Some years ago a friend sent me a paper that has stayed with me, though not for the reasons its authors might have hoped. In The Return on Investment in AI Ethics, Bevilacqua, Berente, Domin, Goehring, and Rossi propose a holistic return on ethics framework for organisations that wish to quantify what they get back when they spend on behaving well. It is a serious piece of work, carefully argued and admirably candid about the difficulty of its own enterprise. I recommend it. I also think it answers the wrong question—or rather, that the question it answers so diligently is one no manager or board member in a healthy organisation should feel the need to ask.
The difficulty it addresses is real enough. Traditional return on investment is calculated on revenue and cost, on the things that present themselves obligingly for counting. The returns to ethical conduct—reputation, trust, the loyalty of staff who would rather work for a firm they need not be ashamed of—are intangible, indirect, and slow. They resist the spreadsheet. The literature on measuring them is now considerable, and it offers an expanding toolkit: balanced scorecards, sentiment analysis, ESG indices, longitudinal impact assessment. My contention is that this toolkit is not a solution to a measurement problem. It is the symptom of a moral one. The cynic, in Wilde's enduring definition, is the person who knows the price of everything and the value of nothing. An organisation that has learned to price its integrity has, in the same motion, begun to forget what it was for.
The Question Behind the Question
When a board commissions a business case for ethical investment, it has already settled a prior question without noticing it has done so. To ask what ethics returns is to assume that its warrant lies in the return—that conduct is justified by its consequences for the share price, the engagement survey, the brand-tracking study. This is the logic of the proxy, and it is the master pathology of contemporary management—one I have explored before, in the board that reaches for the charismatic chief executive as a legible stand-in for the illegible work of institutional repair, and in the romance of leadership that prefers the satisfying narrative of individual agency to the unsatisfying analysis of systems. In each case the organisation substitutes the question it can answer for the one it should. The business case for virtue is the same move performed on ethics: it converts "is this right?"—which admits no clean metric—into "does this pay?", which obligingly does.
We already have a vocabulary that enables us to see what is lost in the exchange by distinguishing the goods internal to a practice—those one can obtain only by engaging in it on its own terms—from the external goods of money, status, and reputation that the practice may also, contingently, attract. When managers and NEDS come to value products and services solely for the external goods they yield, the ethical tends to hollow, however prosperous it appears. Ethical conduct justified entirely by its reputational dividend is a practice in precisely this condition. It is being held, not believed.
Aristotle saw the methodological error two and a half thousand years ago. It is the mark of an educated person, he observed, to look for only so much precision in each subject as its nature admits; we should no more accept mere probability from a mathematician than demand demonstration from an orator. Ethics is the subject that least admits the precision the business case demands. To insist on quantifying it is not rigour but its opposite—a category error dressed in the costume of seriousness.
What Counting Does to the Counted
Suppose we set the philosophical objection aside and grant the board its business case. A harder problem awaits. Measurement is not the neutral mirror its advocates often imagine. It is reactive: the act of measuring alters the thing measured. A study by Wendy Espeland and Michael Sauder showed how law schools, once ranked, reorganised themselves around the criteria of the ranking, gaming what they could and quietly abandoning what the ranking ignored. This process has been called commensuration—the transformation of distinct qualities into a common metric—and it does not so much record value as remake it, discarding whatever will not reduce to the chosen unit.
The organisation that demands a metric for its ethics is thus advertising a prior failure—of trust in its own people to know right from wrong without a dashboard to confirm it.