Schneier on Security
A blog covering security and security technology.
« Scareware: How Crime Pays |
| Bomb-Sniffing Mice »
February 8, 2011
I'd never heard the term "micromort" before. It's a probability: a one-in-a-million probability of death. For example, one-micromort activities are "travelling 230 miles (370 km) by car (accident)," and "living 2 days in New York or Boston (air pollution)."
I don't know if that data is accurate; it's from the Wikipedia entry. In any case, I think it's a useful term.
EDITED TO ADD (2/12): Discussion here.
Posted on February 8, 2011 at 5:46 AM
• 55 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
I dunno, sounds like it's diagonal to the micro-helen.
I wonder if you can mathematically relate mircomorts to microhelens?
@ wiredog: Men have been getting themselves killed for the sake of a little you-know-what since the dawn of time, so it should be possible to arrive at a correlation.
But to take strictly the historical case, just figure out how many extra deaths there were in the Spartan and allied armies vs. similarly aged populations that stayed home, factor in the total number of ships and the average crew complement...
So we could now say that the chances of being killed by a terrorist attack while flying is .00x micromorts. (Redundant I know, but you can see where I'm going with this.)
Neat and clean. I like it.
How to you convert micromorts to petite mort(s)? SCNR
Macromort = Attempting to court positive publicity by antagonising Anonymous.
Come on Bruce, we await your thoughts on this one...
1 millihelen is the energy it takes to launch one ship. So assuming there's a 5 in a million chance of dying while launching a ship, that's 5 micromorts to a millihelen, and 5000 millimorts to a helen.
OK, a microhelen is the unit of BEAUTY sufficient to launch one ship. Helen had 'a face that launched one thousand ships'. And that wasn't the Spartans, it was the Achaeans; much earlier people.
Great post Bruce.
I think Wikipedia should have cited Roseanne Roseannadanna who frequently said "Well, Jane, it just goes to show you, it's always something."
Perhaps it is like the inverse of a megadeath ;-)
How appropriate... since "dead" is about the only way I'd be caught in NYC or Boston.
@Rich and @SteveD,
Did you check out the health tab on that animation? Supposedly the U.S. has such wonderful (albeit overpriced) healthcare. Yet, the risk of giving birth in the US is over 2x the risk of giving birth in the UK and over 3x the risk in Sweden.
The fields of science and engineering have usesd "milli" (1/1000) "micro" (1/10^6) "nano" (1/10^9) and "pico" (1/10^12) for some considerable period of time.
Sometimes it can help to understand things such as the "nanosecond" is aproximatly the time it takes light to travel a foot when designing printed circuit boards. But is it realy relevant outside a field of endevor?
Further does it "scale to far" after all what would be a likley contender for a "picomort"?
So playing Russian Roulette with a six-shooter would be a hexamort?
TimKeck: keep your "milli" and "micro" distinct.
A milli-Helen is the quantity of beauty necessary to cause the sailing of one ship of war. (Under the circumstances suffered by the original Helen.)
The US birth risk is a little misleading,
First off it varies very widely across the US (unlike the UK and Sweden) being worse than some third world countries in some areas...
Secondly it suffers from an "abisca problem" caused by the law of "unintended consiquences".
A natural birth carries certain quantifiable risks prior to the birth, but many that are not quatifiable except on mass post occurance. Likewise various forms of "assisted birth".
However things change with time and improved techniques, for instance the Romans knew about and very occasionaly performed cesarian births, however whilst the child might survive the mother would certainly die this is not the case today.
So before carrying out an assesment to do or not do a given procedure the medical team make certain choices. The issues is quantifing them all correctly.
For instance the use of various drugs that are used to stop the shock of surgery killing you have very wide ranges of risk depending on many factors. One of which is being overweight, however it is very difficult to guess just how overweight somebody is just by looking at them.
Thus a medical practisioner might over rate one risk and underrate another thus actually inadvertantly increasing the risk of their actions whilst thinking they are reducing the risk...
And part of the risk calculation in the US will be based not on proven medical benifit, but on potential loss due to "medical negligence" litigation and what the insurance companies dictate...
"So playing Russian Roulette with a six-shooter would be a hexamort?"
Oh that it were that simple...
The odds actually depend on the method used to "stop" the cyclinder, "when" and at what "angle" the gun is held at...
The wonderful mathematicians at the BBC's More or Less have discussed Micromorts several times.
So would it be more accurate to say that the coward dies a thousand millimorts?
Re: Health tab
If you'll notice the US "Giving Birth" rate is about the same as the England/Wales "Caesarean" rate. Guess which country is doing Caesareans at something like a 70% rate for first birth and 100% for births where a Caesarean was previously used.
I think that by picking and choosing which to show as birth in general and which to show as Caesarean may be skewing the stats somewhat. (Though it's probably just the innocent issue of not having Caesarean birth stats separate from natural birth stats for all countries.)
No, a million millimorts (10^6 * 10^-3). A thousand millimorts is just the one death.
(I remember model train exhibitions having signs saying "warning, do not touch, 12,000 millivolts")
@No One - it was alleged that Caesarean are more popular in the US for insurance reasons. A natural birth is assumed to be a normal procedure - if the child/mother dies somebody must have done something wrong.
A caesarean is an operation - there is an assumption of risk, is somebody dies you have to prove that there was an error - hence the insurance costs are much less.
Similarly the insurance costs for me talking a group mountain biking (a normal activity) is much higher than scuba diving (where there is assumed risk)
@Clive: On scaling down to pico morts.
If driving 230 miles is a micromort, then a picomort will be driving one foot two and a half inches.
Similarly living for 173 milliseconds in Boston/New York will also be a picomort.
@NobodySpecial: it was alleged that cesareans are more popular *than what*? I don't ever remember this being clear: more popular than natural, than in Mexico, than 10 years ago, etc.
It sounds like one of those backwards units of measure, like the Brits used for metering phone calls in the early 1990's (not sure if they still do).
Consumers pay a fixed price for a "unit". One "unit" will buy you maybe 1 minute of local call time, but maybe only 6 seconds over international call time.
It makes my head hurt.
You are correct. My bad. Millihelen it is.
Strange. The Wikipedia entry doesn't define the term "micromort" as a portmanteau or two other terms.
Talking this through with the family, my niece is proposing that on this scale a value of 1 voldemort is assigned to a scenario where the situation is desperate, the odds against us and we are all most likely to perish. In a typical IT context, a fine example would be a project you take over from Accenture.
For CTM (Common Terrorist Management) pencil pushers, we could introduce the "alamort" as the likelihood an incompetent operative gets himself arrested or blown up before being able to wreak havoc. As in the case of the black widow whose bomb belt alledgedly went off when her prepaid cell phone unexpectedly received a text message with new year's greetings from the provider.
"Similarly living for 173 milliseconds in Boston/New York will also be a picomort"
Is it actualy possible to live that long in New York ;)
More seriously I spent some time there a number of years ago, and it was how I discovered the true meaning of a love hate relationship...
Just ask any NY brownstone dweller about "their cockroaches". Some love them so much they actualy wear live ones as an accessory...
Me I hate them especialy when they run over your face during the night...
Must mean a sure thing...if it's a one-in-a-million chance literature assures us that the hero will survive and the villian be vanquished.
Terry Pratchet always claimed that "million-to-one chances happened about nine times out of ten" when a hero was in dire trouble. In one of his books I think the characters determined that the odds weren't high enough to invoke this, so they had to raise the difficulty to get the odds up to a million-to-one.
I cannot help but to point out that one-in-a-million is the probability which is almost certainly going to happen. You often say one-in-a-million-plus-one to avoid this probability flux, carelessly referring to it might invoke it. Now I'm deeply concerned.
( http://wiki.lspace.org/wiki/Million-to-one_chance )
I see an issue with this kind of unit: probability doesn't add up linearly. If you do two activities, each of which has probability 0.1 of killing you, the combined probability is not 0.2 but 0.19 (because if you die once, you don't get another chance to die again)
Perhaps a logarithmic scale would be better: X micromorts have an e^(-X/1000000) probability of not killing you. ;)
This is yet another misunderstanding and misuse of statistics. A one in a million chance of dying is NOT the same as one millionth of a death (which is the literal definition of the idiotic term micromort). This misuse leads to the nonsensical claim that breathing NYC air for two days will have a one in a million chance of killing you. Wrong. Unless you already have severe lung disease, the probability of dropping dead because of breathing NYC air for two days is zero. In fact, the probability of dropping dead because of breathing NYC air for two months is zero. It takes a lot more exposure than that to acquire enough pollutants to cause acute pulmonary failure.
Thank you, Dr. T, for demonstrating 2 micromorts of exasperation.
@vwm: with autoerotic asphyxiation but it takes very large numbers of micromorts.
Re: the slightly off-topic discussion above about risks of giving birth in different countries: One possible explanation might be the US attitude to abortion. If cultural norms mean that abortion, under any circumstances, is looked down on, then that might cause difficult and risky pregnancies to be brought to term which in more liberal countries would be aborted.
Though I'm British, and love the NHS to bits, so I'm entirely happy to believe it's all down to healthcare quality anyway.
@ Dr. T,
"This is yet another misunderstanding and misuse of statistics... ...This misuse leads to the nonsensica claim that breathing NYC air for two days will have a one in a million chance of killing you. Wrong Unless you already have severe lung disease..."
Err sorry I have to disagre in part.
Statistics have some fundemental problem like all sampling systems of grouping unlike items together and using different sampling rates without appropriarte conversion.
Air quality is one of those things that is variable in many ways and at different rates.
For instance if you get a temprature inversion then while it lasts things like oxides of nitrogen get trapped and the concentration can rise very rapidly for a short period of time.
Likewise particulates from partial combustion of hydrocarbons and plant pollen all show seasonal changes which can vary by several orders of magnitude, but also can vary by several orders of magnitude within a day (mushroom spours do this). And it has happened that in some US ports the unloading of a bulk cargo (dried soya products) into silos can cause massive outbreaks of alergic respiratory attacks some of which have killed downwind of the silos.
Thus the risk at any time is very very variable. So to use the millimort scale it might be 0.5micromorts on day X in location A and 100micromorts on day X in location B and on day Y location A might be 100micromorts. And it might easily go up above 100,000micromorts on day Z in a localised area for one particular substance.
Thus if you look at deaths by respiratory related illness (one of the biggies due to smoking) you tend to see a background rate, a seasonal rate, a weather rate and location rate at any given point in time and space.
Now on the lose (and incorect) assumption these are just sinewaves of different frequencies you know that you will get peaks and troughs as they come into and go out of phase.
Also it is known that to some things the human bodies response is markedly nonlinear. If you make another assumption that the deaths are related to the square of the air quality then a simple graph ploting just four sine waves of different frequencies and the square of their combined amplitudes produces a complex graph with a high degree of variability.
If you sample this complex graph you are effectivly performing a low pass filtering function where the high frequency information folds back into the frequency domain below the sampling frequency, if you then sample the samples (which is what statistics tends to do) you get a further lowpass filtering effect. If you look at a graph of the result you will quickly realise that the only visable correlation to the untrained eye between the original high frequency graph and the low frequency statistical result is the area under the curve...
It is why in my original post above I said,
"Further does it 'scale to far' after all what would be a likley contender for a 'picomort'?"
That is, what natural occurance that might be fatal is so constant it can make a reliable prediction on a 10^-12th scale?
The answer is actually 'none' simply because we only have ~7 x 10^9 humans on the earth and the average life expectancy in various locations varies from 35years to 85years...
I'm not too convinced by these stats; to have the inevitable total probability of 1 you need to do about 33 micromorts every day, and given the examples in the article that's going to be difficult unless you're a wine drinking smoker that likes to canoe in coal mines. Or maybe I'm missing something and there are further macromorts to be included too.
heh, use only "large" and "small" as modifiers for death?
if we thought the five color terror-alert was confusing...
maybe it should have been only black and white?
microterror: one-in-a-million probability of a terrorist act
"To have the inevitable total probability of 1 you need to do about 33 micromorts every day."
Hmm I wish I was going to live that long but who knows ;)
The more usual quoted lifes span of "75 years" I have always reckoned to be a bit of a fudge because,
1, 36.5 x 365 x 75 = 999187.5 (~1 million)
2, 60 x 24 x 365.25 x 76years = 39972960 (~40 million minutes)
3, 60 x 24 x 365.25 x 76bpm = ~40 million heart beats a year
And a whole bunch of other numbers that drop out as something easy to remember (like 1000hours ~40days), which enable quick mental arithmetic to give a "sanity check" on "long hand" calculations.
That being said you are right about the "comming up short" of the 36.5 micromorts / day.
There is an argument going around that you can tell a natural death from an unnatural death by the length of time you die.
That is those dying from just "old age" live an apparently normal existance apart from getting physicaly frailer and then just have a rapid decline or short series of illnesses (viral followed by bacterial) and die in 1 month or less. Any longer than two months can be said to be a death by other causes such as an environmental factor the body keeps fighting and a week or less is normaly down to some fairly recognisable trauma or single major infection.
Now if you take out the longterm and very short term you find that life expectancy goes up by something like 20years currently. But importantly they are "20 capable years". Thus having a healthy existance from day 1 and avoiding environmental stressors and keeping the mind active should get you up to just under 100years of acceptable life expectancy currently...
Further some give figures that show that life expectancy is rising by as much as one year in every five due to improvments in medicine and environment. And there appears to be a direct socio-economic tie in with this in that the richer you are the healthier place you tend to live in, you don't mix with unhealthy people and you have good timely access to high end medical treatment.
Which is making quite a few governments with reasonable state pension and health care systems quite nervous as it means the age of retirment will most likely have to move up 15 years in the next 30 years.
And to make this extended working life viable people will have to have much higher accademic abilities so instead of going of to work at 16-18 they will get a higher education and go to work at 21-27. Which means adding another few years to the age of retirment...
Hence the "avoidence" tactic of life long learning proposed by some, that is higher education is taught part time to the majority whilst the person learning remains economicaly self active, or if their socio-economic position is good they either go into family funded full time education or as is happening in some places unpaid "internships" to get the very few top jobs...
The upshot is if your parents are poor "you are going to work to death" in your 70's so that those in a better socio-economic position will be able to lead a comfortable retirement for upto 25years into their early hundreds with a significant number crossing the 110year mark.
Which brings the base down to under 25 micromorts/day for the wealthy and over 40 micromorts/day for the poor...
Dr. T is correct in that "micromort" glosses over the difference between cumulative and independent events. If I were to smoke, each individual cigarette would do cumulative damage to my lungs and heart, increasing the chance that they'd kill me. If I were to drive to and from work each day with a 0.001% chance of dying on the trip, there's nothing cumulative about the chances: either I'll die on a particular commute, or I'll be unharmed by it.
However, the main difference is in how you sum the micromorts, and at low probabilities P(A) + P(B) == P(A or B), for all practical purposes, regardless of whether A and B are cumulative or independent. We aren't trying to be precise here.
As the saying goes, all models are false, but some are useful.
'Did you check out the health tab on that animation? Supposedly the U.S. has such wonderful (albeit overpriced) healthcare. Yet, the risk of giving birth in the US is over 2x the risk of giving birth in the UK and over 3x the risk in Sweden.'
And your challenge for the day is to figure out why these apparently-huge disparity in the risks of childbirth might be.
I'm sure that the numbers are absolutely correct - let's stipulate that they are. Now, ask yourself - does it really seem likely to you that the standard of healthcare in the US, on the average, is so terrible that the risk of childbirth is really that much higher than in the other countries mentioned - all other things being equal?
There's a very large hint in the very animation tab you are looking at.
Minor nit for the usually impeccable Clive:
Sometimes it can help to understand things such as the "nanosecond" is aproximatly the time it takes light to travel a foot when designing printed circuit boards.
A PCB designer who assumes "one foot per nanosecond" applies to signals on a PCB is either designing some very slow boards or destined for re-assignment. Taking a phrase from the automobile ad disclaimers: "Actual speed may vary and will probably be worse" Sometimes much worse.
"traveling 6000 miles (9656 km) by train (accident)"
So, let me get this straight. Driving one (1) person 230 miles in a car is one (1) in one million (1,000,000) chance of death.
On train, this is 6000 miles, so 6000 / 230 == 26, in other words 1/26 times the car or a one in twenty-six million chance.
But wait! There's more!
Act now and you can be one of about 230 (maybe 250) on the train to Hell! Because of this exclusiveness-- less so than the single person driving solo, of course, but still quite exclusive-- the probability of death averages out!
That means your chances of dying on train are one in one thousand million?
(My statistics may be off due to the proportional amount of people on the non-Hellbound trains; but I still find it funny)
We need more trains.
Yes, John, trains and planes are far safer per mile than cars. I believe trains are safer because they're on a fixed path with fixed and known interaction points with others (for the most part) and (commercial) planes because of similar reasons in addition to being piloted by professionals rather than amateurs. (Professional truck drivers also tend to get into fewer accidents per mile than the average commuter.)
What, it's not Voldemort's version of "Mini-Me?"
"Minor nit for the usually impeccable Clive"
As it's a "nit" I'm going to have to scratch my head a little 8)
And say, yes what I wrote does not read very well.
However "light" does travel at slightly less than 3x10^8 m/second in "freespace" which is aproximatly 30cms /nS which is about a foot in "old money", so gives a rough mental value for sanity checking what you do.
However the valid point you are making is that propagation times vary when not in "free space", and a PCB like nearly all transmission lines has a propergation time related to various things such as the dialetric of the board and the actual distance traveled in 1nS can be significantly different (for instance most coax cables for TV or Radio down leads from roof antennas come in with a PF around 0.66).
I'm wondering how big a ship a millihelen would launch. In those days the ships were really quite small. If it's a function of displacement then a modern aircraft carrier could probably take a few thousand Helens. Would it be sensible to take a boat of a constant size ( a canoe? ) and work from there?
With apologies to Dave Barry, Micromort would be a great name for a chamber quartet.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.