A closer examination of risk-benefit analyses reveals
some conceptual difficulties.14 Both risks and benefits lie
in the future. As there is some uncertainty associated with them, we should address their expected values (provided such a model fits the situation); in other words, we should multiply the magnitude of the potential loss by the probability of its occurrence, and similarly with the gain. But who establishes these values, and how? If the benefits are about to be realized in the near future but the risks are far off, how is the future to be discounted in terms of, say, an interest rate so we can compare present values? What if the benefits accrue to one party, and the risks are incurred by another party? Given sufficient information, an individual can decide whether to participate in (or consent to exposure to) a risky activity (an experiment). Individuals are more ready to assume voluntary risks than they are involuntary risks, or activities over which they have no control, sometimes even when the voluntary risks are 1,000 times more likely to produce a fatality than the involuntary ones. Some would advocate that the marketplace should decide, assuming market values can come into play. But today there is no over-the-counter trade in lives. Nor are even more mundane gains and losses easily priced. If the market is being manipulated, or if there is a wide difference between “product” cost and sales price, it matters under what conditions the buying or selling takes place. For example, if one buys a loaf of bread, it can matter whether it is just one additional daily loaf among others; it is different when it is the first loaf available in weeks. Or, if you are compensated for a risk by an amount based on the exposure tolerance of the average person, yet your tolerance of a condition or your propensity to be harmed is much greater than average, the compensation is apt to be inadequate. The result of these difficulties in assessing personal risk is that analysts employ whatever quantitative measures are ready at hand. In regard to voluntary activities, one could possibly make judgments on the basis of the amount of life insurance a person buys. Is that individual going to offer the same amount to a kidnapper to be freed? Or is there likely to be a difference between future events (requiring insurance) and present events (demand for ransom)? In assessing a hazardous job, one might look at the increased wages a worker demands to carry out the task. Faced with the wide range of variables possible in such assessments, one can only suggest that an open procedure, overseen by trained arbiters, be employed in each case as it arises. Conversely, for people taken in a population-at-large context, it is much easier to use statistical averages without giving offense to anyone in particular. Risks and benefits to the public at large are more easily determined because individual differences tend to even out as larger numbers of people are considered. Also, assessment studies relating to technological safety can be conducted more readily in the detached manner of a macroscopic view as statistical parameters take on greater significance. This occurs, for example, in studies by the National Highway Traffic Safety Administration (NHTSA), which proposes a value for human life based on loss of future income and other costs associated with an accident. Yet, studies repeatedly show that “policy analysts do not evaluate the risk of their subjects’ lives as highly as people evaluate risks to heir own (and others’) lives. Consequently, too many risks are taken. It is almost impossible to build a completely safe product or one that will never fail. The best one can do is to assure that when a product fails, (1) it will fail safely, (2) the product can be abandoned safely, or—at least—(3) the user can safely escape the product. Let us refer to these three conditions as safe exit. It is not obvious who should take the responsibility for providing safe exit. But apart from questions of who will build, install, maintain, and pay for a safe exit system there remains the crucial question of who will recognize the need for a safe exit. Providing for a safe exit is an integral part of the social experimental procedure—in other words, of sound engineering. The experiment is to be carried out without causing bodily or financial harm. If safety is threatened, the experiment must be terminated. The full responsibility cannot fall on the shoulders of a lone engineer, but one can expect the engineer to issue warnings when a safe exit does not exist or the experiment must be terminated. The only way one can justify the continuation of an experiment without safe exit is for all participants (including the subjects of the experiment) to have given valid consent for its continuation. Here are some examples of what this might involve. Ships need lifeboats with enough spaces for all passengers and crew members. Buildings need usable fire escapes. The operation of nuclear power plants calls for realistic ways to evacuate nearby communities. The foregoing are examples of safe exits for people. Provisions are also needed for safe disposal of dangerous products and materials: Altogether too many truck accidents and train derailments have exposed communities to toxic gases, and too many dumps have let toxic wastes get to the groundwater table or into the hands of children. Finally, avoiding system failure might require redundant or alternative means of continuing a process when the original process fails. Examples would be backup systems for computer-based data banks, air traffic control systems, automated medical treatment systems, or sources of water for fire fighting.