You would draw on a body of knowledge previously developed to try to judge, possibly through mathematical-like formulae, the reliability of a technique or techniques to produce capitulation, the chance that accurate, actionable information could be extracted from the capitulation, assessment in the keenest way possible of the status of the umbrella investigation, assessment of the certainty of the catastrophic event, assessment of the damage that the event would cause, projection of the longitudinal ramifications of the catastrophe actually occurring, projection of the effect of a "wrong" decision" on future civilized life (e.g., wrong -- not trying (catastrophe occurs): the effects of the catastrophe on future interrogation policy, and much more; wrong -- trying without breaking the subject: the corrosive effects of uncivility perhaps perpetuated but mitigated by the outcry over the catastrophe itself; wrong -- breaking the subject but finding he has nothing to offer or else provides deceptive or useless information: see immediately preceding elaboration; and so on to "right and successful": catastrophe aborted). Beyond my particular ken, there are calculuses and analytical trains that may be able to make these matters more intelligible and usable, which, I suspect, society would opt for (i.e., trying to optimize data in a quite scientific undertaking), all the time informed by every available and helpful value, distilled to a usable philosophy. The basic moral problem as conceived is this: when do you risk the effects of degradation from torture (though please note, like "legitimate" collateral damage from bombing in wartime, say, this may be substantially mitigated by the importance of the given mission) to prevent (as posited) a truly earth-shattering catastrophe taking "x" amounts of lives with all sorts of horrible maiming, "x" being to the limits of your imagination.