The costs of being consequentialist: Social inference from instrumental harm and impartial beneficence

Jim A.C. Everett, Nadira S. Faber, Julian Savulescu, Molly J. Crockett
Journal of Experimental Social Psychology (79), pp200-216

Previous work has demonstrated that people are more likely to trust “deontological” agents who reject harming one person to save many others than “consequentialist” agents who endorse such instrumental harms, which could explain the higher prevalence of non-consequentialist moral intuitions. Yet consequentialism involves endorsing not just instrumental harm, but also impartial beneficence, treating the well-being of every individual as equally important. In four studies (totalN = 2,086), we investigated preferences for consequentialist vs. non-consequentialist social partners endorsing instrumental harm or impartial beneficence and examined how such preferences varied across different types of social relationships. Our results demonstrate robust preferences for non-consequentialist over consequentialist agents in the domain of instrumental harm, and weaker – but still evident – preferences in the domain of impartial beneficence. In the domain of instrumental harm, non-consequentialist agents were consistently viewed as more moral and trustworthy, preferred for a range of social roles, and entrusted with more money in economic exchanges. In the domain of impartial beneficence, preferences for non-consequentialist agents were observed for close interpersonal relationships requiring direct interaction (friend, spouse) but not for more distant roles with little-to-no personal interaction (political leader). Collectively our findings demonstrate that preferences for non-consequentialist agents are sensitive to the different dimensions of consequentialist thinking and the relational context.

Overview of Paper


Much work over the last decade has focused on the psychological processes underlying judgments about whether it is moral to sacrifice one innocent person to save a greater number of people. In recent years, befitting the fundamentally social role of moral judgments, researchers have begun to consider the social consequences of these judgments: how are consequentialist individuals who endorse harming for the greater good perceived? Previous work has shown that the judgment a person makes in these sacrificial dilemmas influences how moral, warm, and competent they are perceived to be, and even how much cooperation is extended towards them in economic games (e.g. Bostyn & Roets, 2017; Everett et al., 2016; Rom et al., 2017; Uhlmann et al., 2013). But while this work has focused solely on judgments in sacrificial dilemmas and has been thought to shed light on the social consequences of consequentialist judgments in general, consequentialism involves much more than just judgments about whether to sacrifice one to save a greater number.

As outlined in the two-dimensional model of utilitarian psychology (Kahane et al., 2018, 2015), judgments in sacrificial dilemmas tap the endorsement of “instrumental harm”, which can be theoretically and empirically distinguished from impartiality dilemmas that tap the more positive, impartial welfare-maximizing dimension (“impartial beneficence”) of consequentialist theories and consequentialist tendencies in ordinary people. Because this previous work on perception of consequentialist agents has focused almost exclusively on sacrificial dilemmas, it has remained unknown whether the preference for non-consequentialists over consequentialists operates similarly in impartiality dilemmas in which someone faces the decision to help someone close to them or a greater number of strangers. In this paper, we sought to fill this gap.


We conduct four studies in which we investigated social perceptions of non-consequentialist and consequentialist agents in both sacrificial dilemmas tapping endorsement of instrumental harm, and impartiality dilemmas tapping endorsement of impartial beneficence. As well as theoretically extending the conceptual space in which non-consequentialists might be preferred, we also investigate this across a much greater range of dependent measures than has previously been used. Specifically, we study partner preference by looking at two different economic games (the Trust Game and the Prisoner’s Dilemma); several distinct dimensions along which the agent’s character could be perceived (warmth; competence; morality); the different social roles in which the agent would be preferred (as a friend, a spouse, a boss, and as a political leader); and the different processes or motivations perceived to influence the agent’s moral decision (reason vs. emotion; strategic considerations; altruistic motivations).

For Studies 2-4, our design, hypotheses, and analysis plan were all pre-registered, and all data, experimental materials, and analysis scripts are available at the Open Science Framework



In the domain of sacrificial harm, our findings strongly confirm previous work highlighting the cost of being consequentialist: the non-consequentialist was seen as more moral, loyal, warm, and competent; expected to be a better friend, spouse, boss, and political leader; and was cooperated with more in a Trust Game and Prisoner’s Dilemma.

The pattern of results in the impartiality dilemmas was much more nuanced. Simply counting overall results suggest that the non- consequentialist has the edge over the consequentialist in the domain of impartial beneficence too: overall, we saw few cases where the consequentialist was preferred, typically seeing either a null effect or that the non-consequentialist was favored. Impartial consequentialists were consistently disfavored for roles involving a direct interpersonal relationship, even if they were not explicitly rated as being deficient in morality and warmth. In all four studies, the impartial consequentialist was thought to be less loyal and thought to make a worse friend and spouse, even if they were not always explicitly rated as being deficient in morality and warmth.

Character Ratings

Character ratings in Studies 2-4 as a function of agent judgment and dilemma type. Results show that in the sacrificial dilemmas, the non-consequentialist was consistently rated as higher in morality (A), loyalty (B), warmth (C), and competence (D). In the impartiality dilemmas, the non-consequentialist was seen as more loyal but was not consistently rated as more moral, warm, or competent. Error bars represent 95% confidence intervals.

Role Suitability

Role suitability in Studies 2-4 as a function of agent judgment and dilemma type. Results show that in the sacrificial dilemmas, the non-consequentialist was preferred for all four social roles. In the impartiality dilemmas, the non-consequentialist was consistently preferred as a friend (A) and spouse (B), but not a boss (C) or political leader (D). Error bars represent 95% confidence intervals.

Cooperation in Economic Games

Transfers in a Trust Game as a function of agent judgment and dilemma type. In the sacrificial dilemmas participants transferred more to a non-consequentialist than a consequentialist, but there was no difference in the impartiality dilemmas.

Cooperation in the Prisoner’s Dilemma in Study 3 as a function of agent judgment and participant judgment. Across both dilemmas, non-consequentialist participants were more likely to cooperate with the non-consequentialist agent than the consequentialist, but consequentialist participants cooperated equally with both agents.


When it comes to sacrificial dilemmas, We show in three studies that non-consequentialists were consistently preferred over consequentialists. We argue that this is perfectly explicable on a partner choice account of non-consequentialist moral intuitions (Everett et al., 2016). The consequentialist rejection of any constraints on the maximisation of welfare means that if killing one’s partner maximises the greater good, then that is what one should do. And yet when selecting a social partner for the purposes of continued cooperative exchange, such a person would seem disastrous. Instead, we have argued, it would be more advantageous for people to seek social partners who maximise good consequences, but also those who also exhibit respect for rights, duties, and the individuality of persons. Given that consequentialists are consistently disfavored in the cooperation market, partner choice mechanisms could explain why our moral intuitions in such sacrificial dilemmas so often lean deontological (Everett et al., 2016). Put simply, non- consequentialist, or deontological, judgments confer an adaptive function by increasing the likelihood of being chosen as a cooperation partner. It pays – in our studies, literally – to be a non- consequentialist in a sacrificial dilemma.

When it comes to impartiality dilemmas, we show impartial consequentialists were consistently disfavored for roles involving a direct interpersonal relationship, even if they were not explicitly rated as being deficient in morality and warmth. Tis makes sense given that consequentialism’s requirement for the impartial maximisation of welfare is often inconsistent with the nature of special relationships like friendship and familial duties that are a fundamental part of common- sense morality (Jeske, 2014; W.D. Ross, 1930). When we enter a close, interpersonal partnership – with a friend, or a team-mate, or a spouse – we expect partiality by agreeing on certain special obligations ourselves and expecting them to be honoured by our friend. We are expected to help our friend when they need it, and we expect our friend to help us when we need it, but consequentialism’s denial of any such obligations is incompatible with what we seek in social partners. This, we think, is why participants consistently rated the impartial consequentialist as being less loyal and as making a worse friend or spouse. In contrast, while it makes sense for non-consequentialists to be favored for direct, interpersonal relationships, it is much more reasonable – even preferable – to favor a consequentialist for distant, impersonal roles like a political leader, and this what we what found in Studies 1 and 2.