Cognitive Ability Tests

General orientation, not legal advice; the position on these tests differs substantially by jurisdiction.

Once the criteria described in “Cognitive Ability Tests” are fixed, the practical challenge is running them consistently across assessors. A coordinator can use productivity tracking platform to plan review time, compare the effort spent at each stage and spot avoidable delays, without confusing administrative efficiency with the quality of the evidence collected from candidates.

For an independent benchmark, compare the process with NIST AI Risk Management Framework; the useful test is whether the local method remains job-related, proportionate and explainable.

Cognitive ability tests measure reasoning, numerical and verbal comprehension. They predict reasonably well and carry specific risks.

What they are good at

Predicting performance across a wide range of roles, particularly complex ones.

Working where candidates have no directly relevant experience, which makes them useful for entry-level and career-change hiring.

And producing comparable scores cheaply at volume.

The adverse impact problem

Score differences between demographic groups are well documented for these tests.

Which means using them as a hard filter can produce substantially different pass rates between groups, and that is a legal exposure in many jurisdictions as well as an unfairness.

This is the central issue with the method and it cannot be designed away entirely.

Its own note covers measuring adverse impact in your own process, which is the minimum responsible step.

Using them less badly

Not as a single hard cut-off. A threshold at a level genuinely required by the role, rather than ranking.

Combined with other methods rather than alone, which reduces both the impact and the reliance.

With practice materials provided in advance, which reduces the advantage held by candidates who have taken such tests before.

And with the job analysis actually supporting the need for it, which is the defence if challenged.

Practice effects

Scores improve with practice, substantially on the first few attempts.

Which means an untrained candidate is disadvantaged relative to somebody who has taken twenty of these during a previous job search.

Providing practice tests to everybody is cheap and levels this. Few organisations do it.

Accessibility

Timed tests disadvantage candidates with several conditions in ways unrelated to job performance.

Extra time is a standard adjustment and should be offered proactively rather than only on request.

And some formats are difficult with assistive technology, which is worth checking before deployment rather than when a candidate raises it.

Where they do not belong

Roles where the job analysis does not identify complex reasoning as a requirement.

As the first filter on a large pool, where the adverse impact falls hardest and the information gained is least.

And in jurisdictions where their use in selection is restricted, which is a real constraint in some places and worth checking.

The honest position

They work, in the sense of predicting something.

They also carry a fairness cost that work samples and structured interviews do not, for similar or better prediction.

Which means the case for using them at all is weaker than their popularity suggests, and an organisation that can achieve its aims with a work sample probably should.

What to check

Does your job analysis support the need for one?

Have you measured pass rates by group?

Are practice materials offered to everybody?

And is extra time offered proactively, or only when somebody asks?

The point

Cognitive tests predict reasonably and carry a fairness cost that work samples do not, for similar prediction.

The case for using them is weaker than their popularity.

Underlying all of this

Almost everything in this collection reduces to one discipline: write down what the job requires, assess that thing directly, record the evidence, and look at your own outcomes afterwards. None of it requires buying anything, and organisations that do those four things consistently outperform ones running longer processes built from instruments chosen before the requirements were known.