Skip to content

Sinnott-Armstrong on Implicit Moral Attitudes

On October 30th, Professor Walter Sinnott-Armstrong of Duke University gave the 2014 Wellcome Lecture in Neuroethics.  His talk, “Implicit Moral Attitudes”, concerned the practical and theoretical implications of recent empirical research into unconscious or sub-conscious beliefs or associations.  Recordings of his talk will be made available soon Audio recording of the talk is available here: those interested that were unable to attend, I will summarise the main points of Sinnott-Armstrong’s talk and some of the discussion that occurred during the Q&A afterwards.

Evaluating people’s moral attitudes is both practically and theoretically important, but faces crucial difficulties.  To understand people’s moral attitudes and determine praise or blame, we have traditionally had to rely (at least in part) on reports.  This is particularly difficult, though, when confronting psychopaths: are they being honest when they make moral claims?  Mere reports are unreliable, as psychopaths are prone to deceive.  More usual cases of convicted criminals present similar problems: how can we know whether a criminal is rehabilitated, or will reoffend?  Given the incentive to lie, reports are unreliable – and behavioural assessment may only be of limited utility.

More theoretical issues have plagued philosophers for centuries.  Moral semantics inquires into whether people make cognitive, truth-conducive claims or non-cognitive expressions of attitudes.  Motivational internalists argue there is a tight link between moral belief and motivation to act, while others deny it.  Moral epistemology deals, in part, with whether our moral judgments are justified, reliable, knowledgeable, etc. And questions of moral responsibility consider whether factors like moral knowledge are necessary for us to legitimately blame agents.

Sinnott-Armstrong argues that, while the notion of ‘attitude’ is typically unclear, empirical analysis of implicit moral attitudes can shed light on all the above debates.  This requires a good tool for measurement of such moral attitudes in agents – one that does not rely on mere reports of judgment.  Several tests have been proposed and tested, including the Implicit Association Test and eye-tracking tests; Sinnott-Armstrong favours what is known as the Process Dissociation Procedure (PDP) test. 

The PDP test (in present context of moral judgment) classifies a set of uncontroversially ‘immoral’ words like murder, steal, and lie; and a set of ‘neutral’ words like walk, bake or swim.  Subjects see a ‘prime’ target word flash, then a ‘target’ word, and are asked to judge whether the target word reflects something morally wrong.  ‘Moral bias’ occurs when a moral word prime makes people unreliable in judging the morality of the subsequent target neutral word.  Though we normally think of bias as a bad thing, this is one case where it is desirable*; the bias originates from strong negative attitudes towards the immoral behaviour.  The PDP then can function as a measurement of the strength of those negative attitudes.  This test (as with other tests of implicit attitudes) is not of purely theoretical interest, but can predict behaviour as well.  According to Sinnott-Armstrong, a modified version of the test (with words relating to same-sex marriage used as primers) was able to predict voting behaviour on a North Carolina referendum against same-sex marriage.

If the PDP test and other implicit attitude tests function well as predictors of behaviour, Sinnott-Armstrong believes that (when suitably developed) they could be deployed in various legal contexts.  Currently, parole boards use crude measures of prisoners’ behaviour and first-personal reports to determine likelihood of reoffense; information on implicit attitudes from these tests could help those boards make more reliable assessment of recidivism chance.  The tests could also be used to evaluate the effects of rehabilitative treatments, insofar as ‘moral bias’ is affected; and we might be able to measure the effect of other interventions like prison on inmates’ implicit moral attitudes – if they weaken one’s moral attitudes, this may be a strike against current incarceration models.

On the philosophical side, Sinnott-Armstrong suggests some revisions to current debates.  Implicit attitudes and expressed judgments clearly come apart (e.g., someone claims to despise racism but evinces implicit racial bias); but given that non-cognitivist models of moral expression seem closer to implicit attitudes, this might provide some evidence for cognitivism when it comes to judgments themselves.  Implicit attitudes are linked strongly to behaviour, suggesting some form of motivational internalism is plausible.  Implicit attitudes are also apparently less susceptible to framing effects than express judgments, perhaps making them more epistemically reliable.  And moral responsibility may have more to do with one’s implicit attitudes than express claims about morality; implicit attitude tests could then inform the degree to which someone is responsible for their acts.

Sinnott-Armstrong’s talk was thought-provoking, and a number of interesting questions were raised by the audience.  Here is a sample, along with a summary of Sinnott-Armstrong’s replies:

  • Doesn’t this suggest moral appreciation centrally involves a form of implicit (moral) bias?  Reply: Yes, but it’s complicated – the bias might take more or less pernicious forms depending on the context.
  • What the grounds for the ‘list’ of purportedly moral behaviour?  Reply:  The list comes from areas of 90% agreement on morality (validated in a prior study).  For present purposes, Sinnott-Armstrong remains neutral on the ultimate grounds for those claims – but given their uncontroversial nature, you should be able to help yourself to your favourite story.
  • Doesn’t this overlook the effect of social context on judgment/behaviour?  Reply:  The account is compatible with alternative explanations of judgments that do not analyze implicit associations; the data are meant to supplement, rather than supplant, such sociological investigation.
  • Do these tests really predict behaviour, and not just reflect reported behaviour? Reply:  Other studies have indeed found true behavioural correlates; for example, implicit bias tests predict how ER doctors triage.

Overall, Sinnott-Armstrong agrees there is much more research to be done on the topic of implicit attitudes.  The hope is that such research can meaningfully impact a variety of policy decisions as well as philosophical debates.


*    At least at moderate levels.  A companion seminar by Sinnott-Armstrong drew attention to the phenomenon of scrupulosity, which is a form of obsessive-compulsive disorder where one is inordinately obsessed with moral and/or religious norms.

Share on