Unconscious Bias at Work: What It Is and What Actually Works
Unconscious bias is an automatic association about a group of people that influences a decision before you’ve consciously thought about it. Everyone has these associations. They form from repeated exposure to patterns in your environment, not from personal malice, and they show up most strongly in fast judgements made under time pressure.
That definition is uncontroversial. What happens next is where most organisations go wrong.
The standard response to unconscious bias is a training session. Sit everyone in a room, explain the concept, run the exercise where people notice their own assumptions, and tick the box. UK evidence suggests that approach, on its own, doesn’t work. This piece covers what unconscious bias is, the types that affect real workplace decisions, and what the research says about reducing its effects. If you’re responsible for hiring, promotion or performance decisions, the last three sections matter more than the first three.
What is unconscious bias?
Unconscious bias is a mental shortcut. Your brain handles far more incoming information than conscious attention can process, so it sorts people and situations into categories automatically and attaches expectations to those categories. Most of the time this is useful. You don’t reason your way through a doorway.
The problem starts when the categories are about people and the expectations are wrong.
Unconscious bias definition: An automatic, learned association between a social group and a set of attributes, which influences judgement and behaviour without the person being aware it’s happening.
Three things follow from that definition, and they’re worth being precise about.
These associations are learned, not innate. They come from media, family, workplace norms and repeated exposure. A person who has only ever seen one kind of person in a senior role will form an expectation about who belongs in senior roles.
They don’t match stated beliefs. Someone can hold sincere egalitarian views and still show a measurable automatic association pulling the other way. This is why unconscious bias is not a polite word for racism or sexism. Those are conscious positions people choose to hold.
They operate strongest under pressure. Time constraints, cognitive load, fatigue and ambiguity all increase reliance on automatic processing. A recruiter scanning ninety CVs in an afternoon is running on shortcuts, whatever their intentions when they started.
Unconscious bias meaning: what the term covers and what it doesn’t
The phrase gets stretched a long way. In workplace conversation it’s used to mean anything from a genuine automatic association through to open discrimination that someone would rather not name directly. That vagueness causes real problems, because the two need completely different responses.
Here’s the distinction that matters in practice.
| Unconscious bias | Explicit bias | |
| Awareness | The person doesn’t know it’s operating | The person holds the view knowingly. |
| Origin | Absorbed from the environment and repetition | Adopted as a belief |
| Where it shows | Split-second judgements, gut feel, “culture fit” | Stated opinions, deliberate exclusion |
| What changes it | Changing the decision process | Conduct policy, and sometimes dismissal |
| Legal position | Effects can still be unlawful discrimination. | Usually clearly unlawful |
That last row catches people out. Under the Equality Act 2010, intention is not the test. If a recruitment practice puts people with a protected characteristic at a disadvantage, it can amount to indirect discrimination regardless of whether anyone meant it to. “We didn’t realise” is an explanation, not a defence.
So when someone asks us to define unconscious bias in a training room, we’re careful about one thing above all. The term describes a mechanism. It doesn’t excuse an outcome.
Is subconscious bias the same thing?
Mostly yes, in ordinary use. Subconscious bias, unconscious bias and implicit bias all get used for the same idea, and if you’re searching for one, you’ll find the others.
Academic psychology tends to prefer “implicit bias” and treats “subconscious” as loose language borrowed from Freud. Government and HR documents in the UK say “unconscious bias”. The distinction rarely matters for practical purposes, though if you’re citing research you’ll find it filed under implicit.
The types of unconscious bias that affect workplace decisions
Most articles list fifteen or twenty types. That’s more useful as a glossary than as a tool, because only a handful do real damage to employment decisions. These are the ones that show up repeatedly when we look at how organisations actually hire and promote.
| Type | What it does | Where it bites |
| Affinity bias | Favouring people similar to you | “Culture fit” judgements, informal referrals |
| Halo and horn effect | One strong trait colouring everything else | A prestigious employer on a CV, or one awkward answer |
| Confirmation bias | Seeking evidence for a first impression | Interview questions that test a hunch you formed at minute two |
| Attribution bias | Reading the same behaviour differently by group | Assertive vs “difficult” in performance reviews |
| Contrast effect | Judging a candidate against the previous one | Back-to-back interviews, CV sifting in batches |
| Name bias | Reacting to a name before reading the CV | Application screening |
| Beauty and height bias | Linking appearance to competence | Video interviews, first meetings |
| Ageism | Assumptions about capability or ambition by age | Promotion, redundancy selection, retraining decisions |
| Status quo bias | Preferring the existing arrangement | Referral schemes, “this is how we’ve always recruited” |
| Conformity bias | Adjusting your view to the group’s | Hiring panels where the senior person speaks first |
Two of these deserve more than a table row.
Affinity bias is the expensive one. It’s the mechanism behind “culture fit”, and it’s socially invisible because liking someone feels like a judgement about them, not about you. Organisations that recruit heavily through referrals amplify it structurally: people refer people like themselves, and the workforce narrows year on year without anyone making a single biased decision they could point to.
Attribution bias is the one that damages retention. The same behaviour gets described differently depending on who’s doing it. Direct becomes abrasive. Persistent becomes difficult. Quiet becomes disengaged for one person and thoughtful for another. It surfaces in performance reviews, where the language is subjective and the consequences compound over years.
What the evidence says about unconscious bias training
This is where our position diverges from most of what you’ll read on this topic.
In December 2020 the Cabinet Office announced that unconscious bias training would be phased out across the UK Civil Service, where it had been mandatory for managers since 2015. The decision followed an evidence summary commissioned from the Behavioural Insights Team, which concluded there was no evidence the training changes behaviour in the long term or improves workplace equality in terms of representation. Ministers encouraged local government, the police and the NHS to review their own approaches.
The central study behind that conclusion was a systematic review covering 492 studies and more than 87,000 participants. It found that changes in unconscious bias measures were not associated with changes in behaviour. You can shift someone’s score on a bias test and see nothing at all happen to how they hire.
The Equality and Human Rights Commission went further, noting limited evidence for behaviour change and a risk of backfiring when participants are given the impression that biases are fixed and unchangeable. If the message a manager takes away is “everyone’s biased and there’s nothing to be done”, you’ve handed them an excuse.
We deliver this training. We’re telling you this anyway, because the finding isn’t that awareness is worthless. It’s that awareness on its own doesn’t change what people do. A session that ends with everyone agreeing bias exists, and no change to a single process, has cost you a day and bought you nothing.
Two claims about unconscious bias you should stop repeating
Both appear on most of the pages ranking for this topic. Neither holds up.
“Take the Harvard IAT to find out your biases.” The Implicit Association Test is a research instrument and a useful one for studying populations. It’s weak as an individual diagnostic. The UK Government’s evidence summary noted it has been criticised for failing to produce replicable results when the same people are retested. If your score changes between Tuesday and Thursday, it isn’t telling you something stable about yourself. Use it as a demonstration that automatic associations exist. Don’t use it as a personal report card, and don’t build an intervention on the score.
“Blind auditions increased women in orchestras by 50%.” This is the most cited statistic in bias training, and it’s shakier than its fame suggests. The original 2000 study by Goldin and Rouse is honest in its own abstract: some estimates carry large standard errors, and there’s one persistent effect running the opposite way. The authors describe the weight of evidence as supporting their conclusion, which is a more modest claim than the headline figure implies. Statisticians, including Andrew Gelman, have since pointed out that the confidence intervals are too wide to support the precise numbers that get quoted.
Anonymised sifting is still a sensible practice. It just doesn’t rest on the proof people think it does, and citing shaky evidence to an audience of sceptical managers costs you the room.
What actually reduces biased decisions
The pattern in the evidence points one way. Interventions that change the decision-making process outperform interventions that try to change the decision-maker.
Structure the interview. Same questions, same order, every candidate, with scoring criteria agreed before anyone is interviewed. Structured interviews are among the best-evidenced predictors of job performance, and they remove most of the room in which affinity bias operates. Unstructured conversation is where gut feel wins.
Score as you go, individually, before you discuss. Panel members write and submit their ratings before the group talks. This kills conformity bias, which otherwise means the panel adopts whatever the most senior person says first.
Fix the criteria before you see the people. Write down what “good” looks like for the role, including which requirements are essential and which are preferences. Doing this afterwards lets the criteria bend around whoever you already liked.
Strip identifying detail at the sift stage. Name, address, school, university and graduation year removed from the version the sifter sees. Cheap to implement, and it removes the input instead of asking someone to ignore it.
Audit your own numbers. Application to interview rate, interview to offer rate, promotion rate and pay by group. Bias is hard to see in any single decision and obvious across two hundred. Most organisations already hold this data and never look at it.
Slow down the decisions that matter. Automatic processing dominates under time pressure. A recruiter given twenty minutes per CV makes measurably different decisions from one given four.
Train managers on the process, not the concept. The useful session isn’t “here is what bias is”. It’s “here is how you run a scoring panel, here is how you write feedback that describes behaviour instead of personality, here is what to do when someone on your panel says they just didn’t click with her.”
That last point is where training earns its place. Not as awareness raising, but as skills practice attached to a process that has actually been redesigned. Awareness without process change is theatre. Process change without training tends to be implemented badly and abandoned within a year.
Where UK employment law sits on this
Brief, because this is context and not legal advice.
The Equality Act 2010 protects nine characteristics: age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex, and sexual orientation. Direct discrimination means treating someone less favourably because of one of these. Indirect discrimination means applying a provision, criterion or practice that disadvantages a group, unless you can justify it as a proportionate means of achieving a legitimate aim.
Unconscious bias tends to surface as indirect discrimination, and intent is not part of the test. A tribunal will look at the effect of your practice and at what evidence you have that it’s fair. Documented scoring, consistent criteria and monitored outcomes are the practical answer, which is the same list as the section above.
Public sector organisations carry an additional obligation under the Public Sector Equality Duty to have due regard to eliminating discrimination and advancing equality of opportunity. That duty is about evidence and consideration, not about running a training course.
Frequently asked questions
What is unconscious bias in simple terms?
It’s an automatic association your brain makes about a group of people, which affects your judgement before you’ve consciously thought about it. Everyone has them. They come from repeated exposure to patterns in your environment, not from deliberate prejudice, and they influence quick decisions far more than considered ones.
What causes unconscious bias?
Repeated exposure. Your brain builds expectations from what it sees often: who appears in which roles, how groups are portrayed in media, and what the people around you assume. Cognitive load makes it worse, because automatic processing takes over when you’re rushed, tired or handling too much at once.
Is unconscious bias the same as discrimination?
No. Unconscious bias is a mental mechanism. Discrimination is an outcome, and under the Equality Act 2010 it can be unlawful whether or not it was intended. Bias can produce discrimination, so the absence of intent doesn’t protect an employer whose practices disadvantage a protected group.
Does unconscious bias training work?
For raising awareness, yes. For changing behaviour on its own, the evidence is poor. A UK government evidence review found no evidence that such training changes behaviour long term or improves representation, and the Civil Service phased out mandatory training in 2020. Training works when paired with redesigned decision processes.
How many types of unconscious bias are there?
There’s no fixed number, and lists range from ten to over twenty. For workplace decisions, around ten matter: affinity, halo and horn, confirmation, attribution, contrast, name, appearance, age, status quo and conformity bias. The rest are mostly variations on these.
What’s the difference between unconscious bias and subconscious bias?
In everyday use, nothing. Researchers prefer “implicit bias” and treat “subconscious” as imprecise, but all three terms describe the same thing: automatic associations operating outside conscious awareness.
How can we reduce unconscious bias in hiring?
Change the process, not just the people. Structured interviews with fixed questions, criteria agreed upon before candidates are seen, individual scoring submitted before panel discussion, identifying details removed at sift, and outcome data monitored by group. These change what decisions are possible, which is more reliable than asking people to notice their own thinking.
Talk to us about bias and decision-making.
Goldmark Training delivers unconscious bias and inclusion training to housing associations, local authorities, charities, care providers and private sector organisations across the UK. Our sessions are built around the decisions your managers actually make, and we’ll tell you honestly when training isn’t the thing that will fix your problem.
Programmes are designed and delivered by Tahira Hussain, a BABCP-accredited cognitive behavioural psychotherapist with nineteen years of experience across social work, mental health, housing and the charity sector, and a TEDx speaker on race and belonging.
Call 07476 988566 or book a chat.
About the author. Tahira Hussain is the founder and director of Goldmark Training. She is a qualified cognitive behavioural psychotherapist registered with the BABCP and the British Psychological Society, with nineteen years of experience across social work, mental health, housing, community development and the charity sector. She delivered “I’m Not Racist But…” at TEDx Northwich and is an executive member of the Manchester BAME Network.
Sources. Cabinet Office written ministerial statement on unconscious bias training, 15 December 2020; Behavioural Insights Team for the Government Equalities Office, Unconscious bias and diversity training: what the evidence says (2020); Forscher et al., systematic review of procedures to change implicit measures (2019); Equality and Human Rights Commission, Unconscious bias training: an assessment of the evidence for effectiveness (2018); Goldin and Rouse, Orchestrating Impartiality, American Economic Review (2000); Equality Act 2010.
