Read on Twitter
Precision in Implicit Social Cognition research can and should be improved 1/n
A thread
A thread
I’m increasingly concerned with conceptual and measurement issues in Implicit Social Cognition research (ISCr).
The reason why I’m concerned is that hardly anything is changing in this research domain.
I’m explaining my main concerns below 2/n
The reason why I’m concerned is that hardly anything is changing in this research domain.
I’m explaining my main concerns below 2/n
• In ISCr, it is considered okay when two measures that correlate by virtually zero (the IAT and the AMP) are said to measure the same construct (“implicit prejudice”) 3/n
Can we imagine two measures of “speed” correlating by zero? We’d sure be heading for serious car and plane accidents. Why not being more concerned about the risk for misdiagnosis ? 4/n
Clearly, one should refer to “IAT” or “AMP” (or to any other poorly correlated) task *scores* and avoid referring to a supra-ordinal category (e.g. "implicit racial preference), when these tasks show so little convergence 5/n
• In ISCr, a same measure is given over half a dozen different names: “name_your_task score”, “implicit attitude”, “implicit evaluation”, “unconscious bias”, “implicit bias”, “implicit prejudice”, “associative attitude”, “implicit preference” 6/n
What’s the reason for this logorrhea? Is it conducive to good communication and sound research advances? 7/n
• In ISCr, definitions may be fully circular. For instance, “implicit preference” may be defined as “behavioral responses on implicit measures”. However, we don’t really know what these measures are and what they measure 8/n
The reason we don’t know is that, in ISCr, the core “implicit” construct is not consensually defined (see Corneille & Hütter, PSPR, 2020). A colleague asked me: “Which is the right measure of implicit prejudice then: the IAT or the AMP”? Here were my tentative responses: 9/n
If you define "implicit" as "indirect", then they both qualify as "implicit measures of prejudice". However, the direct-indirect nature of a measure is agnostic to mental processes, so this definition has little value 10/n
If you define "implicit" as "automatic", the two measures heterogeneously vary along a number of features of automaticity, in all likelihood contributing to their low convergence. Also, by this definition, even a non-speeded self-report may qualify as an "implicit" measure 11/n
If you define "implicit" as capturing "affective gut reactions" (in a physiological sense), then I'm not sure either of these two tasks qualify 12/n
If you define "implicit" as "associative", then it depends on how you define "associative". In any case, "associations" is (possibly) only one among several contributors to task performance 13/n
• Speaking of “associations”, it is another great example of circularity. It is occasionally defined as “the underlying mental construct assessed by implicit measures” 14/n
This can’t be true, when (1) “implicit attitude” measures may correlate by zero, (2) performance on these measures is driven by independent contributor*s*, (3) the set of implicit measures is unconstrained, (4) except ... for “implicit”=indirect=agnostic to mental processes! 15/n
• In ISCr, the most popular task (IAT) is said to be a “test”; however, even their designers feel uneasy with the interpretation of their task outcome. This is apparent in the disclaimer posted on “Project Implicit” 16/n
• Remarkably, and I sincerely admire this note of caution, the “Project Implicit” site indicates that one cannot infer “prejudice” from IAT scores 17/n
• In ISCr, researchers may insist that “implicit=indirect” and bears no relation to mental processes, explaining that they “find themselves occasionally lapsing to use implicit and explicit as if they had conceptual meaning” 18/n
… Yet, they may contradict their own recommendation in their next publications, implying (unsubstantiated) mental inferences from the task outcomes 19/n
• ISCr can and should do better. ISCr is not just inspiring theories. It is also inspiring interventions 20/n
Thanks to @bkeithpayne, @FloridaKateR and @CalvinKLai for being hardcore “implicit attitude” researchers who've been willing to engage in these debates – and thanks too, to the many geographically closer colleagues and friends for the great discussions 21/21
