How Much is a Human Life Worth? $50?
Well, as Bo Diddley says in the classic Trading Places, "in Philadephia, it's worth fifty bucks." That's what it feels like in Philly sometimes, anyway.
Truthfully, most people would consider that a rather low estimate. People make implicit assumptions about the value of human lives. "Every life is priceless" tends to be the default assumption in many cultures.
There's a good AI ethics simulator designed by MIT to help determine how people gauge these values in relative terms. The Moral Machine offers variations on the Trolley Problem to anyone who takes the time to read and enter answers: which of two bad outcomes would you choose? This was built to "crowdsource" AI ethics for self-driving cars and look at broader societal values around AI safety. You can read more about the summary results here.
As it turns out, I have a distinct disdain for lawbreakers, and I have a hard time understanding any scenario where others pick pets over people. But the point isn't for me to convince you to run over Fluffy. A number of interesting observations emerge when you add in gender, age, and other demographic information for both the survey participant and the "victims". There are significant variations by culture in how the relative value of life is perceived.
So maybe every life is not priceless – some are more valued than others. "Women and children first" is often a human reaction to danger. In the Titanic disaster, 74% of women and 52% of children (who were disproportionately in third class in lower decks) survived, against a male survival rate of 20%.
Much of human endeavor actually implicitly prices human life rather precisely in dollars and cents, though we don't consider it in those terms. One of my formative courses in engineering was Engineering Economy, an upper-level course designed to look at how engineers make cost tradeoffs in design of systems that interact with humans.
Non-engineers assume that roads and bridges are designed as safely as possible. And it's true that they typically have high safety factors. Safety factor is a load ratio – how much load can a structure handle before it fails, as compared to its design load. Buildings and roads tend to use safety factors above 2. Aircraft use 1.2 to 1.5 because of the high costs of additional weight. To mitigate that risk, aircraft are thus inspected frequently as a safety measure.
But what about additional safety features like guardrails? One cannot build guardrails for the entire Pennsylvania Turnpike – it would be cost-prohibitive. Every dollar that goes to safety features is one that cannot go to maintenance or initial construction. So engineers must calculate tradeoffs.
In the guardrail example, those calculations go something like this: How likely is a crash at a particular site, and how severe will crashes likely be? This results in a probability estimate for the loss of life at that location in a year. It's much less than 1 -- engineers are not monsters.
Liability lawsuits in cases of death in Pennsylvania have a known average cost, and health costs for injuries are likewise estimable. This creates a tradeoff between the cost of safety features and the risk and economic costs of human injuries. Finance will weigh in at this point with amortization of the initial outlay for construction, and so a rational economic decision can be made about which guardrails, and how many overall, can be added to the highway plan.
In practice, engineers do not figure "eh, we'll only kill three people on this stretch, skip the guardrails." These factors are considered in overall budgeting. But they do absolutely estimate a value for human lives and balance our societal investment appropriately for that estimate.
Sure, Bo, it's worth more than fifty bucks. Even in Philadelphia.
[To be continued...]