Unconscious Bias
Unconscious bias refers to associations, assumptions, and evaluative tendencies that can influence how people perceive and respond to others without full awareness or deliberate intention. Often discussed as implicit bias, it can conflict with consciously held values. Understanding it requires attention not only to individual judgement, but also to power, decision processes, and organisational conditions that can produce unequal treatment.
What it is
Meaning and conceptual boundaries
“Unconscious bias” is a widely used umbrella term, particularly in leadership, human resources, and inclusion work. In research, “implicit bias” more often refers to evaluations or associations assessed indirectly rather than through a person's deliberate self-report. Neither term identifies one single mechanism. Automatic activation, limited awareness, lack of intention, and difficulty controlling a response are related but distinct properties; they should not be treated as interchangeable.
A person may sincerely endorse equality while noticing some candidates more readily, interpreting the same behaviour differently, or granting certain colleagues more benefit of the doubt. This does not mean that every unequal decision is unconscious, or that conscious beliefs have no influence. The practical concern is how assumptions enter decisions and affect other people, not simply whether someone considers themselves fair.
Several distinctions help keep the discussion precise:
- Stereotype: An association or belief connecting a social group with particular characteristics. It may be explicitly endorsed or activated without deliberate endorsement.
- Prejudice: An evaluation or attitude toward people because of group membership. It can be favourable or unfavourable and explicit or implicit.
- Discrimination: Unequal treatment related to group membership or another relevant characteristic. It concerns actions and practices, not merely internal associations. Its legal meaning depends on the jurisdiction.
- Cognitive bias: A broader tendency in judgement, such as confirmation bias or anchoring. It may contribute to social inequality, but is not necessarily about social groups or outside awareness.
- Structural inequality: Disadvantage embedded in rules, resources, networks, or institutions. It can persist even when individual decision-makers intend to be fair.
These distinctions matter because an intervention aimed at awareness cannot, by itself, correct an inaccessible recruitment process or a promotion rule that rewards unequal access to opportunity.
How bias can develop and enter judgement
People learn associations through repeated exposure to social messages, representation, personal experience, and organisational routines. Categories help make a complex world manageable, but can encourage assumptions that outrun the evidence about an individual. Cultural stereotypes may become familiar even to people who reject them; familiarity is not the same as endorsement, and group membership does not make anyone automatically immune to bias.
Dual-process models distinguish relatively automatic processing from more deliberative processing. Daniel Kahneman's “System 1” and “System 2” vocabulary is a useful illustration, not a literal map of two separate brain systems. Bias is not simply a defect of fast thinking: careful reasoning can also selectively justify a preferred conclusion. Slowing down helps only if people reconsider their criteria and evidence rather than rehearse the original judgement more elaborately.
Ambiguous criteria, incomplete information, time pressure, and unstructured discretion can create opportunities for assumptions to influence a decision. These conditions are practical risk points, not proof that a particular person will discriminate. A vague instruction to choose someone with “leadership presence,” for example, leaves considerable room for a familiar style to be mistaken for capability.
Common patterns in professional settings
The following are useful descriptive labels, not separate diagnoses or a set of scientifically identical constructs. The same decision may involve several patterns, and some may be consciously recognised.
- Affinity and in-group preference: Familiarity with a person's background, interests, or communication style is mistaken for trustworthiness or suitability. Informal referrals and “culture fit” discussions are common places to examine this.
- Halo and horns effects: One favourable or unfavourable impression colours unrelated judgements. An impressive qualification may overshadow weak work evidence; one awkward presentation may eclipse strong performance.
- Confirmation bias: Evidence is sought or interpreted to support an existing view. A manager expecting poor performance may notice mistakes and discount improvement.
- Attribution asymmetry: Similar outcomes receive different explanations. One person's success is described as talent and another's as luck; a mistake is attributed to character in one case and circumstances in another.
- Unequal standards and stereotype-linked expectations: Assertiveness, ambition, competence, care, or availability are judged differently depending on assumptions about gender, age, disability, ethnicity, social background, or other characteristics. A complimentary stereotype can still restrict opportunity.
- Visibility, proximity, and recency effects: Being nearby, speaking frequently, or having a recent success is confused with overall contribution. Remote work, part-time schedules, or less visible responsibilities may make this particularly important to examine.
These examples indicate questions to investigate, not evidence that a named colleague holds a particular unconscious belief. Examine actual criteria, decisions, opportunities, and consequences before drawing conclusions.
Impact, power, and accumulated disadvantage
Bias can affect who is heard, whose ideas receive credit, who receives developmental feedback, and who is offered assignments that later become evidence of readiness for promotion. Small differences may accumulate: if someone receives fewer demanding projects, their later experience gap can be used to justify another exclusion. What looks like a neutral assessment of experience may therefore reflect earlier decisions about access.
The consequences also depend on power. A manager's assumption can influence pay or progression; a coach's assumption can shape which possibilities a client is encouraged to explore. Intersecting characteristics can produce experiences that are not captured by considering each category separately. Asking only whether women, for example, are treated fairly overall may conceal differences related to disability, ethnicity, or socioeconomic background.
Unequal outcomes are signals for investigation, not automatic proof of a specific psychological mechanism. Conversely, a person's good intentions do not establish that a process is fair. Responsible analysis connects patterns to how decisions are made, considers alternative explanations, and listens to those affected without requiring them to diagnose other people's minds.
The Implicit Association Test: use and limitations
Introduced by Anthony Greenwald, Debbie McGhee, and Jordan Schwartz in 1998, the Implicit Association Test (IAT) compares response times when concepts share different sorting categories. In a particular task, faster responses for one pairing relative to another can indicate stronger relative associations. The task does not directly observe an unconscious belief or determine why an association developed.
The relationship between implicit measures and behaviour remains an active research area. Associations at group level do not turn a single test result into a reliable prediction of a specific person's conduct. Scores can vary with context, task features, practice, and measurement error. A relatively neutral result does not certify someone as unbiased, and an association favouring one group does not establish an immutable identity or a clinical diagnosis.
Project Implicit recommends treating individual feedback as an opportunity for reflection rather than a perfect or permanent measure. In coaching, participation should be voluntary, interpretation cautious, and privacy explicit. Do not use a test result to label a client racist or sexist, compel a confession, rank employees, or infer suitability for a role. Refusal to take a test is not evidence of bias. Behavioural examples and decision processes are generally more actionable than a score alone.
Awareness is not the same as behaviour change
Awareness can provide a useful starting point, but knowledge, measured associations, behaviour, and institutional outcomes are different targets. An engaging workshop or a positive participant reaction does not demonstrate improved hiring or fewer discriminatory decisions.
Forscher and colleagues' 2019 meta-analysis synthesised 492 studies involving 87,418 participants. It found that procedures could change implicit measures, usually with relatively small effects; most studies examined short-term changes from brief interventions. Changes in behaviour were generally trivial, and changes in implicit measures did not explain changes in behaviour. This does not prove that all inclusion initiatives are ineffective. It does caution against assuming that changing a test score will automatically change conduct.
The Equality and Human Rights Commission's 2018 evidence assessment likewise distinguished awareness raising from changes in bias and behaviour. Evidence for behavioural change was limited. A defensible approach therefore combines learning with changes to everyday decisions, opportunities to practise, and evaluation of meaningful outcomes. No intervention should be presented as permanently eliminating bias.
From individual reflection to process design
Practical safeguards aim to reduce opportunities for irrelevant assumptions to determine outcomes. They should be chosen for the specific problem and evaluated rather than treated as universal cures.
- Specify relevant criteria before evaluating people. Define what competent performance looks like and which evidence is needed. Avoid retrospectively changing the standard to favour a preferred candidate.
- Structure evidence gathering. Use comparable questions, work samples where appropriate, and anchored rating criteria. Accommodate legitimate access needs; identical treatment is not always equitable treatment.
- Separate initial judgements from group influence. Independent assessments before panel discussion can make disagreement and inconsistent standards visible. A diverse panel is not automatically a fair panel.
- Examine access as well as selection. Review how nominations, mentoring, stretch assignments, networks, and feedback are distributed. A formally fair final decision cannot erase unequal preparation opportunities.
- Make decisions explainable and reviewable. Record relevant reasons, allow concerns to be raised, and check patterns over time. Where lawful and appropriate, analyse outcomes using sufficiently protected data; do not collect sensitive information casually.
Perspective-taking, contact, and exposure to counter-stereotypical examples may support learning, but their effectiveness depends on context and cannot be assumed. People affected by exclusion should not be required to disclose painful experiences or educate colleagues. Automated systems also require scrutiny: historical data and proxy variables can reproduce unfair patterns without revealing an individual decision-maker's intentions.
Responsibility without shame or excuse
A productive stance combines curiosity with accountability: people can examine an assumption without being reduced to it, while still taking responsibility for its consequences. Saying that bias is common must not become “everyone does it, so nothing needs to change.” Equally, moral labelling can displace the more demanding work of reviewing behaviour, repairing harm, and changing procedures.
Coaching can support that work, but does not replace safeguarding, grievance procedures, investigation, or legal duties. A report of discrimination should not automatically be reframed as a misunderstanding or a problem with the affected person's resilience. Appropriate organisational or specialist action may be needed alongside, or instead of, coaching.
Sources and further reading
- Forscher, P. S., Lai, C. K., Axt, J. R., Ebersole, C. R., Herman, M., Devine, P. G., and Nosek, B. A. (2019). A meta-analysis of procedures to change implicit measures. Journal of Personality and Social Psychology, 117(3), 522–559. Full research article.
- Atewologun, D., Cornish, T., and Tresh, F. (2018). Unconscious bias training: An assessment of the evidence for effectiveness. Equality and Human Rights Commission, Research Report 113. Evidence assessment.
- Project Implicit. Frequently asked questions. Research-informed explanations of implicit attitudes, stereotypes, the IAT, and interpretation limits; a primary account from the project, not an independent validation of every use of the test.
How it is used in coaching
Contract around a decision, not a label
Start with a concrete challenge: a promotion decision, recurring meeting exclusions, inconsistent feedback, or the client's concern about their own assumptions. Clarify what the client can change, who may be affected, and how progress will be assessed. With an organisational sponsor, agree confidentiality, reporting boundaries, and any limits arising from safeguarding or professional obligations. Avoid promising to uncover hidden beliefs or make a client “bias-free.”
A practical inquiry and action cycle
The following is an illustrative coaching structure, not a validated treatment protocol.
- Reconstruct an actual episode. Ask what happened, what information was available, which criteria were used, and who had decision-making power. Separate observation (“she challenged the deadline twice”) from interpretation (“she is difficult”). Do not require unnecessary disclosure of identifiable or sensitive information.
- Surface the assumption. Explore what the client inferred, how certain they were, and what made the interpretation feel plausible. Ask whether the same behaviour would receive the same description from someone with a different background or status. This is a prompt for inquiry, not proof of prejudice.
- Test alternatives and standards. Look for missing evidence, counterexamples, contextual explanations, and inconsistent expectations. Ask whose perspective is absent and whether the criteria are genuinely relevant to the role. Avoid replacing one untested assumption with another.
- Consider impact and responsibility. Examine what opportunity, feedback, recognition, or treatment followed. Intent may help explain the episode, but does not settle its impact. Identify whether repair, consultation, or a formal organisational response is needed.
- Build a specific safeguard. Agree a behaviour or process change: predefined criteria, independent scoring, an evidence check before using a personality label, or a review of assignment allocation. An implementation intention can help: “If I write ‘not ready,’ I will specify the requirement, supporting evidence, and the opportunity the person had to demonstrate it.”
- Review real-world results. Set a review point and examine whether decisions became more consistent, reasons more defensible, and access more equitable. Include unintended effects and feedback where participation is appropriate and voluntary. Revise the safeguard rather than treating completion of an exercise as success.
Illustrative promotion example
A leader says that an employee lacks “executive presence,” despite strong delivery. The coach asks for observable evidence, the requirements of the next role, and examples of how those requirements were applied to other candidates. The conversation reveals that an informal, extroverted presentation style has been treated as a proxy for leadership, while the employee has had few opportunities to present to senior stakeholders.
The leader and relevant colleagues define role-relevant behaviours, gather comparable evidence, and review access to visible assignments. They do not assume the employee should be promoted regardless of evidence, nor coach the employee simply to imitate the dominant style. The aim is a more defensible decision and fairer development opportunity. This is a fictional example, not a documented case or evidence that every use of “executive presence” is biased.
Working with defensiveness and with lived experience
A client who says “I treat everyone the same” can be invited to examine a particular decision without being pressured to accept a moral label. Ask what evidence would support or challenge their view. Acknowledging discomfort can support learning, but the coach should not collude with dismissal of harm. Encourage specific ownership: what happened, what needs correcting, and what will be different next time?
When a client describes exclusion, first understand the experience and what support they want. Do not insist on a charitable explanation, ask them to prove another person's unconscious motive, or make adaptation to discrimination the default goal. Discuss options, agency, risks, and appropriate support. In team work, create clear participation agreements, do not compel public admissions or personal testimony, and provide ways to raise concerns outside the group.
The coach's own reflective practice
Coaches also make assumptions about ambition, family priorities, communication, capability, or emotional expression. Notice whose story you find immediately credible, whose behaviour you challenge, and when you steer a client toward a culturally familiar definition of success. Ask rather than infer preferences from identity, accent, appearance, or seniority.
Use supervision, feedback, and reflective notes to examine these patterns while protecting confidentiality. Seek understanding of the client's context without expecting them to represent an entire group. Acknowledge and repair a mistaken assumption directly. The goal is not to declare the coach free of bias, but to maintain a practice in which interpretations are tentative, power is considered, and errors can be corrected.
Questions that deepen the work
- What did you observe, and what meaning did you add?
- What evidence would make you change your interpretation?
- Would the same behaviour be described differently if the person's identity or status changed?
- Who received the opportunity to demonstrate the quality you are assessing?
- Is this an individual judgement problem, a process problem, or both?
- What has been the impact, and what responsibility follows from it?
- What safeguard will operate when you are busy, tired, or under pressure?
- How will you know whether treatment or opportunity changed, rather than awareness alone?
Key benefits & applications
- More disciplined judgement: Separating observation from interpretation helps make assumptions and missing evidence visible.
- More consistent people decisions: Role-relevant criteria and reviewable reasons can reduce the scope for arbitrary or unequal standards.
- Greater intellectual humility: Clients learn to question first impressions without assuming that every intuition is wrong or every inequality has the same cause.
- More equitable access to development: Examining assignments, feedback, and sponsorship addresses the opportunities behind later performance judgements.
- Stronger accountability: Attention to impact supports repair and process change rather than relying on good intentions or training attendance.
- More inclusive coaching: Reflective practice can reduce the risk of imposing the coach's own cultural assumptions on a client's goals and choices.
- Better evaluation of inclusion work: Distinguishing awareness, test results, behaviour, and outcomes makes claims of progress more meaningful.
These are potential benefits, not guaranteed effects of a workshop, assessment, or coaching conversation. Sustained improvement depends on practice, organisational conditions, accountable decision-making, and evaluation; reflection alone cannot remove structural inequality.
What our members have to say
Perspectives from coaches and institutes in the CoachScape community.
Member perspectives on this topic will appear here soon.
Bibliography
A curated selection of the most important literature on this topic.
No bibliography has been published for this topic yet.
Submissions are open to coaches and companies and are reviewed by our editors.