Fooling the Classifiers (part 2)

When media outlets get called out for misjudging a character, they tend to overcompensate with over-the-top villainization. Just look at the narrative reversals of Weinstein, Epstein, Sheryl Sandberg, Adam Neumann, Corona-chan, and yes, Elizabeth Holmes.

Nope nope nope, we never claimed the coronavirus was no deadlier than the flu, and you certainly won’t find our deleted Tweets stating such!

Maybe some of these people really are cartoon villains, but I’m sympathetic to Theranos. Right idea, wrong execution.

Build: Medical Devices

Here’s people in China taking a round of COVID serology tests before a group dinner, like some sort of aperitif.

Why can’t we have this in the United States? COVID-19 antibody tests currently have a low sensitivity rate, meaning they deliver a lot of false negatives. An inaccurate test result might lead to a false sense of security. The obvious solution is to test more, and more frequently, but public health officials think Americans are stupid. To keep us safe, government health agencies prevent anyone from getting tested at all. And that is why it took 25 years for the FDA to approve an over-the-counter HIV test.

It’s possible to fool the classifiers. Just modify the intended use. For example, absorbable sutures are considered a Class III medical device, on par with pacemakers and defibrillators. To get around clearance requirements, manufacturers call their product a “practice suture”, for training and taxidermy. Now they can sell the sutures for less than twenty bucks on Amazon. The sutures are packaged in sterile alcohol, because animal carcasses care about germs. Wink wink nudge nudge.

Another example is 23andMe, a DNA test that maps your personal genome. The original inception provided information about a person’s risk for diseases like Parkinson’s or cancer, but the FDA shut them down for a few years because any kit that diagnoses disease must first gain clearance as a medical device.

To get around FDA restrictions, 23andMe pivoted from health diagnostics to measuring intersectionality. You could be 1/1024th Cherokee and not even know it — Buy our DNA test kit and claim your minority status now! Hence the series of identity politics commercials featuring customers who discover their genetic victimhood.

So here’s how we get home COVID-19 serology tests out the door: Claim they’re for measuring disparate impact, or some similarly woke purpose. If the FDA complains, tell them to take it up with the SPLC.

(To be continued…)

9 thoughts on “Fooling the Classifiers (part 2)

  1. The media response to both the Epstein and Weinstein affairs are examples of the general failure of the American thought process.

    Some keyboard operator for the NYT mentions that Epstein related to him in an interview that a disapproval of older men having sex with younger women is a “cultural aberration” and omigod that sealed the deal as Epstein as a perv. Too bad that perv Epstein was actually correct in that observation.

    The real culprits in Epstein’s transgressions were these “girls” parents. “Children” are supposed to be raised and protected by their parents, who don’t generally let them jump on private jets to Caribbean islands with strangers. In contemporary USA childhood now extends into the early thirties so there are maybe a lot more orphans than in the past. Maybe that’s the problem.

    In the case of physically and visually repulsive Harvey, any female that talked to the sleaze had to know that doing business with him involved surrender of some form of chastity but they held their nose in hopes of a big screen appearance and following fame and fortune. There’s a price to be paid if one wishes to join a club that celebrates sex as currency.

    Another nugget of information that seems to have slipped through the media sieve is the real genesis of all this moralizing. That would be the fact that it’s been exposed by Ronan Farrow, son of Mia Farrow, who married super-star crooner Frank Sinatra when she was 21 and he was 50. Ronan Farrow, born in 1987, was possibly the son of writer/director/actor Woody Allen but who knows? The genetic pathways of the Hollywood elite are a maze that extends to Manhattan.

    1. I remember reading Samantha Geimer’s memoir many years ago (the 13-year-old girl raped by Roman Polanski). She doesn’t see herself as a victim. Her telling was that it was normal for young girls to hook up with much older Hollywood producers, it was almost like a privilege. Weird how her story never got rehashed in the #MeToo crusades.

  2. Wikipedia says you’re wrong about the meaning of “sensitivity.”

    You write, “COVID-19 antibody tests currently have a low sensitivity rate, meaning they deliver a lot of false negatives. An inaccurate test result might lead to a false sense of security.”

    But Wikipedia says, “Sensitivity (also called the true positive rate, the recall, or probability of detection in some fields) measures the proportion of actual positives that are correctly identified as such (e.g., the percentage of sick people who are correctly identified as having the condition).”

    https://en.wikipedia.org/wiki/Sensitivity_and_specificity

    If Wikipedia is right, then a low sensitivity rate would indicate a high rate of false positives. A low specificity rate would indicate a high rate of false negatives.

    A low specificity rate on an antibody test would indicate that many test-takers don’t have antibodies to the virus when they actually do, which would deprive those test-takers of a true sense of security, assuming that knowing they have been infected gives them a sense of security.

    1. no. low sensitivity means the test misses a lot of true positives — ie, it is not sensitive enough to detect weak positives. Hence a false negative. I used to work for a company designing pregnancy tests for livestock, trust me on this one :b

      1. Thanks very much for your reply.

        If you go to this CDC site and open up the question which asks: “Under what circumstances should laboratories use either a SARS-CoV-2 viral or serology (antibody) test that has received EUA from FDA?,” part of the answer states: “Most of the PCR-based tests that use two or more targets are likely to have high specificity (few false positives).”

        https://www.cdc.gov/coronavirus/2019-ncov/lab/testing-laboratories.html

        The implication of this sentence is that specificity deals with the proportion of false positives, which further implies that sensitivity deals with the proportion of false negatives. These definitions are consistent with your definition.

        On the other hand, here are links to two BMJ articles that seem to imply that specificity deals with false negatives and that sensitivity deals with false positives.

        Understanding sensitivity and specificity with the right side of the brain
        https://www.ncbi.nlm.nih.gov/pmc/articles/pmc200804/
        https://pdfs.semanticscholar.org/e4ae/e71d6545d432d30520119ca38ea13582ab90.pdf

        Diagnostic tests. 1: Sensitivity and specificity
        https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2540489/

        Therefore, I am still confused. Are all of these definitions by you, the CDC, the BMJ articles and Wikipedia consistent and saying the same thing, which would mean that I am just not understanding this issue at all? Or are you and the CDC saying something which is inconsistent with Wikipedia and the BMJ articles?

        1. I think all the definitions are consistent, and I agree with them. A test is basically a list of characteristics that a sample has to match in order to render a positive. If the list is specific about the characteristics, then you will have few false positives. The test can “cheat” and achieve 100% specificity by always returning negative.

          A very sensitive test would detect all the positives, leaving few false negatives. Again, you can “cheat” and get 100% sensitivity by always returning positive. That’s why both numbers matter.

          The first paper says: “sensitivity = TP/TP+FN, where TP is the number of true positives and FN is the number of false negatives.”
          The second paper says: “sensitivity = true positives correctly identified; specificity = true negatives correctly identified”

          So yeah, it all works out the same.

Leave a Reply