Morning Links

Wednesday, July 1st, 2009
  • What happens to your keys and passwords after you die? Cory Doctorow looks at the various ways of giving loved ones access to your post-mortem online life.
  • On the topic of police dogs, someone in the comments posted this 2007 Grits for Breakfast post, in which a consultant expert on the use of K9s says the dogs are wrong about half the time. No idea how accurate that is, though it’s consistent with what cops from LEAP have told me.
  • Publishers Weekly interviews comic artist Peter Bagge, whose new book is a collection of the editorial comics he has written for Reason over the years.
  • Wired follows up on bCurtis Melvin’s work using Google Maps to annotate North Korea’s geography.
  • WalMart supports an employer health care mandate. Weirdly, this will likely win the company praise from its traditional critics. In truth, this really is an effort to impose expensive, government-enforced burdens on the company’s mom-and-pop competitors. Yet another example of how behemoth companies tend to welcome federal regulation, not shun it. More regs make it more difficult for upstarts to compete.
  • Stock up on Nyquil and Allerest now. The feds may ban them. Ridiculous. When you consider how many people benefit from the acetaminophen’s pain relief properties, 458 deaths per year sounds almost like a rounding error. (MORE: They want to ban Percocet and Vicodin, too.)
  • The Daily Show’s terrific reporting from Iran.
  • Husien Shehada, a 29-year-old unarmed Virginia man, was shot dead while vacationing in Florida this week. Police were apparently investigating reports of a man carrying a gun outside a nightclub. It doesn’t appear that he did anything wrong at all. The police bizarrely then interrogated the man’s brother and girlfriend about whether “they spoke Arabic,” then arrested the man’s brother for beating his girlfriend (he denies the charge). The cop who shot him was back on duty four days later, during which he was involved in a second fatal shooting. He’s now on paid desk duty. More here.
    Digg it |  reddit |  del.icio.us |  Fark
  • 63 Responses to “Morning Links”

    1. #1 |  MacGregory | 

      After a re-read I withdraw my last comment

    2. #2 |  Steve Verdon | 

      Sorry Lee (#2) that is just wrong.

      Are we going to get health care reform? Yes. Would Wal-Mart want to have some say in how it is forumalted? Yes. Why not do it so as to hurt your competitors. If Target has lower costs, then a federal mandate that has higher costing requirements but lower than Wal-Marts is a smart move. I’d also suggest you read up on how guys like J.D. Rockafeller made his money. A large part of his success depended on getting sweetheart deals from rail roads so that he could undercut his competitors. Another way to do that is to favor legislation that raises your competitors costs.

      Regarding dogs:

      Suppose we have 1000 trials, and that 10 of them are postive (bombs, drugs, etc.). What we want to know is:

      Prob(Bomb|Alert)

      We can find this by looking at

      Prob(Alert|Bomb)Prob(Bomb)/Prob(Alert)

      The probability of finding a bomb is thus proportional to the probablity that there is a bomb and the porbability that the dog alerts when there is a bomb.

      Now if the probability that the dog alerts in general is 50%, then the probability that there is a bomb when the dog alerts is equal to the probability of there being a bomb. Using our numbers a 1% chance. In this case, the dog offeres no help in finding the bomb. If the probability of an alert is dropped to say 5% and the probability of the dog alerting when there is a bomb is 80% then we have the following

      Prob(Bomb|Alert) = 0.16

      In other words, even a dog that doesn’t alert that often, and alerts 80% of the time when there is bomb will still likely have a poor track record.

      Unless I’ve messed up my math.

      Based on this I’d say that dogs should be evaluated by an independent agency. Dogs should be checked both in terms of how often they alert when there is a bomb, how often they alert when there isn’t a bomb, and from this determine how often the dog alerts in general.

      What also needs to be kept in mind is that a dog can alert even if the item is not there. That a dog alerts does not mean that there is something there. Also dogs should not be trained for cross purposes. Train a dog to detect bombs or drugs not both. If a bomb detection dog alerts but you find no bomb and drugs, I’d argue that the evidence should be inadmissable since the probably cause due to the dog’s “alert” was for explosives not pot or whatever.

      Of course, it is a bit more problmatic when its the other way around, a drug dog alerting and finding a bomb. Do we really want a bomb builder/terrorist to be let lose? Hopefully this is sufficiently unlikely that it wont happen.

    3. #3 |  Steve Verdon | 

      MacGregory,

      Technically you are on the right path, but just a tad bit off. Yes, we want to use the data we have in hand. However, the dog is allowed to “inspect” all 100 cars, so that is part of our data in hand so to speak. Using field data is tough since we wont know about those cases where the dog has didn’t find drugs or a bomb when indeed there was one.

      I’d also add that dogs should go through not only re-tests, but also refresher training.

    4. #4 |  Light | 

      I’m actually allergic to Acetaminophen and a high dose would probably kill me, however, I’m aware of this and never get anything with Acetaminophen in it…. and its in EVERYTHING.

    5. #5 |  Lee | 

      Doctors cannot prescribe straight hydrocodone for you to pick up at your friendly neighborhood pharmacy.

      I have a couple of pill containers that would disagree with that statement.

      Steve:
      Yep, you’re right I screwd up the quick reading I gave the link. I thought WalMart was backing the a reform plan, not an employer mandate.

    6. #6 |  Aresen | 

      The cop who shot him was back on duty four days later, during which he was involved in a second fatal shooting. He’s now on paid desk duty.

      If I were the coffee boy in his office, I’d be very careful when bringing him his order.

    7. #7 |  Alex | 

      The consultant in the GfB story is based in Austin. Everyone in Austin is high, so if you bring the drug dog he’s going to smell the joint they just smoked. Considering Austin is full of stoned retards, I’d bet that about 52% of smokers smoke with their sacks in the car.

      Also, I had a run-in awhile ago in Austin with a drug dog. I was one of the 48% who doesn’t carry weed while the car smells like weed so no harm, no foul. But that dog immediately identified where the unsmoked blunt was kept and the ashtray. It was quite impressive really. The APD K9 officer asked me (away from the others) when was the last time the truck had been smoked in and where any weed was stored. He said it was to track the dog’s performance and make sure he’s sharp and really had no reason to lie. I’ve had one other run-in with a K9 in the middle of nowhere on the interstate, and the thing went down exactly the same.

    8. #8 |  Alex | 

      As to the statistics, trying to define an exact error rate is pointless. As with all statistics, you can manipulate the numbers to be in your favor. To get a meaningful evaluation, you have to define your objective before seeing the numbers. So a bomb dog that suspects 2 in 10,000 travelers of having a bomb, and he’s right on one of those is pretty awesome (even though some call that 50% error). But if he lets 100 bombs get by, that K9 handlers salary could be better spent in other places.

      The problem with drug dogs and evaluating their effectiveness is that they’re really good at smelling recently smoked pot (and other drugs) and drug users are really good at saying the dog is lying. What drug dogs are not very good at, even though it’s their stated purpose, is detecting 5 kilos of professionally packaged coke.

    9. #9 |  Bronwyn | 

      Lee, you forgot to quote the second sentence, which pointed out that it can be done, but at the risk of having the DEA breathing down the doctor’s neck.

      Even in the absence of an actual threat from the DEA, so many doctors are sufficiently afraid of the perceived threat that they refuse to prescribe opioids properly.

      You, Lee, seem to have hit the physician jackpot. I am sorry to hear you’re in need of pain meds.

    10. #10 |  supercat | 

      Sniffing dogs can be good and accurate if (1) the handlers want them to be accurate, and (2) nobody is trying to jinx them into yielding false positives. In cases involving e.g. getting victims out of collapsed buildings, the dog’s handlers want to avoid erroneous alerts, and so the dogs are probably reliable. In cases where a dog’s claimed alert can be used to claim PC for a search, the dog’s handlers are not opposed to erroneous alerts, and so the alerts are not as reliable.

    11. #11 |  old | 

      458 deaths per year sounds almost like a rounding error.

      Sure a rounding error, you’re a funny guy Balko, but what about the estimated 56,000 emergency room visits, 26,000 hospitalizations. What do those numbers do to the cost of insurance and health care in America?

    12. #12 |  Bronwyn | 

      More important is what those ER visits and hospitalizations do to quality of life.

    13. #13 |  Lee | 

      ANY drugs you take will have side effects.

      It is a very rare drug that will not kill someone.

      The numbers involved with aceto are very very low.