Implicit associations are discerned in experimental settings, whereby the differential speed of pairing items (such as black faces and white faces with positive vs negative terms) is taken to indicate the strength of associations held. Experimental evidence indicates that these implicit associations manifest in other, more worrying contexts: the differential evaluation of CVs with the same (identical!) qualifications but from different genders; the differential hiring recommendations for equally moderately qualified black and white job applicants; the greater readiness to identify as a gun an indeterminate object when in a black, rather than white, man's hand (for an overview, see Jost et al. 2009).
One general question is what the role of these implicit cognitions in individual agency is. Not all implicit cognitions are prima facie problematic in the ways that those described above are. Can they be epistemically innocent, or even valuable? How should we model agency given the role of implicit, as well as explicit processes in the production of action?
A specific question I have looked at in some detail (Holroyd 2012) concerns whether individuals should be held responsible for being influenced by morally problematic implicit biases. I identify a range of considerations that have been appealed to in support of the idea that individuals are not morally responsible for being influenced by such implicit associations:
i) that the agent is not causally responsible for the presence of such associations;
ii) that individuals lack the relevant kind of control over their operation;
iii) that individuals are unaware of their operation;
iv) that implicit biases are not responsive to reasons.
I have argued that with respect to each of these claims, there is reason to hold either that the condition posited as necessary for moral responsibility should not be accepted (i and ii), or that empirical evidence indicates that the condition is at least sometimes met (iii and iv).
For example, an argument such as the following, for exempting individuals from responsibility for being influenced by bias, might be offered (for versions of this argument see Saul, Levy 2012):
(i) Individuals cannot be held responsible for cognitive processes or influences on behaviour and judgment over which they do not have control.
(ii) Manifesting—being influenced in behavior and judgment by—implicit biases is not under an agent’s control.
(iii) Therefore, individuals cannot be held responsible for the influence of implicit biases on behavior and judgment.
One strategy for evaluating this argument is to consider exactly what sense of control is at issue (direct control? rational control? ability to inhibit or intervene? In a work in progress, Dan Kelly and I evaluate different versions of the control argument, and argue that ecological control can be sufficient for responsibility for implicit biases).
In my 2012 paper, I argue that it is plausible to suppose that indirect control is, at least sometimes, the relevant sense of control for responsibility, and that on this interpretation, premise ii will be false. There is some evidence that suggests that for some implicit associations (race and negative/positive word associations), the extent to which it influences action is correlated with the individual's explicit beliefs about the importance of non-prejudiced behaviour or goals to treat people fairly (see Devine et al 2002, Moskowitz & Li 2011). We might exert control over the manifestation of implicit biases indirectly, then, via change in our beliefs and goals. Insofar as individuals may have indirect control in this way, then it is not obvious that there are grounds related to lack of control for not holding individuals responsible for implicit biases.
In future work I aim to look in more detail at the kinds of control we might have over implicit cognitions, whether implicit cognitions differ from each other in important ways (in progress, with Joseph Sweetman) and the role of holding each other responsible in regulating the expression of implicit biases.