Robot cars, falling bridges, and the psychology of risk assessment
Bret posts an amusing hypothetical conversation at Great Guys weblog between an insurance adjuster and the owner of a car that was hit by an experimental robotic vehicle. The technical challenges of creating an autonomous land vehicle got me thinking about the ways that people ascribe risks to various activities that are often contradicted by the actual statistical probabilities of risks associated with those activities, likek flying and driving:
Relative to driving a car, flying is exceedingly safe, yet people continue to fear plane crashes more than everyday commuting. I used to think that this was related to the catasprophic nature of plane crashes, where hundreds of people are killed at once, whereas automobile accidents rarely kill more than four or five people at a time, though their cumulative daily toll attests to the risks involved.
But I've re-assessed my thinking based on the recent collapse of the I-35 bridge in Minneapolis, and the public outcry over unsafe bridges in its wake. I now think that, paradoxically, people associate higher risk value to rare, unexpected tragedies than to frequent, expected ones. People fear being killed by random serial killers more than they fear being killed by an acquaintance or a family member, even though the latter is statistically much more likely than the former, and no less likely to be brutal and fearful.
As horrible as the Minneapolis bridge collapse was, such collapses are exceedingly rare. Between 1983 and 1995 there were four highway bridge collapses that killed a total of 28 people. The news coverage has publicized the frightening fact that about one in four of the nations bridges are classified as structurally deficient, as was the Minneapolis bridge, yet if that were a meaningful statistic we would be witness to daily collapses across the US. We aren't. Even for structurally deficient bridges, a total collapse is an exceedingly rare occurrence.
Yet our nations politicians are falling over themselves in fingering blame for the collapse and in calling for comprehensive action to make sure that the I-35 bridge collapse never happens again. Should our politicians be calling for, in the words of James Lileks, a "Marshall Plan for the Nation’s Roads"?
If they did, it would represent one of the most massively expensive overreactions to a perceived problem since Mayor Quimby instituted the flying "Bear Patrol" to quell the fears of irate Springfield residents resulting from the intrusion of a single confused bear into the city.
The disaster does expose one glaring gap in our current infrastructure management environment: inspections are a very poor method for determining if, and when, a bridge will collapse. The above referenced figure that 1 in 4 US bridges have been rated deficient, which works out to around 125,000 bridges, and the fact that only five bridges have experienced a catastrophic collapse since 1983 means that we would have to upgrade 125,000 bridges to be sure that none of them would be likely to collapse in the near future. Is that money well spent?
What we need is a way to predict when a bridge is in danger of collapse. Inspections can't do it. An inspection is a static view of a structure at a single point in time. To have better predictive capabilities, we need real time data on how a bridge behaves under daily traffic and environmental stresses. How much does it vibrate? How much does it sag under a load? Is it expanding properly under heat, or are its expansion joints clogged with debris? Is the bridge deforming over time?
I'm envisioning a method whereby a bridge can be wired with sensors to generate all these data points and stream the data to computers for analysis. A healthy bridge should have a set profile of responses under the full range of load and environmental conditions. A bridge that is losing its ability to maintain itself should display a data profile outside of those healthy boundaries. It doesn't make fiscal sense to spend the hundreds of billions of dollars it would take to rebuild every deficient bridge when most of those bridges could last decades into the future without failing. We need our engineers to design bridges that will tell us when they are getting ready to fail. Call it "just in time" infrastructure management. Lets take the opportunity afforded by this disaster to develop smarter ways to manage our infrastructure. We don't need a feel-good overreaction to satisfy our fears of the moment. We don't need a "Bear Patrol for our Nations Bridges".
The robotics for a jumbo jet to take off, fly halfway around the world and land all by itself are in production now, but the robotics to drive a car around a parking lot are still in its infancy. That says something about the relative difficulties between the tasks of flying and driving. Yet people who fear to fly probably feel safe driving their car on the freeway. How can human psychology be so bad at risk assessment?
Relative to driving a car, flying is exceedingly safe, yet people continue to fear plane crashes more than everyday commuting. I used to think that this was related to the catasprophic nature of plane crashes, where hundreds of people are killed at once, whereas automobile accidents rarely kill more than four or five people at a time, though their cumulative daily toll attests to the risks involved.
But I've re-assessed my thinking based on the recent collapse of the I-35 bridge in Minneapolis, and the public outcry over unsafe bridges in its wake. I now think that, paradoxically, people associate higher risk value to rare, unexpected tragedies than to frequent, expected ones. People fear being killed by random serial killers more than they fear being killed by an acquaintance or a family member, even though the latter is statistically much more likely than the former, and no less likely to be brutal and fearful.
As horrible as the Minneapolis bridge collapse was, such collapses are exceedingly rare. Between 1983 and 1995 there were four highway bridge collapses that killed a total of 28 people. The news coverage has publicized the frightening fact that about one in four of the nations bridges are classified as structurally deficient, as was the Minneapolis bridge, yet if that were a meaningful statistic we would be witness to daily collapses across the US. We aren't. Even for structurally deficient bridges, a total collapse is an exceedingly rare occurrence.
Yet our nations politicians are falling over themselves in fingering blame for the collapse and in calling for comprehensive action to make sure that the I-35 bridge collapse never happens again. Should our politicians be calling for, in the words of James Lileks, a "Marshall Plan for the Nation’s Roads"?
If they did, it would represent one of the most massively expensive overreactions to a perceived problem since Mayor Quimby instituted the flying "Bear Patrol" to quell the fears of irate Springfield residents resulting from the intrusion of a single confused bear into the city.
The disaster does expose one glaring gap in our current infrastructure management environment: inspections are a very poor method for determining if, and when, a bridge will collapse. The above referenced figure that 1 in 4 US bridges have been rated deficient, which works out to around 125,000 bridges, and the fact that only five bridges have experienced a catastrophic collapse since 1983 means that we would have to upgrade 125,000 bridges to be sure that none of them would be likely to collapse in the near future. Is that money well spent?
What we need is a way to predict when a bridge is in danger of collapse. Inspections can't do it. An inspection is a static view of a structure at a single point in time. To have better predictive capabilities, we need real time data on how a bridge behaves under daily traffic and environmental stresses. How much does it vibrate? How much does it sag under a load? Is it expanding properly under heat, or are its expansion joints clogged with debris? Is the bridge deforming over time?
I'm envisioning a method whereby a bridge can be wired with sensors to generate all these data points and stream the data to computers for analysis. A healthy bridge should have a set profile of responses under the full range of load and environmental conditions. A bridge that is losing its ability to maintain itself should display a data profile outside of those healthy boundaries. It doesn't make fiscal sense to spend the hundreds of billions of dollars it would take to rebuild every deficient bridge when most of those bridges could last decades into the future without failing. We need our engineers to design bridges that will tell us when they are getting ready to fail. Call it "just in time" infrastructure management. Lets take the opportunity afforded by this disaster to develop smarter ways to manage our infrastructure. We don't need a feel-good overreaction to satisfy our fears of the moment. We don't need a "Bear Patrol for our Nations Bridges".
17 Comments:
I suspect that what appears to be statistical incompetence in fear/risk assessment reactions is nothing of the sort.
It is a feeling of not being in control.
If you are a passenger in an airplane, there's nothing you can do; you are trusting thousands of people, none of whom you know.
In your car, you can persuade yourself that if trouble arises, you can drive yourself out of it.
Similarly with crime. People may have violent relatives but they think they can manage them. There's no managing a random murderer, although Second Amendment fans imagine that by going armed they can.
If you think that the percentage of bridge collapses v. sick bridges is a low number, consider the number of armed citizens vs. the number of armed citizens who successfully use their firepower to prevent mischief.
This comment has been removed by the author.
Duck wrote: "I now think that, paradoxically, people associate higher risk value to rare, unexpected tragedies than to frequent, expected ones."
That almost has to be true in modern society. If we feared things we had to do all of the time (like driving) we'd become paralyzed. So we get used to driving, etc.
Also, the news cycle may play a role. Things that happen frequently (like auto accidents), are by definition, not news. Only things that happen rarely, like airplane crashes, are news. As a result, people are exposed to rare events, much more often than common events.
Also, I find it kind of weird that the fact that all bridges are inspected and the cognate fact that almost all bridges fail to fall down is spun as a failure of inspection as a concept.
I'm envisioning a method whereby a bridge can be wired with sensors...
Sounds good, but surely that will mean someday in the future in the wake of another disaster we will see reports that say:
The above referenced figure that 1 in 4 US sensors have been rated deficient..., and
What we need is a way to predict when a sensor is in danger of failing...
Harry nailed it. We are going quite mad collectively in our expectation that we can extirpate all risk, chance and happenstance from life. Who can remember the last time a major tragedy was met with mourning and reflection as opposed to lawsuits and angry political activism?
Harry
The inspection program isn't a failure, it just doesn't give detailed enough information. It's like a description of a suspect in a bank robbery. The description might say "the suspect is a white male in his thirties, around 6 feet tall with a medium build". That only narrows it down to about 20% of the adult population. If that's the only information the investigation will ever produce, then you either have to lock up every adult male that fits the description or just wait until he strikes again and hope to catch him in the act.
If the inspections don't lead to additional investigative steps that can give the engineers enough data to decide that they must fix the bridge now or they can let it go until the next inspection, then it doesn't do any good. As it is now, we only know which bridges need to be replaced after they collapse.
'As it is now, we only know which bridges need to be replaced after they collapse.'
You're assuming a fact not in evidence -- that no inspections have led to corrections that prevented collapses.
I don't know what caused the I35 collapse, but it isn't hard to imagine something that happened (as the result of a truck crash, for example) since the last inspection that led to the breakdown.
Or, it may have been caused by a hitherto unknown defect of aging or use.
My colleagues are sadly silly in demanding or expecting to know within hours or days what caused complex systems to fail. My brother, whose work it is to analyze catastrophes, knows better.
There is also Henry Petroski's argument (in 'Why Things Fail') that if some bridges don't fall, the engineers are failing society by overdesigning, which is wasteful.
The book is a lot more intelligent than my 15-word summary and I recommend it highly. It is about engineered systems generally, but bridges are his focus.
It can be compared to the credit union run by a friend of mine. He is proud that it has never had a loan default. I say, if he's that tight, then he's failing to do one of the jobs of a community lender, which is to create opportunity.
Last, we can always suspect misfeasance. Maybe the inspections was incomplete, hasty or corrupt.
This comment has been removed by the author.
erp
The repairs weren't to fix the structural defects, they were making surface repairs to the road deck.
... as a regular user of the I-35 bridge, I can attest that that bridge, due to its unusual design (which has since been discontinued) has always struck me as being under-built, especially compared to the 10th Ave bridge which is twice its age or even the stone arch bridge nearby which is three times the age of the I-35 bridge. This type of bridge is prone to catastrophic failure, it had been deemed deficient and was not repaired or replaced due to short-sighted politicians. This is what has people so upset, knowing that it will happen again unless measures are taken. It isn't as expensive as you might imagine, a national 10¢ per gallon tax on gasoline would probably cover it. We're throwing away far more in Iraqi infrastructure than any domestic bridge program could cost.
Duck, thanks for the update. I removed my comment when I realized my error.
I still think that even given all of the engineering talk above, the option of sabotage can't be ruled out and I'm afraid there will be a cover-up, similar to that of TWA Flight 800, should that be the case.
We know why TWA 800 exploded. There wasn't any coverup.
Why did it explode?
Spark from decayed wiring harness in a fuel tank full of fumes.
There's a great deal of controversy about that conclusion.
erp: TWA 800 wasn't brought down by a missile.
Harry: worn, not decayed; and outside the tank, according to the NTSB, not inside. The theory is that chafed wiring outside the tank shorted and induced a spark in the fuel sensing system inside the tank. It seems pedantic to point that out, I know, but it bears on the difficulty of catching a problem that hasn't happened yet by visual inspection. You may remember that after TWA 800 the FAA ordered wiring inspections for the whole 747 fleet, and sure enough they found a number of 747s and a whole lot of 737s with the same symptom. I can tell you from experience that you'll always find chafed wiring, unless you're inspecting a brand-new aircraft, and I'm pretty sure you'll also always miss some too.
Post a Comment
<< Home