Claims that deontology may very well be a confabulation are notorious, but there is loose talk around them. Clarification work needs to be done.
Greene (2008) proposes two conditions for confabulation:
- find a factor that predicts ones judgments
- the factor that predicts ones judgments is not plausibly related to the factors that are believed to be the bases of those respective judgments.
To illustrate this, he imagines a case relating to romantic choices. Alice goes on many dates and evaluates the people she likes as brilliant, kind, charming, and the ones she does not like as self-absorbed. At the same time, those who were evaluated positively are exceptionally tall, while the ones rejected are less than six-foot-four. As it turns out height is a near-perfect predictor of Alice’s preferences. However, she believes that her romantic choices are based on personality traits. Thus, Alice’s talk about personality traits is a confabulation.
The first condition is straightforward, as one just needs to identify the influencing factors in controlled experiments. The second is not. How we should understand the idea that the factors are not plausibly related? It is clear that height is not related to personality traits. But it does not follow from this alone that the factors that Alice believes are the bases of her judgments are actually a mere rationalization. What follows is that two kinds of factors are the bases of her judgments, and that she invokes only one, being unaware of additional influencing factors. This is not quite a mere rationalization. The second condition needs to prove that the “official” factors do not have a significant enough causal contribution to be a predictive factor, contrary to what a confabulator believes. Therefore, “plausibly unrelated” must refer to a relation of causal influence on belief or action. This understanding is in line with the first condition which requires identifying a factor that is the main causal influence and with the psychological characteristics the confabulator’s factors do not drive ex ante his judgments. The model can be stated as:
(1) Alice believes that plausible factors X, Y, Z preferentially support her romantic choices.
(2) Alice’s romantic choices are preferentially supported by W.
(3) Alice is unaware and does not believe that W drives her choices.
(4) X, Y, Z do not preferentially support Alice’s romantic choice.
(5) Therefore, Alice’s citation of factors X, Y, Z is a confabulation.
Greene suggests that something similar to Alice’s thought process reflects the way deontological philosophy is done. Deontologists seem to be confident that certain factors, such as thoughts of duties and rights, generate deontological judgments, but in fact something else is decisive to issuing these judgments. Now, we get the following:
(1) Deontologists believe that plausible factors X, Y, Z preferentially support characteristic deontological (CD) judgments.
(2) Characteristic deontological judgments are preferentially supported by W.
(3) Deontologists are unaware and do not believe that W drives CD judgments.
(4) X, Y, Z do not preferentially support CD judgments.
(5) Therefore, deontologists’ citation of X, Y, Z is a confabulation.
I will call this the neutral confabulation model (NC) because the two conditions do not stipulate the nature of the predictive factor W. But this is not the relevant model. Greene insists that the existence of a certain type of moral emotions in conjunction with confabulatory tendencies gives rise to deontological philosophy. He draws the distinction between currency emotions, specific to consequentialism, and alarm-like emotions, specific to deontology. Currency emotions are relevant input in a weighing process, whereas alarm like emotions are highly salient, blunt, simple and forces one to issue strong commands (“Don’t do it!”, “Must do it!”) We now get what I will call the alarm-like emotion based confabulation model (AEC), in which W is qualified as an alarm-like affective factor in premise (2).
One famous example of moral confabulation which seems to fit the AEC model is the incest experiment. Haidt and colleagues (2000) devised scenarios to elicit responses that involve two separate processes: an averse emotional reaction followed by a post hoc justification. In the incest story, Julie and Mark (brother and sister) decide to make love. Julie was already taking birth control pills and Mark used a condom. After hearing the story most subjects immediately judged the action to be wrong, and began to justify their response by pointing out psychological and inbreeding risks.
The incest story elicits a judgement of moral condemnation which is driven by an alarm-like emotion such as disgust. However, subjects immediately abandon their justifications when features from the story are reinforced. They point out psychological and inbreeding risks, but after the experimenter reiterates the use of birth control, and the fact the story presupposes no psychological harm, they admit failure of justification. It is highly implausible that these results mirror the way deontological philosophy is done. It is immediately transparent that the reasoning involved is badly done. Subjects are neither confident about their justifications, nor do they provide at least prima facie plausible reasons. When retracting the proposed justifications, subjects say something like “I do not know, I can’t explain it, I just know it’s wrong.” This is indeed a confabulation driven by alarm-like emotions, but one that ultimately is not endorsed by the subjects themselves, and lacks a minimal plausibility.