Computational Models Suggest Heroism Is Suboptimal for Saving Lives in an Emergency

Maybe George Costanza was right.

by Michael Byrne
Jun 25 2017, 2:00pm


We all like to imagine that we'd be the hero. In the fuel-leaking wreckage of a downed airliner, we wouldn't freeze or shut down. We wouldn't run. We'd put others before ourselves, carrying away endangered children and wounded survivors with heretofore unknown reserves of strength and resolve. Ultimately, we would prove to be the excellent, exceptional people that we imagine ourselves to be. Fortunately, most of us will never have to reconcile that self-conception with the reality that more than anything we'd just prefer not to die ourselves.

That impulse—the one that drives us to leapfrog children and-or other vulnerable humans who may be blocking the way to the closest emergency exit—might not be particularly flattering, but its results could wind up being a net positive in terms of overall survival during an emergency. This is according to a new paper from researchers at the University of Waterloo published in Expert Systems with Applications. With help from new computer models, the Waterloo group makes the counterintuitive finding that that putting others first in a disaster can actually cost lives overall. Selfishness may (sometimes) be optimal.

Before getting into the study findings, some context is important. In Japan, in particular, catastrophic flooding driven by tsunami and tropical storms is a clear and present public safety concern. Of course, 2011 saw over 15,000 deaths courtesy of the Tōhoku earthquake and tsunami, but a bad typhoon year (such as 2012 and 2006) can easily see deaths in the hundreds. In catastrophes like these, the only real way to bring people to safety is evacuation. And survival-by-evacuation is much more of a computational problem than it may seem.

Eishiro Higo, a civil engineer and lead author of the new study, has been researching the problem of evacuation and survivability for some time. In 2012, Higo presented a paper describing the Vitae System, a framework for modeling "an evacuation and rescuing process under survivability-critical states." Generally, it's a way of predicting the likelihood of an individual's self-rescue based on three somewhat blurry factors: survivability, vitality, and conviviality.

Higo et al/2012

Survivability in the Vitae System is what it sounds: how bad the shit actually is. More than anything else, how deep is the water? Vitality is then a metric of stamina, or how long a disaster victim can hold out in a temporary shelter, such as the top of a tree. And then conviviality basically just corresponds to how well a disaster victim can communicate with other survivors, rescuers, and the environment itself.

In the Vitae System, all three of these factors are backed by formulas that return scores for vitality, or how likely you are to survive a disaster. The new paper expands on Higo's previous work by applying the Vitae System, particularly the third metric of conviviality, to a simulated disaster in which an underground mall and adjacent subway station become flooded.

"Helping people from a safe location is still good behaviour and the result is actually much better"

Higo and co. tested three survival strategies (or algorithms) within this scenario. The first is total selfishness—what we might call the George Costanza strategy. No one helps anyone. Your reaction is limited to getting as far away from danger as possible as soon as possible. Next is the group strategy, which is the opposite. Survivors help other survivors help other survivors, even if their own survivability is not yet assured. Finally, there's the rope strategy. Here, survivors first get themselves to relative safety and then toss a rope to other would-be survivors.

The third strategy won handily, resulting in far more lives saved overall. Given a worst case disaster warning period, where evacuation starts 12 or more minutes after the start of the flooding, over 80 percent of evacuees fail given the group strategy. If they follow the rope strategy, however, that figure drops to about 45 percent. Of course, this presumes that the subway station had a rope at the top of its staircase to begin with.

"Foolhardiness is not a good strategy for rescuing," Higo offered in a statement. "In very critical situations, we have to be kind of selfish, but we can still help others if we have proper equipment and proper strategies."

"We have to identify what is brave and what is reckless," he said. "Helping people from a safe location is still good behaviour and the result is actually much better."