The Obedience Replication

Structure of the calculation. Act utilitarianism asks whether this act, on this occasion, maximises expected welfare across all affected parties. That means the participants (180, plus the confederates and the experimenters), the wider research community, and future students who may or may not trust psychology.

Benefits. Knowledge about obedience to authority, and specifically the value of resolving disputes about Milgram's data. But replication value is not automatic: if the original paradigm is confounded, reproducing the confound adds little, and the expected benefit should be discounted accordingly - consequentialists must reason about expected benefit, not hoped-for benefit.

Harms. 'Discomfort' is the wrong unit. The relevant states are distress at believing one has harmed a person, guilt, and - the documented risk in Milgram's own follow-ups - a durable revision of self-image. These are not evenly distributed: some participants will be unaffected, a minority may be significantly harmed, so a mean is misleading; utilitarians must attend to the tail.

Externalities. Trust is a commons. A pool of 180 gossip-linked undergraduates means that deception, once discovered, contaminates the source of future research - a real cost that standard calculations omit.

Verdict. On plausible numbers the calculation is negative, chiefly because cheaper designs exist. A virtual-reality version, or a lower maximum voltage, or an honest framing about 'authority and compliance', secures most of the knowledge at a fraction of the harm. A utilitarian who ignores available alternatives has not done the sum.

The formula of humanity. Kant requires that persons be treated as ends in themselves, never merely as means. Deception is the paradigm case of treating a rational agent as a means: the participant's capacity to deliberate about what she is joining is deliberately bypassed so that her compliance can be observed. Her rational agency is not merely unused; it is the object of manipulation.

Consent. The student representative's point is Kantian in form: consent to an undisclosed purpose is not consent. The participant can agree to 'memory and reaction time'; she cannot agree to what she does not know, and a signature obtained under a false description is precisely the case Kant calls treating someone as a thing.

Universalisation. Try the maxim: 'when important knowledge is hard to obtain, deceive subjects to obtain it'. Universalised, subjects would cease to believe experimenters, and the practice would destroy the condition of its own possibility - a contradiction in conception rather than a mere bad consequence. This is a stronger result than the utilitarian's, because it does not depend on numbers.

Two refinements a strong answer should make. First, Kantian ethics is not absolutist about all concealment: where nothing is withheld that the person has a right to know, no wrong occurs - but here the deception concerns the very risk being imposed. Second, the confederate raises a distinct issue: she consents, so she is not used merely as a means, though the committee should ask whether a researcher's power over a student confederate makes that consent fully free.

Conflict of interest. The supervisor sitting on a five-person committee judging her own student's application is a Kantian failure of impartiality as much as a procedural one: the applicant is being judged by someone with an interest in the outcome, not by the law of universalisability.

Yes, materially. The utilitarian's case rests on expected benefit: knowledge gained × probability the study actually delivers it. If the original's validity is disputed - questions about demand characteristics, about whether participants believed the shocks were real, about selective reporting - then the expected benefit of a faithful replication falls. Replicating a possibly confounded paradigm may produce data that resolve nothing, while incurring the same harms.

This is not an objection to utilitarianism; it is a feature of it. Consequentialism is a theory about which acts are right given how the world is, and how the world is includes what we know. A utilitarian decision made under uncertainty must be an expected-value decision, and expected values change when the evidence about likely outcomes changes. Someone who defended the study on the grounds that 'Milgram was important' would be reasoning from actual rather than expected consequences, and from reputation rather than welfare.

The deeper lesson: option value. Because less invasive designs preserve the possibility of obtaining the knowledge later, the violent design must clear a higher expected-benefit bar - it must be better than the alternative, not merely better than nothing.

Against. A critic may say that making rightness this sensitive to contested evidence makes ethics hostage to shifting expert opinion, and that the honest consequentialist conclusion is: run the study on the design most likely to be informative, which may be no study at all until the dispute is settled. That is a fair reading - and it is the one the committee should adopt.