designer then interprets, categorises, combines and reports those answers. A published result is therefore not a direct picture of public opinion. It is a representation constructed through a chain of receivers, judgments and transformations.

Three different representations

A person may hold several related but different representations: 1. What I believe: My private opinion, preference, intention or judgment. 2. What I think other people believe: My representation of the social normal around me. 3. What I am willing to tell you: The representation I choose to transmit in this situation. These may overlap, but they need not be identical. If I believe my answer could expose me to criticism, embarrassment or exclusion, I may modify it, soften it, avoid it or refuse to answer. This does not always require deliberate deception. The anticipated response of the questioner may influence what becomes available to say before I formulate my answer. What is transmitted may therefore depend partly on the expected cost of transmitting it.

“How will you vote?”

Consider two polling questions: How will you vote? and: How do you think your neighbours will vote? The first asks the respondent to report a personal intention. The answer may feel revealing. It may identify the respondent with a candidate, party, group or set of values—and create a risk of being judged. The second question appears to be about other people. It reduces the respondent’s personal exposure and asks them to construct a representation of the surrounding social normal. That representation may contain useful information gathered from conversations, signs, behaviour, local experience and weak signals that the respondent cannot fully identify or explain. But it may also contain: • projection of the respondent’s own preference; • assumptions about people they barely know; • stereotypes; • impressions shaped by unusual or memorable encounters; • media representations of public opinion; and • beliefs about what other people are willing to admit. The second question does not reveal the same thing more accurately. It produces a different judgment. That difference may nevertheless be useful.

The question changes the task

A question is not a neutral container into which a pre-existing answer is placed. Different wording may change: • what the receiver notices; • which memories become available; • the comparison being made; • the unit or scale used for judgment; • the social consequences that become salient; • the threshold for selecting an answer; and • the representation ultimately transmitted. Ask whether someone is “concerned”, “worried” or “afraid”, and the receiver may construct three different judgments. Ask whether a service is “good”, and different respondents may compare it with different standards: what they expected, what they experienced previously, What they paid for or what they believe was realistically possible. The answer cannot be fully understood without considering the question, the context and the receiver’s model of normal.

Available answers shape available

opinions

Surveys commonly ask respondents to fit their judgment into a predefined set of categories: • agree or disagree; • satisfied or dissatisfied; • likely or unlikely; • approve or disapprove; • select one candidate; • rate something from one to ten. These options make answers easier to collect and compare. But they also define the units through which the response can be expressed. A receiver may hold a conditional, uncertain or internally divided representation that does not fit any available category. Selecting the nearest option can make the final answer appear clearer and more settled than the underlying judgment was. A precise percentage can therefore be produced from answers that were individually approximate, reluctant or interpreted in different ways. Measurement does not merely record a representation. The available units help shape the representation that can be recorded.

Silence is also produced

Some receivers do not answer. They may not have noticed the survey, understood the question, formed an opinion, trusted the questioner or considered participation worth their limited time and attention. Others may have an answer but judge that revealing it carries too much risk. Non-response is therefore not empty space. It is an outcome produced somewhere within the interaction between question, receiver and context. However, its meaning cannot be assumed. Silence may indicate uncertainty, indifference, mistrust, fear, lack of access, lack of time—or something else entirely. The people who answer may differ systematically from those who do not. A survey can represent only the signals it receives, but the missing signals may matter.

Anonymity helps - but does not

remove the receiver

An anonymous survey may reduce the expected cost of answering honestly. It does not remove every influence on the response. The receiver may still: • judge themselves through the imagined eyes of others; • avoid acknowledging an uncomfortable belief; • infer what the survey is trying to discover; • provide the answer they believe a reasonable person should give; • remember selectively; • misunderstand the question; or • construct an opinion only when asked for one. A person can answer sincerely without the answer being a stable or complete representation of what they believe. Honesty does not eliminate uncertainty.

From individual answers to “public

opinion”

Survey findings pass through several stages before they reach the public. Questions are selected and worded. Participants are reached. Answers are coded. Missing responses are handled. Samples may be weighted. Results are grouped, summarised and presented. At each stage, someone decides which differences matter and which can be treated as equivalent. Two answers may be placed in the same category. A small change may be described as important or insignificant. Uncertainty may be made prominent or allowed to disappear behind a headline figure. “Public opinion” is therefore a useful representation—not a single object waiting to be discovered. To interpret it well, we should ask: • Who was able and willing to respond? • What exactly were they asked? • What did the available answers allow them to express? • What comparison or model of normal did they use? • What may have influenced what they were willing to transmit? • How were different responses combined? • Which uncertainties were preserved, and which were removed?

Beyond political polling

The same pattern appears wherever people are asked to report what they think or experience. Organisations: An employee may privately identify a serious problem but report that everything is satisfactory if criticism appears unsafe or futile. Safety: A worker may notice that something is not quite right but hesitate to report it because the signal is difficult to explain, appears too weak or could expose them to blame. Education: A learner may say they understand because they believe that is the expected answer—or because they cannot yet identify what they have misunderstood. Healthcare: A patient may understate pain, symptoms or behaviour because they fear judgment, do not share the clinician’s scale or have adapted to a condition that now feels normal. Customer research: A customer may select “satisfied” while quietly deciding never to return. The survey answer and the later action may draw on different judgments and thresholds. In each case, asking a question produces a signal. Understanding that signal requires attention to the receiver and the conditions under which it was produced.

Better questions - or several

different questions?

There may be no single perfect question that reveals a person’s “true” opinion. A more useful approach may be to ask different questions that approach the representation from different positions: • What do you believe? • How confident are you? • What might change your mind? • What do you think most other people believe? • What do you expect them to do? • What makes it difficult to answer? • Is your answer different in private and in public? • Which available option comes closest, and what does it leave out? Differences between the answers are not necessarily errors to eliminate. They may reveal the structure, uncertainty and social conditions surrounding the judgment.

What NL adds

The Noticing Loop encourages us to look upstream of the recorded answer. It asks: • What became available to the receiver? • Against what model of normal was it judged? • What difference attracted attention? • What resources were available to investigate it? • What threshold made one answer good enough? • What consequences did the receiver anticipate? • What representation was eventually transmitted? • What was lost when that representation was categorised and combined with others? Polling does not simply measure opinion. It creates a communication situation in which the respondent must decide what the question means, what answer is defensible and what can safely be revealed. The resulting answer may tell us something important—but we should be clear about what it represents.

Questions to explore

1. When does a reported opinion differ from a private one? 2. Can indirect questions reveal information that direct questions suppress? 3. How does anticipated judgment change what people transmit? 4. What can disagreement between different polling methods teach us? 5. How should uncertainty and conditional answers be represented? 6. What might non-response tell us—and what must we avoid assuming? 7. Can surveys measure changes in what people notice, not only what they say? 8. How can organisations make weak or uncomfortable signals safer to report? 9. When does aggregation clarify a pattern, and when does it conceal meaningful difference? A survey answer is not simply extracted from a person. It is constructed by a receiver, under particular conditions, for a particular audience — and then interpreted by another receiver.