I have the following case:

John Smith, M.D., is a psychiatrist with a private practice. He has been extremely successful in helping child-abusing parents. Many of his clients have sent him others; at present, nearly seventy per cent of his clients are people who have physically, mentally, or sexually abused their children.

Dr. Smith's provision of effective therapy is not the only reason for his exceptional concentration on child-abusing clients. He also systematically refuses to report child-abuse cases to the authorities, and he includes no mention of child abuse in his clients' files. Because of this, a number of his clients are serious abusers who have not previously sought counselling and who will not take their abused children to physicians because of their fear of being reported to authorities. Dr. Smith believes that by rigorously protecting confidentiality, he is able to help precisely those people who are most likely to injure their children.

Now I have to use the Act-Utilitarianism and define in which way Dr. Smith action is affecting the people affected by his action. Now my instructor told me to use numbers in terms of the amount of happiness that either created or lost. For instance, Dr. Smith action creates 5 point happiness for the parents but -6 for the affected children. Now I could just randomly assign numbers to those cases, but that does not seem right. So I am asking you guys how you would approach this case by using numbers to define the amount of happiness. Thank you.


  • That doesn't seem right to you? Well, welcome, this is Utilitarianism. I don't see where these numbers should come from other than your imagination.
    – iphigenie
    Jul 22, 2013 at 8:07
  • You may want to go with a rigorous approach, first defining your scale and then assigning from that scale. You will have to assign utility in a Bayesian probabilistic fashion because different people will have different reactions to the same approach. For information on where to get an appropriate Prior Probability, see meta.philosophy.stackexchange.com
    – shieldfoss
    Jul 22, 2013 at 9:11

1 Answer 1


The type of answer you give ideally would reflect what your instructor wants. One can approach this sort of thing at many levels of detail. In the absence of a clear idea, I suggest dealing with ranges and conditions.

For example, you can contrast the number of child-abuse cases he would see in the case where he does vs. does not report abuse, and compare the fraction of time the abusive situation stops when he deals with it vs. when it is reported. Then you can come up with inequalities that must hold in order for it to be a net positive to abused children; if the inequality fails, you then will have a formula for the relative benefit to adults vs. detriment to children that must hold for it still to be a net positive.

For example, let's suppose we're using red-light cameras on all intersections, and let's suppose that when a car runs a red light it has a fraction c of being involved in a fatal collision. Let's further say that your average citizen has a rate r of running a red light (measured in number of red-light runs per unit time), which will drop to rw if you send them a warning; rt if you send them a ticket; and 0 if you take away their license. If you see a car run a red light, what should you do?

Well, you could consider taking away the license of anyone who runs a red light. Presumably it is better to have a license than not--let's call that benefit L, but it's also better to be alive than dead, and we'll call that benefit A. If we take a license away the cost is -L*T (where T is the elapsed time) and the benefit is A*T*r*c. Thus, we should do this if and only if

A*T*r*c - L*T > 0
r*c > L/A

Even though A is presumably much larger than L, if rates are sufficiently low and the chance of a fatal collision is sufficiently low, it may yet be worth it. Now, in reality we could actually always send a warning (let's assume the cost is negligible), so we can always have rw instead of r. If we give someone a ticket, presumably that has some negative cost to them of -V. Is it better to send a ticket?

A*T*rt*c - A*T*rw*c - V*T > 0
(rt-rw)*c > V/A

So the answer is that firstly, you should only send a ticket if it's actually more effective than sending a warning (since V and A are both positive), and secondly it'd better be the case that the inequality above holds.

You can then, with conditions clearly mapped out, suggest which category things are likely to fall into. For example, in the traffic ticket case you might argue that a well-designed warning could be more effective than sending a ticket at correcting behavior, so that should always be the first resort.

Of course you can always make things more and more complicated, but this sort of reasoning is about the bare minimum you need to make sense of utilitarian considerations.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .