The idea that people might have implicit, unconscious
racist, sexist, or ageist biases is one that’s gotten a lot of traction in the
last decade or so. With the rise of movements like Black Lives Matter, the
racial portion of possible implicit bias — setting aside, of course, explicit
bias, has gotten even more traction.
Project Implicit, grounded at Harvard but which also has
itsown website, is the biggest, and original, tester for such attitudes, looking
at not only the three above, but possible implicit bias against sexual
orientation and more.
That said, the question is, how well can we test for this?
Massimo Pigliucci has raised this issue, per one of the
links on a Friday roundup. Two different meta-analyses within that piece both
look at the racial bias test. (And yes, both of them focus on that,
the first citing only race-based testing as explicit examples and
the second specifically saying it was only about racial bias testing. The idea that the first is also about race only is reinforced by
a pieceI posted on Massimo’s site last week, meta-analyzers, while having bits of
problems with the other tests, don’t see the same testing issues in the same
degree with them.)
That all said, let’s dig into this issue, both of implicit bias in general, and why racial implicit bias (presuming it exists, as I shall shortly show we have good reason to believe) may be hard to test accurately, and again, that problem being most difficult with racial bias.
First, my idea of consciousness, as should be known to
long-term visitors to Massimo’s sites, is broadly similar to that of Dan Dennett —
subselves poking up from the subconsciousness, multiple drafts of
consciousness, etc. I have no problem with the idea of implicit biases. In
fact, I’d be shocked if humans did NOT have such biases. Shocked. (For more on that, see the last two links at the bottom.)
Second, let’s look at the main such biases in modern
America, as most commonly tested for by
implicit association tests, whether
crafted by Project Implicit or not — racism, sexism, ageism, sexual orientation.
Sexism and ageism appear to be largely culturally driven. In pre-literate
societies, the elders were valued for their wisdom, doubly so in
pre-agricultural societies. In “neutral” societies, let alone matrilineal ones,
women were valued for inheritance, clan/tribe structure and more,
sociologically; and especially contra a myth of Ev Psychers, weren’t denigrated
for doing non-hunting work, etc.
Racism, though, is different. Vis-à-vis human xenophobia and
outgroups, it appears to be more biologically embedded. Please don’t reference
Hume at this point; I know well that is ≠ ought. BUT … until a desirable
“ought” of cultural evolution fully trumps an “is” of biological evolution, the
“is” is still in play. That’s reality, folks; we’re still blobs of protoplasm,
not Platonic Ideals.
OK, that’s our background.
(Sexual orientation bias is probably the second-hottest
button issue. And, it may have a moderate degree of biological background, if
connected to generalized feelings of disgust. But, in turn, it’s wrapped up
with yet other issues, and that would be too distracting.)
Third, with the above said, can racism-based implicit bias
be tested? I say yes. The critiquers of IAT don’t say no; they just say that IAT
is overstating what it has found, as far as degree of bias and potential
conscious-level effects of that. And, they do focus on race-based bias.
Fourth, why is such testing problematic? Well, racism is the
hot-button issue in America, beyond the biological background, versus ageism or
sexism. The United States is quite arguably the most racially diverse nation in
history, with minorities becoming an ever-greater portion of that. And, America
has nearly 250 years not just of slavery, but of race-based and race-justified
slavery, behind that.
Ergo, it’s understandable that “social justice warriors” or
“special snowflakes” in places like modern American academia might
over-critique themselves to ferret out actual, would-be, or feared-to-be, bits
of such bias. That’s doubly true with the conformist nature of academia in many
cases.
That still doesn’t mean that such bias is nonexistent.
What it does mean is that we have:
1.
Low accuracy in testing for it;
2.
Low accuracy in determining any explicit results
of it (this applying to the other IATs, not just race.
Or, as noted by Wikipedia at the third link above (additional pull quotes also from there):
A recent meta-analysis[30] has
concluded that the IAT has predictive validity independent of the predictive
validity of explicit measures. However, a follow-up meta-analysis[31] questioned
some of these results, finding that implicit measures were only weakly
predictive of behaviors and no better than explicit measures. Some research has
found that the IAT tends to be a better predictor of behavior in socially
sensitive contexts (e.g. discrimination and suicidal behaviour)[32] than
traditional 'explicit' self-report methods,[33] whereas
explicit measures tend to be better predictors of behavior in less socially
sensitive contexts (e.g. political preferences).
That seems to sum up the main issues pretty well.
First sidebar:
The IAT has also demonstrated a reasonable amount of
resistance to social-desirability
bias. Individuals asked to fake their responses on the IAT have
demonstrated difficulty in doing so in some studies. … Distinct from faking
(the deliberate obscuring of a true association), some studies have shown that
heightening awareness about the nature of the test can change the outcome,
potentially by activating different fluencies and associations.
In other words, whatever the IAT is measuring, regardless of the accuracy of measurement as correlated to an alleged specific implicit bias, it's measuring something "real."
At the same time, per the second half of that second pull quote, it does seem moderately susceptible to "priming," which itself an issue that still gets some degree of debate.
Fifth, that said, what’s the answer?
The answer is to replace so-so science (not “bad science,”
in this case, I say, but “so-so science,” viewing it on a continuum rather than
two polarities) with better science. This is not bad science, and it’s
certainly not pseudoscience. (That said, if Anthony Greenwald and Mahzarin Banaji keep digging in their heels, it's going to become semi-bad science soon enough, then bad science after that.)
So, IAT’s founders need to admit their shortcomings, as well
as admitting what seems to be a primary reason for their shortcomings.
Then, social scientists in general need to work on better
testing. They also need to ask the self-reflective question of: “How much
better can we make this testing?”
I would guess “moderately better.” Science can’t know
everything, nor can it get everything 100 percent right. (Sorry, British
astronomers.)
Sixth, the same applies to asking how much we can improve
our knowledge as to how much implicit biases have explicit results. “Moderately
better” is likely the best we can do.
With that all said, we should support better science in this
instance. The answer to not-good science that’s nonetheless science is always
better science, not rejection of existing science.
And, since this is my blog, per one commenter on Massimo’s
site?
Maybe we could substitute “Israeli” and “Palestinian” for “black” and “white.”
And, yes, I went there. And, in that specific order, stating that specific version of the parallelism.
Speaking of, I also don't get how someone so in love with depth psychology can resist the idea of implicit biases, when, especially with Freud's theories, the idea of implicit bias fits like a T.
To pivot this to politics ...
We leftists and left-liberals must not let conservaDems, classical liberals, or whomever claim that the Democratic Party caved into identity politics at the expense of the working class.
Democratic Party local leaders stress that's
a false dichotomy. And they're right.
And, we who vote outside the duopoly box must insist on this even more.
