jumbled risk analysis thoughts

[gfish]

I guess a study came out recently, because I've now seen 'texting while driving make you 21 times more likely to get into a wreck' quoted several times recently. And, yes, I have no doubt texting while driving is quite stupid. But that phrasing bugs me a lot, because it implies an arbitrary threshold. (And nothing annoys computer scientists like arbitrary thresholds!)

You know what else makes you 21 times more likely to get into a wreck? Driving 21 times more. Sending a text takes, what, a minute? So in that minute you're exposing yourself to the same risk as in the next 21 minutes of driving. Of course, no one would ever imply you shouldn't drive for 21 minutes. Driving is dangerous, period. It's one of the most dangerous things most people ever do. You can't ignore that, and pretend that risk doesn't exist. How many people are horrified at the thought of riding a motorcycle occasionally, but buy a house out in the exurbs which forces them to drive 10+ hours a week?

The texting stat as reported annoys me because it implies that the risk taken while driving is perfectly acceptable, but anything more than that is absolutely irrational. It implies that there is 'safe' and 'unsafe', instead of a continuous range of risk which everyone has to evaluate themselves.

Comments

[novalis.livejournal.com]

It's not reasonable to say that everyone has to evaluate it themselves, since crashes can affect other people (even non-drivers).

[kathrynt.livejournal.com]

Your analysis makes sense, if and only if the relative risk is compared minute-by-minute to get the factor of 23. I looked but can't find out if that is indeed the comparison. It could be that an hour of driving, during which texting takes place, is 23 times as risky as an hour of driving during which texting does NOT take place. But like I said, I don't know. This is yet another example of the shitty nature of science reporting.

[technoshaman.livejournal.com]

It's a statistic, and statistics are by definition not about reality, because they are lossy.

Just as driving 21 times more does not *necessarily* make you 21 times more likely to get into a wreck. The lossy statistic disregards experience, fatigue, the thoughts of "oh, I'm just going a mile down the street rather than 21 miles in from the burbs, I don't have to be as careful"...

Personally, I think it's rather silly to try to ban individual behaviors as too risky. I don't know if it's a city ordinance or a state law, but I know my ex got cited for "inattention" for failure to see a "Stop Ahead" sign... which resulted in her rear-ending another car. *Simple* laws like that, where you throw a human in the loop (and potentially 12) to decide whether you screwed the pooch or were legitimately snookered by the circumstances... much better than "you can't do this, you can't do that" nanny laws, which by Kantian imperative lead to everything that isn't prohibited is compulsory. Nay, f*** that.

But the use of meaningless statistics (a phrase which is often redundant) in FUD? Pitch it.

[encapsulate.livejournal.com]

Driving while blindfolded increases risks of accidents... I'd say at LEAST 10 percent. MINIMUM.

[ilmarinen.livejournal.com]

Yeah, I'm really unclear on how that statistic is broken-out...if that 21x increase is during texting, or per driving session, or what. Which makes the discussion pretty absurd.

But yes, no-one does proper risk analysis. I'll often engage in high-risk behaviors for short periods of time, and I *acknowledge* that as reasonable. One has to integrate risk over time to evaluate full exposure.

So, if I living my entire life crossing the street at unmarked crosswalks at rush hour, I'd expect my life expectancies to radically decrease. However, I'll still cross the street, I just don't hang-out there.

-B.

Or another interpretation:

[stolen-tea.livejournal.com]

I think they clearly meant that, when all drivers are divided between those who have even once texted while driving, and those who have never texted while driving at all, the former group is 21 times more likely to get into a wreck than the latter.

Because texting while driving is clearly the mark of a lunatic with callous disregard for their own safety and that of others, and you can't expect that kind of person to drive safely at all, let alone be one of the Elect.

[spinnerin-ftw.livejournal.com]

Here's actual info on the study: http://www.nytimes.com/2009/07/28/technology/28texting.html

I thought this was an interesting discussion of actual and perceived risk around cell phone use and driving in general: http://www.nytimes.com/2009/07/19/technology/19distracted.html

When complete strangers wander up and comment in your lj...

[stellar-muddle.livejournal.com]

What the statement is probably missing, is a "per km traveled or hours spent traveling". The sort of thing that the scientist who wrote the report probably stated in the first place, but that the reporter thought was "waffling on" and "unnecessary qualifiers".

Similar statements can be made risk by travel mode eg in New Zealand, motorcyclist risk of death or injury per 100 million km traveled is about 16-18 times the risk for car drivers for the same distance. The risk exposure per time spent traveling is similar. They never mention that the death/injury risk per time spent traveling for bus passengers is several times less than that of the drivers...

(basal_surge's partner, but commenting due to a professional interest in transport risk analysis)

[callmem.livejournal.com]

That is so true. I notice lots of people think my rockclimbing and veganism are of questionable safety, but really the most dangerous thing I do is commute (and drive to climbing destinations).

It's like the Thurber story about the fly from Fables for Our Time. The fly avoids the spiderweb because he sees no other flies there, but lands on the flypaper 'cause he thinks it's a party. Moral: There is no safety in numbers, or anywhere else.