Bernard Williams, answered from the texts and cited to the page.
Utilitarianism cannot accommodate the integrity of the agent. The agent has projects and commitments that are constitutive of who they are; the demand to maximize aggregate utility treats these projects as merely one input among many to a calculation that does not fundamentally privilege them.12 George the chemist, asked to take a job in chemical and biological weapons research so that someone less squeamish does not get it, is not aided by the utilitarian calculation that licenses his taking the job; the calculation tells him to abandon what makes his life his.
Jim, asked by the captain Pedro to shoot one Indian to spare nineteen others, is not aided by the utilitarian verdict that he should shoot; the question is what kind of agent he becomes by being the one who pulled the trigger. The integrity objection is not the claim that utilitarianism produces the wrong verdicts. It is the deeper claim that utilitarian reasoning, by treating the agent as a conduit for the calculation, misunderstands what it is to be a person who lives a life.
Sophisticated consequentialism (Hare's two-level theory, indirect utilitarianism, agent-relative permissions added on) preserves utilitarian verdicts while trying to recover the integrity intuition; the recovery is not stable, because the underlying structure remains the calculation that the objection identifies as wrong.34
For, to take the extreme sort of case, how can a man, as a utilitarian agent, come to regard as one satisfaction among others, and a dispensable one, a project or attitude round which he has built his life, just because someone else's projects have so structured the causal scene that that is how the utilitarian sum comes out?Utilitarianism For and Against, p. 35
Here I want only to bring out the kind of treatment that Hare's model gives to these aggregated preferences. This process of correcting preferences (their idealization, as we might say) is appropriate to models in which all preferences become notionally one person's preferences, such as the World Agent interpretation of Ideal Observer theory.Ethics and the Limits of Philosophy, pp. 87–88
Some answers are needed to that, too - and answers which stop short of relapsing into the bad faith of supposing that one's responsibilities could be adequately characterized just by appeal to one's roles.Utilitarianism For and Against, pp. 30–31
It is important to make clear what the objection is. It is not an objection (how could it be?) to all styles of reflective or second-order thought, nor to the mere idea that such thoughts might be interwoven with other thoughts in the course of practice. Hare is of course right to say that a general in action can jump between tactical and strategic levels of thought Seven (MT 52).Philosophy as a Humanistic Discipline, p. 80
The Critique of Utilitarianism (1973) is one of the most-cited essays in twentieth-century moral philosophy and inaugurates the integrity-based critique of consequentialism. Hare's two-level theory and the broader indirect-utilitarian tradition were partly responses; Singer's Practical Ethics defended a more direct utilitarianism that the integrity objection explicitly targets. Scheffler's The Rejection of Consequentialism (1982) developed the agent-relative response. Effective altruism (MacAskill, Ord) is the contemporary popular descendant of the utilitarian tradition the objection was aimed at. The objection remains live; the field has divided between consequentialists who treat it as a difficulty to be answered within the theory and non-consequentialists who treat it as showing the theory cannot be made coherent.