Empirical evidence for moral Bayesianism

ABSTRACT

Many philosophers in the field of meta-ethics believe that rational degrees of confidence in moral judgments should have a probabilistic structure, in the same way as do rational degrees of belief. The current paper examines this position, termed “moral Bayesianism,” from an empirical point of view. To this end, we assessed the extent to which degrees of moral judgments obey the third axiom of the probability calculus, $\mathrm{if}\mathrm{P}\left(\mathrm{A}\cap \mathrm{B}\right)=0\mathrm{then}\mathrm{P}\left(\mathrm{A}\cup \mathrm{B}\right)=\mathrm{P}\left(\mathrm{A}\right)+\mathrm{P}\left(\mathrm{B}\right)$, known as finite additivity, as compared to degrees of beliefs on the one hand and degrees of desires on the other. Results generally converged to show that degrees of moral judgment are more similar to degrees of belief than to degrees of desire in this respect. This supports the adoption of a Bayesian approach to the study of moral judgments. To further support moral Bayesianism, we also demonstrated its predictive power. Finally, we discuss the relevancy of our results to the meta-ethical debate between moral cognitivists and moral non-cognitivists.

KEYWORDS:

• moral judgments
• meta-ethics
• degrees of belief
• moral Bayesianism
• desire as belief thesis
• moral cognitivism
• moral non-cognitivism

Acknowledgments

We would like to thank Brian Weatherson, David Enoch, Itzhak Aharon, Jonathan Najenson, Ryan Doody, Thomas Polzler, and three anonymous reviewers for their thoughtful comments and suggestions on this article. Earlier versions of this paper were presented in the annual meeting of the European Society for Philosophy and Psychology, Hatfield, England (2017) and in Tel hai Seminar, Department of Philosophy, Tel-Hai College, Israel (2018). We thank the participants of these events for useful discussions. This work was funded by The Israel Science Foundation grants (grants 1471/20) to Anat Maril, by The Israel Science Foundation grant (grants 1042/13) to Ittay Nissan-Rozen, and by a fellowship from the Sidney M. Edelstein Center for History and Philosophy of Technology and Medicine at the Hebrew University of Jerusalem to Haim Cohen.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/09515089.2022.2096430.

Notes

1. Smith referred also to a third type of degree associated with moral judgments, their degree of stability over time. Smith called this type of degree “robustness.” This third type of degree will not play a role in this paper.

2. Different non-cognitivist accounts point to different types of attitudes. See, Van Roojen (2018) for a good review.

3. Alternatively, the non-cognitivist can argue that most people, most of the time, fail to obey the normative ideal. Although in some contexts this is certainly the right move to take, we believe it is uncontroversial that, other things being equal, one should aim to avoid such conclusions.

4. This is an implicit assumption, for example, in almost all discussions of decision-making under normative uncertainty (see for example, Lockhart, 2000; MacAskill et al., 2020; Riedener, 2021).

5. As an anonymous referee commented, however, it seems unlikely that most moral psychologists would deny that people can be more or less confident in their moral judgments. A more plausible view would be that – much as many philosophers believe is the case with respect to non-moral beliefs – moral judgment may take either a binary (accept/reject) or a graded form. However, accepting that moral judgments may come in degrees of confidence does not imply accepting that these degrees have a probabilistic structure. This latter claim is what we examine in this paper.

6. The persistent disagreement among philosophers regarding the status of moral judgments may be taken to suggest that holding incorrect meta-ethical views regarding one’s own meta-ethics (as reflected in one’s behavior) is a very common phenomenon.

7. To limit the duration of the online questionnaires to 10 minutes, we limited each to a maximum of 26 questions.

8. An anonymous referee has raised the worry that our results might reflect responses that are trivial in some sense. The reviewer suggested that this might be the case with respect to some of the desire reports, which might reflect complete indifference with respect to the dilemmas (thus, leading the participants to report degrees very close to 50%). A related worry might be that, when it comes to beliefs, participants might report degrees close to 100% and 0%. To address these worries we checked the percentages of participants who reported degrees between 45% to 55% (6% for beliefs and 2% for desires) and the percentages of participants who reported degrees of either 100% or 0% (9% for beliefs and 15% for desires) and established that, even if the analysis is limited to the range of values that do not fall into one of these two categories, we obtain the same pattern of result as reported above.

9. Participants classified dilemmas as either moral or non-moral dilemmas. They were asked to rate each dilemma on a 7-point scale ranging from 1 (does not require a moral judgment) to 7 (requires a moral judgment). Finally, we defined a dilemma as belonging to the moral domain if more than 70% of participants scored it 5 or higher.

10. Level of importance was defined as follows: “High” for responses greater than 60, “Low” for responses lower than 40, and “Neutral” for responses between 40 and 60.

11. An anonymous referee has raised the worry that participants may have interpreted the word “appropriate” in a non-moral way (e.g., as referring to social conventions, or legal status). We do not believe this is the case, for three reasons. First, the questions were posed to the participants in Hebrew. The exact formulation was “עד כמה ראוי בעיניך ש … ?”. The Hebrew word “ראוי” bears a distinctive moral connotation lacking in the English “appropriate” (we debated whether to translate it here as “morally appropriate” but chose to stick with a verbatim translation). Second, in the Hebrew formulation there is an emphasis on one’s personal assessment. The verbatim translation of the Hebrew question is: “how much appropriate, in your eyes is … ?” Third, as explained above, the moral dilemmas were classified as such by participants in a pilot experiment.

12. Following Note 8, here too, we have checked for “trivial” responses (i.e., responses around 50%, 100%, and 0%). Responses around 0% or 100%, were observed for beliefs in 8% of the responses, for desires in 13%, and moral judgments in 19%. Responses around 50% were observed for beliefs in 5% of the responses, for desires in 5% and for moral judgments in 5%. Even if the analysis is limited to the range of values that do not fall into one of these categories, we get the same result-pattern as reported above.

13. Cases in which the degree of belief in either a given proposition, x, or the proposition “x is morally appropriate,” are equal to 1 or 0.

14. Importantly, both internalists and externalists about moral motivation can, should and usually do accept this claim. The difference between internalists and externalists about moral motivation is that the former, but not the latter, take the connection between moral judgments and moral motivation to be necessary. However, externalists too may, and usually do, accept that typically (but not necessarily) moral judgments restrict moral motivation. In the context of the DBT, both internalists and externalists should accept it as a descriptive thesis that usually holds, but only internalists should be troubled by Lewis’ result according to which it cannot always hold for a rational agent.

15. Participants classified dilemmas as either moral dilemmas or non-moral dilemmas. They were asked to rate each dilemma on a 7-point scale ranging from 1 (does not require a moral judgment) to 7 (requires a moral judgment). Finally, we defined a dilemma as belonging to the moral domain if more than 70% of participants scored it between 5 and 7.

16. Following Note 8, here too we have checked for “trivial” responses (i.e., responses around 50%, 100% and 0%). Responses around 0% or 100% were observed for beliefs in 8% of the responses, for desires in 17% and for moral judgments in 16%. Responses around 50%, observed for beliefs in 3% of the responses, for desires in 6% and for moral judgments in 6%. Even if the analysis is limited to the range of values that do not fall into one of these categories, we get the same result pattern as reported above.

Additional information

Funding

This work was funded by The Israel Science Foundation grants (grants 1471/20) to Anat Maril, by The Israel Science Foundation grant (grants 1042/13) to Ittay Nissan-Rozen, and by a fellowship from the Sidney M. Edelstein Center for History and Philosophy of Technology and Medicine at the Hebrew University of Jerusalem to Haim Cohen.

Notes on contributors

Ittay Nissan-Rozen

Ittay Nissan-Rozen is a senior lecturer at the philosophy department and the PPE (Philosophy, Economics and Political Science) program at the Hebrew University of Jerusalem. His areas of expertiese are the philosophy of social science, normative decision theory, formal epistemology and formal ethics. He works mostly on questions related to uncertainty and the role it plays , and should play in our everydays lives, in science and in public policy.

Anat Maril

Anat Maril is a Prof at the Hebrew University of Jerusalem, departments of cognitive science and psychology. Her research focuses on understanding knowledge structures in the human mind. Haim Cohen, a PhD candidate at the Hebrew University of Jerusalem, The Federmann Center for the Study of Rationality, and the department of cognitive science. This paper is related to his thesis research on moral judgments, beliefs, and desires, written under his thesis advisors, Prof Anat Maril and Prof Ittay Nissan-Rozen