Implicit Bias: Measuring It and Its Origins

While I found the implicit bias tests interesting, I couldn’t help but wonder how they designed these tests… Perhaps this wondering is more of a result of my training as a graduate student, but it led me to some interesting tangents…

First off, the tests. I noticed that they never explored all permutations of the test conditions in any one session (i.e., good left + bad right, bad right + good left, Thing1 left + Thing2 right, Thing2 left + Thing1 right, good/Thing1 left + bad/Thing2 right, good/Thing2 left + bad/Thing1 right, bad/Thing1 left + good/Thing2 right, bad/Thing2 left + bad/Thing1 right… the idea of permutations is reflected in the image below). Therefore, I couldn’t help but wonder if results might be skewed because they never tested all of these combinations, leading to implicit bias imposed by the test-makers themselves! Or perhaps they actually tested this and learned that only a subset of these combinations are necessary to detect implicit bias (which I would definitely be interested to read about! I love this kind of research. This might also be accomplished using an idea similar to counterbalancing from usability engineering, but I didn’t see any evidence of that between the multiple IATs I took).

Similarly, I couldn’t help but wonder how they take mistaken keystrokes, misread words, and previous “training” into account. First and foremost, mistaken keystrokes could be just about anything (e.g., finger twitch), but I would guess it’s safe to assume that they’re meaningful given the intent is to measure people’s instinct or gut reactions. However, this brings me to the other two… For example, more often than not, when the word “Terrific” came up and in the split second I started to read the word, I would often just take the first half of the word (“terr”) and assume a negative word (e.g., “terrible” or “terror”). This meant that, more often than not, I would classify this word as “bad,” even though “terrific” is a good thing. That is, my judgement of this word had nothing to do with any biases I had (believe me, I categorized it with bad/left and bad/right very consistently) and more to do with the implied time pressure; I was reading the word too fast to get an accurate judgement of what it said. (On that note, would their results be different with a vocabulary of words that are 5/6 letters or less? Or at least a vocabulary with more consistent lengths of words?)

Moving on to the “training” aspect, the first two sessions of any test trained good on one side + bad on the other, and Thing1 on one side + Thing2 on the other. This “training” certainly give people practice for what the tasks are (and may also be used for baseline results– who knows), but it also makes future tasks that mix these orderings more difficult; simply put, by the time that the left/right orderings have been switched for any particular variable, participants have already gotten significant exposure to and practice for the ordering being the other way around. This would definitely impact results… Do the test-makers account for this?

Image result for training

So, all of that aside, I also began wondering where implicit bias comes from… We all definitely have it, at least on certain subjects to some degree, but how do we “get” it? It seems to me like it’s something that is learned somehow… But where? Family, upbringing, and neighborhoods seem like obvious choices; they certainly have a heavy influence in how we all form our opinions and become the people we are. But what about less obvious, perhaps even more sinister, things?

Here, my mind took a turn to associations and how we learn them. For example, a child might learn from parents that stove = hot (and should not touch), but this lesson is often not truly learned until the hot = ouch association is learned (usually though touching it, just like mom told you not to). So, just how far can these associations go? In what ways do they play a role in how we interact with and understand the world? Take colors for another example… What’s your favorite color? Bet you just thought of one. Maybe several. Why are they your favorite colors? Might take you a second, but I bet you can think of a reason. Can you remember when that reason became the reason why your favorite color is what it is?

Image result for colors

Ok, now, what about light vs dark? Think back to just about every book, fantasy story, sit com, cop show, or anything else you can think of. Light = good, hope, success, happiness, and dark = bad, dangerous, scared, evil. (On that note, can you think of an example of a color used to signify poison that isn’t green or purple? I’m getting away from the point here, but I also think color associations are fascinating…pg22….) This means that from our infancy, we were taught that good things are light and bad things are dark. (You can usually tell in an instance whether a character in a movie or TV show is good or bad based on this, even if it’s just the color of their hat that’s different, and even if the character hasn’t said or done anything yet.) This idea is continuously reinforced. Everywhere. How much of an influence does this kind of “training” or “teaching” have on how we perceive people with lighter-toned skin vs people with darker-toned skin? (Ignoring the long, dark history behind this prejudice, social upbringing and family influences, etc. that probably have a much greater influence over how we perceive lighter-toned people vs darker-toned people… But I can’t help but wonder, just how much influence does it have? Where and how do we learn these prejudices?) How much influence does this have on the Skin-tone IAT? Would people whose favorite color is black perform differently on this IAT (or maybe the Race IAT) than someone else whose favorite color is white (but is equal in as many other aspects as possible)? (Although, now that I think about it, I’ve heard of people’s favorite color being black, but never of a favorite color being white. Where does that difference come from??)

Image result for poison

(Another tangent: I wonder if some of these associations are why there is so much attraction to certain things, simply because you’re not supposed to like them on some level. The “bad boy.” Villains, even ones like Elphaba or Dr. Horrible or Malcolm Reynolds that are the star of the show. What about satire and certain types of humor? Adults liking children’s shows? Hmmmm…..)

Image result for good vs bad


Edit: Just happened to come across this highly relevant video from a science group I follow on Facebook:



  1. I have wondered about possible flaws in the IATs too. Just as an example while the instructions direct participants to act as fast as they can, are they actually considering the response times in the results? Given that people have different ideas about “as fast as possible”, one would think it should be somehow modeled and factored in. Without having done rigorous research about the integrity of the IATs, I have decided to regard it as circumstantial evidence for the existence of bias, rather than hard proof. Something that plants the good seeds of doubt in our over-trained minds.

    PS :
    Speaking of adult’s liking (evil) cartoon characters: Shego in Kim Possible


    1. Sorry for the late reply– it’s been one of those weeks!

      Yes, I agree that the IAT seems to have its flaws, and prior to class last week, I had more doubts about the IAT’s ability to uncover implicit bias than I did when I wrote this. I would still be interested to learn more details about what data they use and how they use it to determine whether implicit bias exists and how strong that bias is.

      And talk about a flash back to the past!! I loved Kim Possible as a kid– it was a great show!


  2. I enjoyed reading your thoughts on the IAT. I worked as an undergraduate research assistant in a lab that studies (explicit and implicit) bias. The IAT is by no means a perfect measure. To your point about the training block (which isn’t used in scoring) affecting performance on the subsequent task: blocks are presented in a random order, which should eliminate bias across participants. Researchers typically also set a certain threshold of accuracy for a participant’s data to be included in your analysis; the idea here is that if someone’s accuracy is very low, perhaps they weren’t paying attention to the task at all or read the instructions incorrectly. Another task used to assess attentional bias is a dot-probe task (Here’s an example if you’re interested: These tasks can be used together- in order to provide evidence using 2 different methods.

    Liked by 1 person

  3. Sorry for the late reply… It’s been a very busy week!

    Thank you for the additional information about the IAT and how they score implicit bias… I find this whole topic and research area fascinating! On the subject of the dot-probe task, I’m curious to know how the results from that kind of IAT test compare to the results from the test that we took. I would assume that the results are very similar (it is a rather similar study design after all), but, again, this whole subject is rather interesting to me, so I can’t help but wonder if results would come out differently and, if so, how different?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s