1 Decisions

As educators, we need to make decisions. For example, we have to decide

  • how much partial credit to give
  • how much help to give a student
  • how detailed to make our explanation
  • if a student complaint is whining or legitimate
  • who to recommend apply as a TA
  • who to encourage to stay in the field

Those decisions are based on many things. Each one has so many inputs, it would overwhlem our brains small amount of working memory. For example, in deciding how much help to give we might wish to consider

  • are they making progress?
  • have they tried in earnest on their own?
  • are they interested and engaged?
  • if I help, will they still have much to learn?
  • is this motivated by learning or grades?
  • are they fishing for the answer?

Because there are more things to consider than I can think about at one time, I trust my subconscious mind to help me out. A simplistic, but informative, approximation is that I have a schema for each criterion, each saying good choice or bad choice based on that one criterion alone; and that a part of my subconscious does a weighted average of all those criteria’s votes to give me a gut instinct (more correctly a schemata instinct) about what choice feels right.

2 Estimators

How does my subconscious mind evaluate each criterion? By appeal to many other schemata which collectively form an estimator for that topic.

Let’s consider an estimator for is engaged in the topic at hand. A perfect estimator has the property that if the truth is they are 30% engaged then the estimator also says they are 30% engaged. Perfect estimators are so rare as to be practically non-existent.

A fair estimator has the property that if the truth is they are 30% engaged the estimator might report any other number: sometimes down 10% to say 20%, sometimes up 28% to say 58%, and so on; but the average error is 0; that is, after many many runs it over-estimates and under-estimates about as often and by about as much.

A biased estimator is neither perfect nor fair. It does not estimate the right value, and it frequently estimates incorrectly in the same direction.

The words perfect, fair, and biased are also used in less technical settings, but we’ll only need this formal numeric definition for our purposes.

The joint working of a group of schemata, averaged by your subconscious and provided as a single estimate of a criterion or a decision as a whole, is an estimator and can be fair or biased.

3 Biased subconscious processing

Two primary causes can bias our subconscious processors, leading us to make unfair (that is, consistently incorrect) decisions.

3.1 My skill at interpreting

Consider how my subconscious estimates is this student engaged?

One estimator of engagement is facial expressions. Engaged students react quickly and correctly to what I say; disengaged students react less and less consistently. But how do I know what their expressions mean? I’ve learned by looking at a lot of faces.

But that skill is not uniformly distributed. If a student has a face that is biologically like those I’m used to seeing, I’ll rapidly recognize the student’s expressions, emotion-driven changes in complexion, etc. If a student was raised in a similar culture, I’ll easily interpret their culturally-significant cues like eye motion, head shakes, etc. If a student does not match the faces I’ve learned on, I’ll have more trouble noticing these things and misinterpret them more often.

Thus, my skill will cause me to have a biased estimator: people with faces and from cultures like those I see the most my subconscious will estimate as being more engaged than people with faces and from cultures I am less familiar with.

3.2 Expected as estimator for good

Good and bad are complicated concepts the subconscious cannot readily handle, but expected and surprising are much simpler for the subconscious to grasp. And in many, many circumstances, expected is good estimator of good and surprising is a good estimator of bad: a surprising shape in the road is usually a bad sign for a driver, a surprising taste in food is usually a bad sign for health, a surprising question on a test is usually a bad sign for future grades, etc.

This estimator is so universal in its application, your brain applies it everywhere and in ways you don’t want it too. Thus, a student who looks, sounds, or dresses unlike the bulk of students you’ve encountered before your subconscious labels as a bad student. A face that strongly resembles someone you used to know or saw on TV who was violent or boring or annoying or delightful or bright or funny your brain will assume will have those same characteristics, and will mark them as good for roles where those characteristics are an asset and bad for roles where they are not.

4 Implicit bias

To combine the above, an implicit bias1 is

  • a bias in your decision making
  • caused by your subconscious using a biased estimator, generally either
    • an estimator your experience has trained better for some than for others, or
    • an estimator that confuses common with good.
  • and is an inevitable part of having a working brain.

In general, implicit bias will cause you to treat people who your subconscious identifies as being either (a) in the minority in the field or (b) unlike yourself as though they were inferior, less able, less engaged, less deserving, etc. It may also cause you to subconsciously assign traits to people based on other people you’ve known or been exposed to whom they resemble in some way.


  1. or unconscious bias; there is a difference, but we won’t get into it here.↩︎