Getting Over The Bell Curve

The Bell Curve (1994) made some broadly reasonable claims : There is such a thing as general aptitude, one can measure it, and folks with high general aptitude tend to do well at a broad range of tasks. 

Then it strayed into some generalizations about heritability and population differences, which led it to misguided and controversial public policy prescriptions, which attracted criticism, which itself was arguably misguided and probably even fabricated. 

That is how aptitude tests became toxic, which is a pity because they work.  

There is more research than we can cite on how aptitude tests are great predictors of candidate effectiveness (see this, or this, or the hundreds of references in either one).  

Our fundamental belief is that job applicants have a semi-static aptitude to learn new skills, and we have designed tests to help measure the ability and willingness to learn. If employers and platforms (like ours) identify these candidates and invest in them, they will be able to pick up acquired skills as needed. 

This is why we have always used cognitive aptitude tests as our first and primary measure to screen candidates, and why we are doubling down. After many years of hiring based on the cognitive aptitude test and employing high-aptitude workers, here are our learnings:

1. Cognitive tests are definitely valid performance predictors : None of our highly effective hires had scored low on the aptitude test. Basically, we should have trusted the test more. 
2. We need more aptitude tests, not fewer: Some high-aptitude hires did not quite rock their role, but only because we did not correctly weight all elements of aptitude (critical thinking, programming). Again, our large datasets have taught us to trust the tests, rather than our subjective biases.
3. English language testing isn’t “biased” if your job involves English: There are tests (like Raven's matrices) which measure only non-verbal reasoning. However, hiring someone based just on those sets them up for failure. Most real jobs involve a little bit of abstract thinking, and a lot of communicating those abstract thoughts with very non-abstract people. Basic conversational aptitude helps (we know because we test for that too), but unless you can grasp deep ideas and communicate them back clearly and quickly, your colleagues will leave you to your abstract thoughts. 

Aptitude test scores are semi static : A lot of controversy over aptitude tests is about strong claims of determinism. We believe population level differences are bunk, but individual aptitude in any given adult is semi-static. Scores can be improved somewhat by practice, but not very much and not very fast. In some cases, scores on aptitude tests are held back by fluency in human and programming languages. Retaking the test after acquiring such fluency helps. 

Unfortunately, the most common way to change the score is not more practice (which we encourage), but cheating (which we don’t). That is a topic for another post.