Utilitarianism, Welfare, and Information

Utilitarianism is often criticized for making assumptions about welfare comparisons that are too strong to be plausible. Roughly speaking, it assumes that all goods can be precisely measured and compared. This criticism applies both to classical utilitarianism, and to Harsanyi’s more sophisticated version.

However, joint work with Kalle Mikkola and Teru Thomas provides a response. Our version of utilitarianism, a generalization of Harsanyi’s, has almost unlimited flexibility when it comes to welfare comparisons. It allows for all kinds of incomparabilities. The post connects this flexibility with uncertainty.

Protected: Paradoxes of Decision PHIL2100 Seminar 3

There is no excerpt because this is a protected post.


Crudely put, utilitarianism is the thesis that one world is better than another if it contains greater total welfare. This theory of distribution has been much criticised. But it received a major boost in Harsanyi’s 1955 utilitarian theorem.

Harsanyi’s theorem seems to mainly be about expected utility theory. But this makes it subject to versions of some of the traditional criticisms, particularly in the assumptions it makes about welfare comparisons. But joint work with Kalle Mikkola and Teru Thomas provides a reply.

Continuity and Completeness

Expected utility theory has three main axioms: completeness, continuity, and independence. Completeness is dubious in many normative settings, so what happens when completeness is dropped? This post, based on joint work with Kalle Mikkola, explains that there is a surprising difficulty.

Moral philosophers often reject completeness because they think that some goods are not fully comparable. But they still carry on making continuity arguments. They should be cautious about this.


Ethics involves complicated problems that overlap with formal epistemology, philosophy of probability, decision theory, and game theory. No one could deny the relevance of mathematics to those subjects, so it would be remarkable if mathematical methods were not central to the study of ethics.

My own work especially involves axiomatic methods. Ordinary language is not well suited to the precise statement of axioms, and it is almost impossible to work out the consequences of even small sets of simple axioms without some mathematics.