(Please visit one of our advertisers)

No donations or subscriptions are required

   OR   
   
Subscription choices:
Board Karma = 40  (3488 positive of 3870 votes is 40 %pts higher than a neutral 50%)
All Things (Safety Oriented) Motorcycle   
Username:
Password:
Save Password
Forgot your Password?

You can the entire collection of Safety Tip articles in a 33 Megabyte PDF Portfolio

 All Forums
 Wendy Moon Archive
 2007 Blog posts
 04-18-System failures, power laws, icebergs and the culture of crashing
Member  Discussion Topic Next Topic  

Moonrider
Female Junior Member
26 Posts


St. Louis, MO
USA

Honda

VFR 750F
Peer Review: Blocked

Posted - 02/06/2011 :  2:15 PM                       Like
(Posted 04/18/2007)

Dinosaur had been trying to get me to read Deep Survival by Laurence Gonzales for months so I brought it with me on my trip a few weeks ago. He had, rightly, pointed out its application to street riding and I would have as well until that discussion on crashes at Georgia's Update. After that, I couldn't help but see it in terms of the training course.

In the book, Gonzales discusses Charles Perrow's theories on system accidents that which appears to be "accidental" really arise out of unexpected interactions of force with components of the system. An accident then is a result of a systemic failure simply because of the complexity of the system. Iow, judgments, conditions and acts that each can be inconsequential in themselves combine in such a way that accidents large and small occur. I would suggest that almost all modern disasters can be seen as system failures. While Gonzales used different accidents, I prefer to illustrate it with the Johnstown Flood in 1889.

Johnstown, PA., was a steel town of about 30,000 people nestled in a narrow valley in the Allegheny Mountains and was almost completely wiped out by the flood. Due to the geography and climate, though, Johnstown experienced at least one flood per year from snowmelt or rainfall and had stood.

While the best known cause of the flood was the failure of the South Fork dama small dam about 14 miles north of Johnstown. It had been built by 1853 to create Lake Conemaugh as a water reservoir but had ended up being part of an exclusive hunting and fishing retreat for the very wealthy. The dam had been holding just fine, then for at least 36 years, and even in 1889 wouldn't have failed had it not been for other seemingly unrelated acts: the dam had been lowered to accommodate a road and the water level in the lake kept higher than it had been. A fish screen had also been installed to keep fish in the lake for the wealthy members' benefit but the fish screen trapped debris and then waning interest in the private club led to less maintenance. While those things led to the dam breaking, they didn't cause Johnstowns destruction.

On May 30th a severe storm moved into the mountainous area dropping 6 to 10 inches of water in less than 24 hours, swelling local creeks and raising the level of the lake. Even so, that amount of rain didn't cause the dam to break nor would have Johnstown been destroyed simply by the increased water in the Little Conemaugh river or Stoney Creek River that joined it at Johnstown.

The Little Conemaugh River ran through a very narrow valley that narrowed even further at points. When the dam broke sending 20 million gallons of water into the already swollen river, the water couldn't spread out across flat land and lose momentum. Instead, it was channeled into a funnel and the river's height grew exponentially. Trees and houses washed away along the banks and jammed the even narrower passages forming temporary dams, which built the height and force behind then. When these debris dams broke, the flood was even more powerful as it rushed downriver.

Even then, though, it would not have been so deadly had man not narrowed the banks of both the Stoney Creek River and the Little Conemaugh to better accommodate the factories along the banks. But since the valley was narrow, almost all inhabitants had built very close to the river. And there was a low railroad bridge across the river. The wall of water that hit Johnstown was about 40 feet high in places and was traveling 40 mph. It knocked the town down like a pile of Lincoln Logs and the bridge trapped the debris forming a blockage 30 feet high, then started on fire which killed many who hadn't already been trapped in the debris and drowned.

In one fell blow over 2,200 people were killed including 99 entire families and over 1,600 homes were destroyed in minutes. The flood, then, is a perfect example of a system failure: many of the factors were not necessarily problematic in and of themselves, many had occurred before and separately and not caused severe problems. It was only when they all happened to come together at the same time that the town was destroyed. But had even one element had not been there, the results would have been less deadly.

Other disasters have similar confluences of happenstance, error, innocent or seemingly unrelated or unsuspected actions, judgments and conditions that come together in, for example, A Perfect Storm or the Columbia space shuttle disaster or the Titanic or Hurricane Katerina. As emergency management experts know, such large scale disasters no matter what their causes will happen more often simply because all systems are growing more complex. And, as Gonzales describes so beautifully, climbing accidents and airplane crashes.

The same is true of the simple multi-vehicle crash: it's not just one thing, it's a myriad of things that create the crash. And we know for every motor vehicle death, there's far, far, far more fender benders.

I'm suggesting it's true of crashes on the range. They are the results of system failures: a confluence of many factors that, in themselves, are not necessarily dangerous or recognized as such, that don't necessarily led to a crash on their own but, combined, led to oneand even a deadly one.

Sand piles on the range.

In Deep Survival, Gonzales talked about Per Bak's sand pile experiments, which, if you recall, I began to talk about last fall and then went off into talking about Calhoun's Rats. Or at least I meant to as that was the whole point of talking about the rats to get to the sand piles. At any rate, Bak studied how a sand pile grows and collapses using a computer model. Sand piles, as we know, only grow to a certain height before they collapse in either small slides or bigger ones thus becoming wider at the base and then begin growing higher again. Pour more sand or pour it faster and the process speeds up but it doesn't change its nature.

So even though slides are an intrinsic part of sand piles and to be expected, there's no way to predict whether the next collapse will be a bigger one or a smaller one or where in the pile it will occur. Otoh, the collapses follow that same power law: there X smaller collapses for every big one so we can accurately predict there will be a large collapse but not accurately predict when it will occur.

This is true of some kinds of natural disasters. For example, seismologists know there will be a large quake on the San Andreas that's certain but when it will happen cannot be predicted. We have such terms as storm of the century and 100-year flood because we know these kinds of events occur in some sort of pattern of small and medium size ones. There are power laws to accidents and disasters and to the stock market and other phenomena as well as Mark Buchanan points out in Ubiquity: Why catastrophes happen.

As Buchanan uses it, a power law is simply the average frequency of a given ______ is inversely proportional to some power of it's size. And, as it turns out, there's scale invariance: no matter what size the event is, it has the same structure and properties. Large events are merely magnified copies of small ones and small ones shrunken copies of large ones. Forest fires or avalanches or earthquakes for example don't behave differently depending on their size. In all these types of systems, there's far, far more small events than large ones whether we're talking earthquakes, forest fires, slides on a virtual sand pile or...well...rider training.

Power law to range crashes?

So let's look at the crashes on the range: I've already talked about the essential sameness of most crashes on the range (and that they merely duplicate the same crashes out on the street). The difference between a non-injury and injury stop 'n flop is the student body position relative to the motorcycle as it goes down. The actual run-off that results in death is no different in any way than a run-off that ends in a non-injury crash or the student regaining control. Iow, it's the ends that differ, not the means: there's scale invariance in the how and the what and the why and a power law revealed in the results.

Just like in motor vehicle crashes on the road, there's far, far, far more non-injury crashes to injury crashes and far, far more injury crashes to AIS 4 and beyond crashes. If all state programs and Harley-Davidson coughed up their crash data, we would be able to find out exactly what that power law was. And this is why:

The key change in the BRC is to allow minor errors to go uncorrected trusting that the student will learn better if they figure out how to do it correctly on their own. This, by itself, increased the number of errors unchecked on the range.

However, we also know no matter where they occur it's an invariable rule that error(s) lead to a crash without them no crash would occur. Logically and inevitably allowing more errors increase the probability that errors that produce crashes will occur. But not every error produces a crash iow, there's a vast number of errors hidden behind every crash that occurs. It suggests that the number of allowed uncorrected errors in the BRC then to switch metaphors momentarily are the unobserved bulk of the iceberg while the actual crashes are the observed tip. Iow, there may be a power law for errors leading to crashes as well.

A culture of error leads to a culture of crashing.

Since we do know the number of crashes has increased dramatically, we can infer that the number of errors corrected and uncorrected have increased exponentially and really are that bulk of the iceberg. So whether the BRC has been dumbed down or not, or which curriculum is better for teaching skills or not is, in a sense, separate from the results on the range: it's the policy about errors that is clearly associated with the crashes. And we see that in the limited data we already have: two or more crashes are now common in classes across the nation where one wasn't even standard before. So more uncorrected errors produce more crashes. Iow, rider education, under the M$F, has become a culture of crashing because it is a culture of error: errors and thus their results are accepted as normative.

Evidence for this is found in the difference between the crash rates between the RSS and BRC. The RSS philosophy held that errors should be actively corrected and also had a far, far lower crash rate. And there was only one death from rider error that significantly occurred after 28 years of that practice. If the range, then, is really like a sand pile or earthquakes or the stock market, it took that long for enough combinations of factors or system failures to finally produce the Johnstown Flood or the Big One of range crashes.

Conversely, though, even though 16% less students have been trained, at least 5 deaths and 2 near-fatals have occurred in 4 years. Iow, the combinations of factors are cycling through faster and faster meaning more and more errors and minor crashes are occurring on the ranges or such deadly system failures would not be occurring so close together.

Iow, the BRC philosophy is like pouring more sand and faster: it increases what would normally happen over a much longer time.

And, like the sand pile, the vast majority are "small slides"minor crashes resulting in no injury while there are a few and a predictable portion of large slides that will occur even though when small or large slides occur cannot be predicted.

If there is a power law operating in range crashes, then, the vast increase in errors not only will result in an increase in non-injury crashes it means that there will be a proportional increase in crashes resulting in minor injuries and a proportional increase in crashes resulting in injuries that require professional treatment and a proportional increase in deaths.

This can't be true!

I know, I know this sounds absolutely ridiculous! If you're like me, it sounds as if somehow it's separate from human will and action and that simply cannot be! Because, if it is true, that suggests there's this massive, undetected force at work that's playing out on ranges and in crashes. Surely, there's got to be variation that indicates the human element and some years there's a higher percentage of one crash or another.

However as bizarre as it sounds from the data we have, it turns out that there does seem to be the same sand pile effect and power law at work in range crashes. When the data from Kawasaki Jockey's state is examined, the rate of injuries have kept perfect pace with the increase in number of crashes. From 2001-2004, the percentage of injury crashes ranged from a high of 26.44 to a low of 24.22% averaging out to be 25.6% even though the number of crashes rose 21% during that same time. It's utterly uncanny coincidence that's just as disturbing to contemplate or a there really is a power law operating.

And so we return to the analogy of the iceberg. An Antarctic berg typically has a 1 to 5 ratio: 1 part above water to 5 parts below. An Atlantic berg is 1 to 7. Either way, that's a whole lot of errors. Then, of course, there's a further iceberg effect non injury to injury crashes: injury to non injury crashes, in Kawasaki Jockey's state is about 1 to 4 year after year.

Can it be true?

So can this actually be true that there is a power law with crashing in training? If we look which I've agreed not to discuss in detail at the field test of the BRT compared to the BRC results from two regions in the Illinois program, it does seem that there is a power law operating there as well both in the BRT and the BRC. And it's the same power law with scale invariance.

And, as we've seen with the RSS vs BRC crash stats, the same scale invariance and power law is also seen. So it's not just Kawasaki Jockey's state though that and the IL BRT/BRC results give us the clearest picture.

What's more, over 50 years of driver's education supports this idea of a power law and it's relationship to error and crashing. Hundreds of millions trained over 50 years in 50 states with untold number of curriculums public school programs, commercial school programs, private training. And yet the numbers of driver ed students killed are incredibly small and there's more injury crashes (though less of both with safer crashing cars) and far, far more "minor" crashes such as clipping a car while learning parallel parking or being rear ended by other cars. So we see the same pattern in driver's ed as the sand pile, as on the range, as with earthquakes X small to a very small number of Y and even tinier number of Z "slides". Even so, the number of even minor crashes are miniscule as in the 0.0014 range to the percentage of students trained. The difference? Errors are immediately corrected to the extent that driver ed cars have duplicate brake and accelerator controls for the instructors. While duplicate controls are not possible on a motorcycle hence the crash rate has to be higher in rider training the key difference is how errors are treated and the ability to stop errors as they are occurring. Iow, not only is the same general sense of proportion of deaths/injuries/crashes found in driver's ed, the difference in attitude towards error clearly results in a fraction of the crashes.

Iow there's good reason to suspect there really may be a power law operating and it's intriguing and should be studied. Because, if it's true, there are serious ramifications with increased number of students being trained:

The power law works not by increased students but increased errors.

Expanding the number of training sites, putting more students through training may appear to be the "more sand on the pile faster" that's increased the fatalities. But it's not.

Using M$F's published numbers, only one rider error death occurred in 1.67 million trained while, since the widespread adoption of the BRC and only 1.4 million students trained at least 5 deaths and two near-fatals and an increased number of serious injuries have occurred. The "more sand faster" is the number of crashes more errors leading to more crashes eventually produce more injury and deadly crashes.

Iow, it's not the number of students it's the number of errors that triggers, so to speak, the power law because the errors lead to more crashes while the injuries and severe injuries are proportional like an iceberg to the number of crashes.

Iow, the deaths really are the very tip of the iceberg while the bulk of allowed errors is hidden below the water. As a result, if minor errors remain uncoached and the tacit or overt policy of tolerating more crashing continues, more grievous injuries and deaths will they must occur and occur proportionally over time.

Ramifications.

This of course has serious ramifications. One of which is: the more training that occurs the more sites, the more students pushed through with the same errors allowed policy, the more injuries and deaths will occur. Mandatory training without a change in curriculum or a change in the philosophy of the current curriculum in the majority of sites will inevitably increase the numbers of deaths in training as a function of the power law.

The issue then may not be whether the BRC is dumbed down in terms of operator skills and hazard awareness. The real issue may be whether M$F's interpretation and application of adult learning is extremely dangerous and increases the instability of the "sand pile" of the rider training range.
   Discussion Topic Next Topic  
Jump To:
All Things (Safety Oriented) Motorcycle © Master Strategy Group Go To Top Of Page
  This page was generated in 0.7 seconds. Powered By: Snitz Forums 2000 Version 3.4.05