Act vs Rule Utilitarianism

Home » Blog » Philosophy »Act vs Rule Utilitarianism
Act vs Rule Utilitarianism: A Critique of Harsanyi’s Assessment

Act-utilitarianism states that we should always perform the act that maximizes the total sum of wellbeing. This seems, prima facie, the best way to maximize the total sum of wellbeing. However, John Harsanyi (1908), Nobel laureates in Economics, denies so and argues for rule-utilitarianism according to which we should always follow the rules, such that if everyone follows them, the total sum of wellbeing will be maximized. He contends that rule-utilitarianism leads to much greater total sum of wellbeing, and therefore is correct. My doubts for Harsanyi are threefold. First, his assessment is biased towards rule-utiltarianism. Second, he does not take into account sophisticated act-utilitarianism. Third, his inference involves a controversial assumption in decision theory.

Harsanyi’s first claim is that a society of act-utilitarians is unable to produce optimal outcomes through coordination, whereas a society of rule-utilitarians can. Harsanyi (1908) asks us to imagine two possible worlds: an act-utilitarian world in which it is common knowledge that everyone is an act-utilitarian, and a rule-utilitarian world in which it is common knowledge that everyone is a rule-utilitarian. Common knowledge [that X] means that everyone knows that X, everyone knows everyone else knows that X, etc. In other words, each utilitarian agent has to regard the other utilitarian agent’s strategies as given.

Consider the following scenario (Harsanyi 1908): 1000 voters have to decide the fate of an important policy, which needs 800 affirmative votes. They all support the policy, but going to vote will incur minor costs. Communication is forbidden, and they cannot find out whether others have voted or will vote. In the act-utilitarian world, a voter will vote only when she is fairly sure that exactly 799 other voters will vote, which is highly unlikely. Imagine a voter k, who doubts that everyone will vote, estimates that the probability of any one of the voters will vote is 99%. This means that k is most unlikely to vote because the probability that exactly 799 will vote is almost zero (99/100^799). In contrast, for the rule-utilitarian world, there are only two choices—“everyone votes” and “everyone not votes”. All agents will choose to follow the rule that “everyone votes” because this rule will yield an exceedingly higher total sum of wellbeing. Harsanyi concludes that the rule-utilitarian world is vastly superior to the act-utilitarian world in maximizing the total sum of wellbeing. Although agents in the rule-utilitarian world is, in a sense, acting irrationally—if you are certain that everyone else will vote, from the point of view of maximizing the total sum of wellbeing, it is irrational to vote—Haryanvi maintains when an individual commits himself to a rule-utilitarian strategy, “he is making a rational commitments because by doing so he will contribute to achieving a higher level of social utility” (1908, pp.124). In other words, rule-utilitarianism is correct because it is good that everyone is committed to rule-utilitarianism.

However, the voting scenario’s setup seems unfair to act-utilitarianism. Since everyone knows everyone is a rule-utilitarian in a rule-utilitarian world, every agent will definitely choose to follow the rule that “everyone vote”, which creates a far better outcome. This seems to assume a rather high level of common knowledge, which favours rule-utiltarianism. But in reality, such common knowledge is very difficult to achieve. On the other hand, the voters are deprived of the knowledge of other’s voting behavior. In reality, however, there is usually good evidence whether the vote is going to be tight or not. If act-utilitarians have access to this knowledge, it will be much more likely to result in a good outcome. Also, in Harsanyi’s act-utilitarian world, agents will only vote when exactly 799 voters vote—“if more of them vote then his vote will not be needed” (120). In reality, however, a sophisticated utilitarian will understand that it is highly improbably that she can always achieve an optimal outcome. In the face of uncertainty, she often has to play safe and satisfy with a good enough consequence (Jackson 1991, pp.468). To ensure a socially important policy can pass, a sophisticated utilitarian may still go to vote, even if she believes that the policy will probably (but not certainly) pass without her vote.

Harsanyi’s second claim is that many important things will be lost in a society of act-utilitarians, and hence, it fails to maximize the total sum of wellbeing. People will break a promise or take other’s property whenever it promotes the social good. Without trust, the society will lose the benefit of the practice of promising, such as the ability to form reasonable expectations about other’s future behavior (which will create a sense of security and make it easier for us to plan for our future), and the opportunity to increase social good through coordination. Without property right, our incentives to engage in socially desirable activity will be greatly reduced. Moreover, special duty—including parent-children relationship—does not have a place in a society of act-utilitarians. However, such relationships have strong informational and motivational advantage, allowing the agents to be aware of other’s special need and develop emotional tie. On the other hand, rule-utilitarianism is able to acknowledge these beneficial effects by suggesting that they correspond to rules such that if everyone follows them, sum of wellbeing will be maximized. In other words, we should generally keep promises, respect property right, and look after our own children according to rule-utilitarianism. Again, rule-utilitarianism performs better in utilitarian terms.

However, Harsanyi is erroneous in assuming that act-utilitarians only follow a simple and naïve model of decision-making. As a matter of fact, sophisticated act-utilitarians can actually take the aforementioned benefits into account. In a world of sophisticated act-utilitarians, the agent will not easily break a promise because they realize that their interactions with others are often continuous rather than one-off, and these interactions will become more profitable if trust is involved. In other words, there is a good act-utilitarian reason to generally keep our promises. The exact same reason can be applied to property rights. There are also good act-utilitarian reasons to keep special relationship: epistemologically speaking, since you know people who are nearest and dearest to you much better, it is easier to benefit them, whereas it is much harder to benefit a stranger; motivationally speaking, as a matter of evolution, we generally perform much better when we have close tie with people (Jackson, 1991). Failing to take into account sophisticated versions of act-utilitarianism, Harsanyi is unsuccessful in showing that rule-utilitarianism performs better than act-utilitarianism.

My last worry for Harsanyi is that both of his arguments involve the following inference: it is good that everyone is committed to rule-utilitarianism; therefore rule-utilitarianism is correct, and we should act in accordance to rule-utilitarianism. But this is controversial. Imagine a billionaire will pay you one million dollar—all you have to do is to commit, at tonight, to drink a toxin, and you can change your mind anytime (Kavka 1983). The standard conclusion, in modern decision theory, is that it will be good for you to have the commitment tonight to drink the toxin, but it will be irrational to actually drink it. In other words, even though it is good to have certain commitments, is does not necessarily imply that you should act on them as well. Harsanyi owes a further account on why we should be committed to rule-utilitarianism in practice.

To conclude, Harsanyi’s comparison between act-utilitarianism and rule-utilitarianism is unfair. His argument also involves a doubtful inference.


Jackson, F. ‘Decision-theoretic consequentialism and the nearest and dearest objection.’ Ethics 101 (1991): 461–482.

Kavka, G. ‘The toxin puzzle.’ Analysis 43 (1983): 33–6.

Harsanyi, J. ‘Rule utilitarianism, rights, obligations and the theory of rational behaior.’ Theory and Decision 12(2) (1980).

Further Readings

Is AI a Threat to Humanity? 4 Facts You Should Know

Does the Use of Enhancement Drug Diminish Our Agency?

Jumpstart Your Productivity with These Hacks

Blog Search