Knowles Knowledge Centre•Terminology
Record Audits vs Awareness Training, Implicit Association Tests and Blind Processes
Four approaches, heavily overlapping, routinely used as if they were interchangeable. Here is what each one actually optimises for — and which to ask for when.
The four are not synonyms, and they do not even operate on the same object. Awareness training optimises for understanding: people know that unconscious associations exist. Implicit association testing optimises for measuring associations, at population level. Blind processes optimise for a decision’s inputs. A record audit optimises for something none of the others produce — one number about one person’s actual practice.
Choosing the wrong one is not a semantic problem. It leads to briefing the wrong intervention, buying the wrong course and measuring the wrong thing — typically whether people understood the session, when the question was whether any decision changed.
About this guide
The four approaches
Each has a defensible definition. The trouble starts when they are used loosely.
Inclusions, exclusions and overlaps
What each approach covers, what it explicitly does not, and where it shares ground with the others.
Includes
Excludes
Overlaps with
Awareness training
Concepts, mechanisms, examples, a shared vocabulary
Whether anybody decides differently afterwards
Record audits, as the thing that makes a number discussable
Implicit association testing
Population-level evidence that associations exist
What any individual will actually do
Record audits, in aim only — different object entirely
Blind processes
What information reaches the decision-maker
Decisions where the information cannot be removed
Record audits, as the usual remedy when a gap is found
Record audits
A written rule, twenty cases, a third-party score, one number
Attitudes, associations, intent, and anything unrecorded
All three of the above, at the point evidence is needed
The second row is the one worth pausing on. A record audit and an association test look like the same project — both produce a number about a person — and they measure entirely different things. The test measures an association; the audit measures a set of decisions. Only one of them is about anything anybody did, and only one of them its authors would endorse using this way.
Why the terms get confused
This is a known problem with a name.
The US Institute of Education Sciences calls it the jangle fallacy: assuming that two things are different because they carry different names. “Unconscious bias,” “implicit bias,” “subconscious bias” and “hidden bias” describe substantially the same territory with different emphasis, while “cognitive bias” is a genuinely wider category that is frequently sold as the same thing.
Two practical consequences follow. First, a provider offering “implicit bias awareness” and one offering “inclusive decision-making” may be selling the identical session, and comparing them on the label tells you nothing. Second, an organisation searching for one term will miss relevant provision listed under another.
The reliable test is not the label. Ask what each participant leaves holding: an explanation, a test score, a redesigned process step, or a count of their own departures from a rule they wrote down. Two providers using different vocabulary will answer that very differently — and only one of those four answers is about a decision anybody made.
Choose this when…
The decision matrix. Match the situation on the left to the approach, not the other way round.
If the situation is…
…then the need is
Typical intervention
People do not accept that unconscious associations operate at all
Awareness training
Concepts, mechanisms, worked examples, discussion
The organisation needs evidence that this is worth attention
Implicit association testing
Aggregate measurement, reported at population level only
A specific decision takes in information it does not need
Blind processes
Redaction, anonymised sifting, screening at the input
A pattern exists and nobody knows which decision produced it
Record audits
A written rule, twenty scored cases, and one number
Every decision-maker shows the same discrepancy
Probably an unworkable rule
Redesign the rule once, centrally, rather than auditing everybody who departs from it
Faithful application of the rule would not change the pattern
Not an individual need
Examine what the rule encodes, and the pool it draws from
The last two rows exist because none of these four approaches helps when the rule is the problem, and the commonest error in this field is treating a faithfully applied rule as a failure of the people applying it. A method for testing that is set out in how to identify the gap between your stated rule and your last twenty decisions.
What goes wrong when the approach is mis-selected
Three failure patterns account for most wasted spending in this area.
Buying awareness training as behaviour change
This is the characteristic error of the field. A session explains how associations form, participants find it interesting and largely accept it, and the allocation decisions next quarter are indistinguishable from last quarter’s. Awareness reliably delivers a shared vocabulary and a shared baseline, which is genuinely useful and is not what it was bought for.
Using a test score to diagnose an individual
An instrument designed to show that associations exist in a population is used to tell one manager something about themselves, and the result is either defensive dismissal or a guilt that attaches to nothing they can change. The researchers behind these measures have warned against precisely this, and the warning is easier to find than the training that ignores it.
Blinding a decision whose exposure is upstream
A shortlist is anonymised with real care, and the same people are shortlisted, because the pool was narrow before anybody read a name. The blinding worked; it was simply applied to the only stage of the process that was not the problem.
Providers choose whichever label is currently attractive, and most provision in this field is awareness-shaped, because it is deliverable at scale, demonstrates well, and requires no access to anybody’s records. Judge a proposal on one question: does any participant leave with a count derived from decisions they actually made? If not, it is an awareness session whatever the title says.
Frequently asked questions
Conclusion
The four approaches cover overlapping territory for different purposes.
The practical value of getting this right is not precision for its own sake. It is that each approach implies a different intervention, a different measure of success and a different kind of failure when it is used in place of one of the others.
A brief written in the wrong approach will attract the wrong proposals, and the same four names will take the interesting work next year.
The five-stage structure behind the record-audit column is set out in you cannot inspect your own bias, but you can audit your record. Organisations that have located the exposure in a discretion rather than a rule can review the Understanding Unconscious Bias Training Course in Singapore.

