When is a behaviour problem really an association problem?

The absence of an action is an observation, not a diagnosis. These hypothetical examples show why the same dashboard outcome can call for very different interventions.

A strong response in an association task does not establish that associations caused the missing behaviour. Without comparable evidence about ability, opportunity and habit, a convincing psychological story remains a story.

Michie, van Stralen and West developed COM-B as the central system of the Behaviour Change Wheel. Behaviour depends on interacting conditions: capability, opportunity and motivation. Capability includes physical and psychological ability. Opportunity includes physical and social circumstances. Motivation includes deliberate reasoning as well as automatic processes. [S01]

That last distinction matters for association research. Associations may contribute to motivation, but they are not the whole of motivation. Someone may value a new safety routine while repeatedly falling into an old habit. A citizen may approve of a sustainable option while continuing to choose the familiar route because it is easier to access. A team member may disagree with a change because its practical consequences have never been explained, rather than because of a hidden negative attitude.

The Theoretical Domains Framework offers a more detailed diagnostic vocabulary: knowledge, skills, social role and identity, beliefs about consequences, emotion, environmental resources and social influence are distinct domains. These distinctions prevent 'resistance' from becoming a convenient label for everything that an organisation does not yet understand. [S02]

An association connects an action, organisation, object or situation with a particular meaning. That meaning may come to mind readily without being actively considered at every moment. For one team, digital reporting suggests 'efficiency'. For another, it suggests 'bureaucracy'. Elsewhere it might represent 'accountability' or 'professional pride'. The outward instruction is identical, yet the learned context differs.

But an association is not automatically a causal mechanism. An interviewee who says 'bureaucracy' may be describing a genuine obstacle, or rationalising an action that was mainly prevented by limited time. Likewise, an indirect association measure cannot independently reveal why someone behaved as they did. Both interpretation and measurement context matter. [S06] [S08]

A useful study therefore separates accessible meaning, the circumstances that activate it, and the subsequent observed action. Connect these layers through a defensible research design. Do not infer an individual employee's hidden motives from a single word or response-time score.

After a new digital workflow is introduced, a team leader hears remarks such as 'another form', 'they do not trust us' and 'now every minute has to be justified'. Those comments deserve attention. But they do not show how common the interpretation is, how strongly it influences action or whether the most outspoken colleague represents the whole group.

Ask where the interpretation came from. Did the organisation recently introduce new monitoring? Are people actually judged on registration data? Was a tool initially offered to support professional judgement and later repurposed for individual control? A negative association can be grounded in an accurate memory of organisational behaviour. If so, the remedy may be to change that behaviour rather than to polish the wording.

This is why association research can be uncomfortable for an organisation. A message about empowerment will not be credible if the policy removes genuine autonomy. When a negative meaning is justified by experience, the experience itself must change. A new communication campaign cannot substitute for governance, fairness and trust.

'We want more engagement' is too vague to measure. 'Before leaving for the next site, a field employee performs a safety check and records any hazards' specifies an action. Record who, does what, where, when, and under which relevant conditions. State why this behaviour matters and distinguish the action from the attitude you hope people will have towards it.

Also distinguish first-time compliance from a routine that survives ordinary work pressure. Completing the check during a training workshop is not the same as remembering it after a difficult shift. Document exceptions: locations without coverage, equipment failures, situations in which the check is not applicable and occasions on which an employee has a legitimate reason not to proceed.

Agree on a behavioural outcome before the intervention, such as the share of eligible work moments with a complete and accurate check. Establish baseline data and assess reliability. Avoid inventing Neurofactor effect sizes. Make sure reporting arrangements cannot become an unfair method for ranking or disciplining individual workers. [S01] [S12]

First ask whether a person could carry out the action if they wanted to. Watch the task in its natural setting. Are instructions understandable? Can the worker reach the device? Is there enough time? Is the task physically safe? Are other targets rewarding the opposite behaviour, such as speed at the expense of an essential safety step?

This sounds like operational analysis rather than neuroscience. That is precisely the point. Taking an association hypothesis seriously requires ruling out simpler explanations. A worker who cannot access a record during the shift has not demonstrated a lack of motivation for that action. A resident unable to navigate an inaccessible local-government website has not demonstrated an unfavourable attitude towards participation.

Record barriers without making a judgement about the character of the people involved. Remove those that can be removed directly. Only then ask whether a remaining meaning barrier deserves further investigation. This sequence protects employees from being psychologised when the system is actually failing them.

Suppose an action is regularly missed. That frequency alone does not tell you why. Create alternative explanations that could each be weakened by specific evidence. The following is a hypothetical diagnostic framework, not data from a Neurofactor project.

Treat every row as a candidate explanation rather than a box you must fill. One individual may face several barriers at once. A team may change from one mechanism to another over time. A study becomes more credible when the researchers have documented what might prove them wrong.

'Which negative associations stop our staff?' is a leading research question: it assumes associations are the obstacle. A more useful question is: 'What combination of skills, opportunity, habits, social expectations and meanings explains the missing behaviour?' The association hypothesis becomes one option among several.

Specify in advance which findings would count against it. If lack of time predicts non-compliance and shortening the task produces a consistent improvement, motivational messaging may be unnecessary. If resources and skills are adequate, while a particular negative meaning is consistently related to behaviour in comparable situations, a targeted test of meaning becomes more interesting.

Even then, association does not imply causation. A worker may dislike the system because they have repeatedly failed with it, rather than fail because they dislike it. Another unmeasured factor may explain both responses. Interpretive confidence should match the design: descriptive comparisons, longitudinal observations and controlled interventions offer different levels of evidence. [S03] [S12]

Employees with the same job title can face different barriers. A seasoned worker may know the check but interpret it as distrust. A newcomer may want to comply but find the instructions confusing. Someone on the late shift may have no functioning connection. Calling them all a 'resistant segment' turns different problems into a convenient marketing category.

Useful segmentation starts from behaviour, conditions and carefully examined meanings. Age, education or a personality label should not be treated as a substitute for evidence. Relevant groups may differ by work location, task, exposure to a system, length of experience or team routines, but those distinctions should be tested rather than assumed.

Keep segments provisional. Re-examine whether they remain useful after the environment changes. Never turn an indirect association measure into a permanent individual profile or a hiring decision. Particularly in employment and public services, the purpose of diagnosing behaviour is to improve the system, not to classify human beings as psychologically cooperative or uncooperative.

A good interview explores a real recent occasion. What happened immediately beforehand? What did the person see? What did they expect? What was difficult? What action did they take? Leave room for enthusiasm, indifference, frustration and practical obstacles. Asking 'Did the monitoring make you feel mistrusted?' places the interpretation in the respondent's mouth before you have evidence for it.

Then compare interviews with task observation, process data and, where justified, a carefully validated association measure. Explicit answers may be influenced by social desirability, but indirect measures have their own limitations. Fazio and Olson discuss the context sensitivity and construct interpretation of such measures. Hofmann and colleagues examined how relationships between explicit and implicit responses depend on several factors. [S06] [S08]

A set of quotes is not a behavioural experiment. A response-time graph is not an explanation of organisational practices. You gain insight when independent evidence sources illuminate one another and when contradictions are openly investigated.

This case is entirely fictional. It includes no measurements, results or confidential information from any Neurofactor client. An imaginary service organisation introduces a thirty-second digital safety check for mobile teams. Its legitimate aim is to identify risks early. Completion remains patchy after launch, and management assumes staff simply do not care about safety.

Observation points to two different situations. On some sites, connectivity fails and people must enter the same details again after their shift. In other teams, the app works, but the check is described over coffee as 'another way for head office to keep an eye on us'. Managers have been displaying individual completion lists in team meetings, strengthening precisely that association.

The plausible diagnosis is mixed: an opportunity barrier alongside a possible association barrier grounded partly in organisational conduct. Another cheerful campaign will not automatically solve either problem. First make the process functional and agree on legitimate, transparent use of the data. Then investigate whether the meaning changes when the actual experience becomes more trustworthy.

The fictional case illustrates why a behavioural diagnosis must lead to distinct, testable responses. Every row below describes a hypothetical possibility, not a reported intervention result.

Several responses may be needed together. If feasible, introduce changes in stages or use a design that distinguishes their effects. That way the organisation learns more than 'the project improved the dashboard'. It learns which mechanism was plausibly changed and which barrier remains.

Step 1 - Specify the action. Describe a behaviour that can be observed at a relevant moment. Step 2 - Map the work system. Check time, access, understanding, equipment, incentives and social arrangements. Step 3 - Establish the baseline. Measure what happens and where without assigning a psychological reason.

Step 4 - Investigate meaning. Use open, context-specific questions and, only if suitable, indirect association tasks. Step 5 - Keep alternative explanations alive. Describe expected evidence for capability, opportunity, habit, norm and association barriers. Step 6 - Design discriminating interventions. Where possible, vary a single mechanism or use a design that allows effects to be separated.

Step 7 - Measure behaviour and unintended consequences. Track the agreed outcome under ordinary working conditions. Step 8 - Decide whether to scale. Examine feasibility, acceptability, effectiveness and maintenance. These eight steps are a practical application of established diagnostic principles, not a newly validated Neurofactor clinical or psychological scale. [S01] [S02] [S12]

EEG can be a valuable complement in particular settings, but there is no reason to begin with it if a simpler method answers the key question. The point is to remove uncertainty, not to make the research look technologically impressive.

A sophisticated toolbox is only useful when the study specifies what each tool contributes. Ask which uncertainty remains after practical barriers have been documented. [S06] [S07] [S12]

'Record every hazard' might sound like professional responsibility when a trusted colleague says it. The same words from a manager who previously punished every reported mistake may sound like another monitoring order. The instruction has not changed; its social context has.

Investigate more than the message. What happened during earlier change projects? What is rewarded in practice? Can staff challenge an instruction without being penalised? If the organisation talks about safety while praising only speed, its reward structure belongs in the diagnosis. An association intervention may therefore require a management decision or a fairer system of accountability, not another creative treatment.

Group norms are relevant too. Some colleagues may fear being seen as 'a person from head office' if they adopt the new routine. Do not assume such a norm exists from one colourful quote. Examine whether it actually shapes behaviour and under which conditions. Choosing a charismatic ambassador without changing the underlying organisational reality may be insufficient. [S02] [S10]

Different diagnoses require different intervention functions. The Behaviour Change Wheel distinguishes options such as education, training, environmental restructuring and enablement. The Behaviour Change Technique Taxonomy helps name the active ingredients in a replicable way. [S01] [S11]

Treat the final column seriously. Any intervention can create unintended effects. The purpose is not to force one technique into every scenario; it is to create a defensible match between what is happening and what the organisation changes.

Suppose technical barriers in our imaginary case have been removed. The organisation now wants to know whether the association 'control from headquarters' still matters. Version A retains the original instruction: 'Every employee must register their check for management review.' Version B gives a factual, non-coercive explanation: 'Use the check to spot hazards for yourself and colleagues. The data are used to improve safety, not to rank individual employees.'

Version B is ethical only if that is genuinely how the data will be used. Otherwise it is a misleading message, not a credible intervention. First study whether the variants activate different meanings at all. Do not assume that warmer language is automatically more effective.

Then compare actual completed checks, error rates and task experiences over an agreed observation period, ideally using an allocation design that limits alternative explanations. If people prefer B but their behaviour does not change, associations may not be the principal barrier. If both meaning and action improve, the result deserves attention, but its causal interpretation still depends on sampling, randomisation, fidelity and context.

A communication campaign can receive excellent evaluation scores yet do little to change day-to-day action. In an experimental meta-analysis, Webb and Sheeran examined whether increasing intentions subsequently changed behaviour. Their synthesis showed that intentions and actions do not change in identical proportions. This is evidence across varied study contexts, not a forecast for one specific Neurofactor project. [S03]

Sheeran and Webb also describe the intention-behaviour gap: people who genuinely intend to act may still encounter missing cues, poor opportunities or competing routines. That is why a rise in declared willingness is not sufficient proof of implementation. [S04]

Choose behavioural endpoints before launch. They might include verified completion, correctly executed steps, behaviour under high workload and continued adherence after the initial attention fades. Report attitudinal or associative change separately. Both kinds of data can be useful, but they answer different questions. A successful campaign impression does not necessarily produce a safer shift.

Wood and Rünger describe habits as responses that are learned through repetition in stable contexts and can operate efficiently without continuous deliberate control. An employee can approve of a new procedure and still carry out the old routine when the daily cue arrives. What looks like reluctance might simply be a well-practised sequence. [S05]

An if-then implementation plan makes a trigger explicit: 'If I collect the keys for the next site, then I open the safety check first.' Gollwitzer and Sheeran's meta-analysis supports the usefulness of implementation intentions for goal pursuit in many contexts, while leaving room for variation between tasks and settings. [S09]

Habits can also be socially meaningful. 'This is how experienced professionals work' can link an old routine with identity. In that case investigate both the learned cue-response pattern and the social story around it. Changing only the message is unlikely to displace a stable routine; changing only the routine may miss a legitimate concern about what the new system represents.

A reminder, a clear default or a well-positioned cue may support people who understand a task, have the necessary resources and want to make a relevant choice. None of these fixes a platform they cannot reach or a form they cannot understand. The Neurofactor knowledge-base entry on nudges explains the concept and its practical boundaries.

Treat the choice environment as an organisational responsibility. Make the action understandable, accessible and proportionate. Where appropriate, retain a meaningful choice, particularly in public services. Check that people with lower digital literacy, disability-related access needs or unusually demanding work circumstances are not made worse off by the design.

Name the intervention precisely. The taxonomy developed by Michie and colleagues offers terms for individual behaviour change techniques, supporting more consistent reporting and evaluation. 'We ran a nudge campaign' says too little about which cue, feedback, planning step or environmental change actually occurred. [S11]

A new procedure can look successful during the first week because supervisors are paying unusual attention. Later, compliance may decline. Staff might also find a shortcut that increases recorded completion while reducing the quality of the actual safety review. This is why dashboard activity cannot be the only endpoint.

Separate implementation as intended, actual behaviour, acceptability, feasibility and maintenance. Proctor and colleagues distinguish implementation outcomes from later service or effectiveness outcomes, a useful reminder that adopting a procedure and achieving its real purpose are not the same thing. [S12]

Use a baseline and comparable follow-up periods. Randomise groups or stagger a rollout when appropriate; consider seasonality, workload and differences between sites. Investigate unintended burdens, privacy concerns and avoidance strategies. The entry on correlation versus causation covers the basic inference problem. An improvement that happened during campaign week was not necessarily produced by that campaign.

1. Calling an observation 'resistance'. Name the missed behaviour and admit that the explanation is unknown. 2. Treating preference as action. Agreeing that safety is important does not mean the check is carried out. 3. Psychologising an operational barrier. Time, access, clarity and realistic autonomy are part of the causal system.

4. Turning an IAT score into a personality label. A relative measure cannot offer certain individual predictions. 5. Leaving the organisation unchanged. If it uses data to control people, friendly words alone cannot establish trust. 6. Reporting only an average. Examine differences across tasks, shifts, sites and accessibility needs. 7. Stopping the evaluation at launch. Track repetition, errors, side effects and maintenance.

The recurring mistake is becoming committed to an intervention before learning which barrier exists. Strong behavioural work can conclude that associations were not the cause, that no campaign is required or that management practice must change. Those are valuable findings, not failures of the research.

A behaviour problem is an association problem only when a meaning attached to the action, sender or situation is demonstrated to be a relevant barrier. Check capability and opportunity before explaining non-compliance as a motivational defect. Test how habits, group norms and associations relate to the behaviour in its real context. Then choose an intervention that addresses the best-supported mechanism and evaluate real action, not simply favourable reactions. Neuroscience and indirect measures can contribute specific evidence, but they cannot replace sound diagnosis. For the underlying framework, read the behavioural diagnostic framework. An indirect association method should be selected only when the research question genuinely requires it. Sometimes communication helps change behaviour. Sometimes change begins when an organisation stops asking the wrong question. [S01] [S03]