Sunday, October 15, 2017

Sunday Bullets

Monday, September 25, 2017

Old Pink In New Bottles: Caroline McCarthy's Failed Bromides

Caroline McCarthy's Medium piece rings every klaxon almost immediately. Her complaint that Damore doesn't use collaborative work as an attraction to women is possibly reasonable, but the underlying justifying link to the National Coalition of Girls' Schools is so full of cant and repeatedly failed approaches, it's impossible to take seriously. "Seeing women’s historic contributions inspires today’s girls", we are told, yet does no one remember the beatification of Ada Lovelace? Of Grace Hopper? And yet, since the mid-1980s, the overall fraction of women in CS has been in decline. She accuses Damore of using research that "was perhaps informed by the agenda-driven pseudoscience that permeates the deepest dregs of Reddit and 4chan "; if you can't attack the man's footnotes, why not manufacture a fantasy list of enemies he's in bed with? (She does correctly mention his bizarre tweets about the KKK, but they weren't on the scene here.)

She states, without any justification, "There’s no question that we need more female computer scientists." As ever, my reaction to this is, why? Why should we have to tailor entire curricula to the needs of people who have no apparent interest in the subject? She cites Stuart Country Day high school as an all-girls' program that has tailored their approach to women in computer science, but what is their track record there? That is, have they had any actual success getting girls who otherwise are not interested in computer programming into the field? Or did they end up like the author, who found it "so un-engaging and isolating and boring that I dropped it before it could bring down my GPA"?

Eventually, she confesses that "there was merit to quite a few of the points James Damore raised, and discrediting the research he cites (rather than simply disagreeing with his conclusions) will hurt rather than help women’s advancement in computer science." Coming late as it does, this seems like so much belated and minimal acknowledgment of the obvious; it recalls Cordelia Fine's sleazy tactics in Testosterone Rex. The lure and futility of pink lacquer continues.

Update 2017-09-26: I didn't spend a lot of time digging through her links, but I want to focus on her cite of the National Council of Girls' Schools in reference to this passage:
The world is desperately seeking to plug the leaky STEM pipeline from its shortage of women, and girls’ schools are playing a critical role. Girls’ schools lead the way in graduating women who become our nation’s scientists, doctors, engineers, designers, and inventors. Girls’ school graduates are six times more likely to consider majoring in math, science, and technology and three times more likely to consider engineering careers compared to girls who attend coed schools. Why? Because girls’ schools support collaboration and all-girl classrooms foster female confidence and aspirations.
The underlying link about considering engineering careers (see p. 38) says that "Engineering also produces the largest single-sex/coeducational differential when it comes to career choice, where 4.4 percent of women from single-sex independent schools aspire to become engineers, relative to 1.4 percent from coeducational schools." In other words, whatever boost such education may yield, it comes nowhere close to reversing the 20% female matriculation rate in CS and engineering disciplines, or the ten times figure needed to surmount female frustrations in the university and subsequent job search process (assuming we take results as representative, which they may not be). And as McCarthy observes, this solution does not scale, for the simple reason that Freddie deBoer raised: terrific outcomes in education almost invariably stem from selection bias. In this case, the kinds of girls who can afford to go to all-girls schools have families with means to afford tuition.

But ultimately, it seems to me that the most salient test of Damore's thesis is and remains the fact that the work is compelling unto itself for men, but not for women. If, as McCarthy suggests, she's only ten years away from her collegiate days, why not have a go at it again? The world isn't lacking for outlets for talented coders; yet she stays out of the business. Why? The answer seems obvious: either the work is its own reward, or it is not. For McCarthy, and many women, it is not.

The second thing at the NCGS website is a discussion of "growth mindset", a topic that has had a rather difficult and muddled empirical and philosophical history; one recent (n=624) study even shows
Children’s own mindsets showed no relationship to IQ, school grades, or change in grades across the school year, with the only significant result being in the reverse direction to prediction (better performance in children holding a fixed mindset). Fixed beliefs about basic ability appear to be unrelated to ability, and we found no support for mindset-effects on cognitive ability, response to challenge, or educational progress.
 From the outside, "growth mindset" looks like a smoke and mirrors foundation upon which to build such dubious concepts as "stereotype threat", which itself has had problems with reproduction. In the end, these have little explanatory power next to the simple story McCarthy herself tells: disinterest.

Betsy DeVos Rescinds "Dear Colleague" Letter Title IX Guidance

As you've probably heard by now, Betsy DeVos has rescinded the infamous "Dear Colleague" letter charging universities to investigate sexual assault cases. Unsurprisingly, California has passed a law retaining the old standard (SB 169 text),  and a number of universities will either defend the old standard or even adhere to it. The show ain't over, but it's a serious step in the right direction.

iOS 11 Early Returns

Some first thoughts on iOS 11:
  • Appears to fix problems with Reminders not synchronizing in the cloud (but let's give it a while, this always worked well in the first day or so after a reboot).
  • UI changes are a mixed bag:
    • Round buttons for the calculator vs. a grid before, meh. This appears in multiple locations.
    • Email has a much more visible (but real estate-hungry) bold title for each mailbox.
    • The Control Center is now super busy, partially reverting to the status quo of iOS 6 (and 7?) where individual apps all had control of audio out (the AirPlay logo that occupied a tiny piece of each app's real estate). I find this not only confusing but annoying; if volume is still in the Control Center, why not audio out? 
  • The ability to select which part of a Live Photo (still photos are now stored as short video clips) looks like it should be great, and overdue; this should have been rolled out once Apple started storing stills as Live Photos.
  • The Health app still sucks, failing to deliver core functionality that has been part of FitBit and other fitness wearables for years. Particularly, the only place one can set goals is inside the Apple Watch's tiny user interface. A horrible, horrible design flaw that makes the Apple Watch almost unusable as a fitness tracker.
  • Air Play 2 adds a functionality long available in Sonos products, the ability to play the same music on multiple speakers throughout the house. This is better developed in MacOS iTunes 12.7.0, but the fact that they're doing it at all is a good sign.
  • Battery life takes a huge hit, with Wandera reporting battery consumption twice the rate of iOS 10.
  • I have also had problems with iOS 11 interacting with my Ford SYNC for music. SYNC will play any given song for 20 seconds and then stop. It exacerbates a long-extant problem forgetting which song in which playlist is active, and starting at the top of the alphabetical song list when first plugged in to the truck's USB port by doing this if you push the vehicle's play button after the song stops. A horrible failure that can be worked around by using Bluetooth audio at the cost of some (mostly unnoticeable) compression.

    Update 15:44: It appears that this problem is now gone.

Friday, September 22, 2017

Jerry Coyne Reviews Cordelia Fine's Testosterone Rex

Pretty much what you'd expect, leaving the deets to Stuart Ritchie's review.
Before I started TR and then while I was reading it, I wrote two posts (here and here) about Fine’s claim that there’s no evolved differences in male and female behavior. I also criticized her completely muddled and erroneous claim (based on bogus statistics) that sexual selection doesn’t work because the “Bateman experiment”—showing a greater variance in reproductive success among male than among female fruit flies—was wrong. Well, it wasn’t wrong, it was inconclusive, and later work, as Ritchie notes, has supported the sex difference in reproductive-success-variance that’s a crucial assumption of sexual selection. Bateman’s result was just a one-off that tells us nothing. Sexual selection is alive and well, and supported by tons of data. Nevertheless, Fine’s argument, which is really dumb if you know even a bit of biology and math, persuaded many people, including a Guardian reviewer, and Ritchie takes it apart in his review.
In that second link (which I missed earlier), Coyne quotes a review by the politically-motivated P.Z. Myers ("Myers has always rejected biology that is ideologically unpalatable to him"):
In a rare occurrence at his site, the commenters, usually a choir of osculatory praise, gave him pushback. In fact one,  “Charly”, did the math correctly and showed that males in relationships with multiple females (bigamous or polygamous) have the potential to have more offspring than do monogamous males, supporting the ideas that men are selected to compete for women. (Duh!) Charly ended his calculations with this statement: “But maybe my reasoning and math is wrong, I am sure someone will point flaws out.”

In the next comment, Myers admitted that Charly’s math was actually right—math that invalidates Fine’s argument—but then he said this:

And there we have it, ladies and gentlemen, brothers and sisters: an admission that the biology is right, at least in theory, but the person who did the calculations is immoral.
This is what we're up against. 

Wednesday, September 20, 2017

California's Dreadful AB 485 Masks And Exacerbates The Problem Of Puppymills

My friend Linda Kaim pointed me at an AKC complaint that California's AB 485 would cost people the "freedom to choose your own dog". As usual, it pays to be skeptical of anything from the AKC, and a quick read of the text of the bill shows that the circumstances behind it are anything but benign:
122354.5.  (a) A pet store operator shall not sell a live dog, cat, or rabbit in a pet store unless the dog, cat, or rabbit was obtained from a public animal control agency or shelter, society for the prevention of cruelty to animals shelter, humane society shelter, or rescue group that is in a cooperative agreement with at least one private or public shelter pursuant to Section 31108, 31752, or 31753 of the Food and Agricultural Code.

(b) All sales of dogs and cats authorized by this section shall be in compliance with paragraph (1) of subdivision (a) of Section 30503 of, subdivision (b) of Section 30520 of, paragraph (1) of subdivision (a) of Section 31751.3 of, and subdivision (b) of Section 31760 of, the Food and Agricultural Code.

(c) Each pet store shall maintain records sufficient to document the source of each dog, cat, or rabbit the pet store sells or provides space for, for at least one year. Additionally, each pet store shall post, in a conspicuous location on the cage or enclosure of each animal, a sign listing the name of the public animal control agency or shelter, society for the prevention of cruelty to animals shelter, humane society shelter, or nonprofit from which each dog, cat, or rabbit was obtained. Public animal control agencies or shelters may periodically require pet stores engaged in sales of dogs, cats, or rabbits to provide access to these records.

(d) A pet store operator who is subject to this section is exempt from the requirements set forth in Article 2 (commencing with Section 122125) of Chapter 5, except for the requirements set forth in Section 122135, paragraphs (3) and (4) of subdivision (a) of, and paragraphs (5) and (6) of subdivision (b) of, Section 122140, and Sections 122145 and 122155.

(e) A pet store operator who violates this section shall be subject to a civil penalty of five hundred dollars ($500). Each animal offered for sale in violation of this section shall constitute a separate violation.

(f) For purposes of this section, a “rescue group” is an organization that is tax exempt under Section 501(c)(3) of the Internal Revenue Code, and that does not obtain animals from breeders or brokers for compensation.
 As we've already seen, there's a lot of ways to hide revenue in 990s, including as executive salaries and other compensation. The way this bill is written would appear to permit puppy mills to rebrand themselves as "rescues" (or to create new intermediary "rescues") who would launder the dogs, shifting the actual profit center. Because of the perverse way USDA APHIS rules regulate commercial dog breeders, ironically it would be small-scale breeders (PDF) who would be most affected by the new language:

Q. Under the final rule, what is the new definition of a retail pet store?

A. In the final rule, “retail pet store” means a place of business or residence at which the seller, buyer, and the animal available for sale are physically present so that every buyer may personally observe the animal prior to purchasing and/or taking custody of it after purchase. [Emboldening in the passage above is mine. - RM] By personally observing the animal, the buyer is exercising public oversight over the animal and in this way is helping to ensure its health and humane treatment. Retailers who sell their pet animals to customers in face- to-face transactions do not have to obtain an AWA license because their animals are subject to such public oversight. Under the AWA regulations, a “retail pet store” is also a place where only the following animals are sold or offered for sale as pets: dogs, cats, rabbits, guinea pigs, hamsters, gerbils, rats, mice, gophers, chinchillas, domestic ferrets, domestic farm animals, birds, and coldblooded species.
Essentially, this eliminates small-scale breeders from selling in California, because their puppies are not from a "rescue", and because they meet the APHIS definition of "retail pet store". That this is approximately insane is par for the course; it is, after all, California.

Cordelia's Not-So-Fine Book Testosterone Rex Wins Royal Society Prize

No, seriously, Cordelia Fine's Testosterone Rex won the annual Royal Society science book prize.  Fine staunchly advocates the blank slate theory of human sexual differences, i.e. that cultural norms drive behavioral and other visible differences. That Fine's book is the purest tosh has been addressed elsewhere, as for example West Hunt's long-form review, Stuart Ritchie's review in Quillette, or Jerry Coyne's criticisms.  She botches her analysis of, the subsequent meaning of, and corrections/improvements to Angus Bateman's groundbreaking fruit fly research in the 1940's. She ignores research on congenital adrenal hyperplasia on other mammals, including among our near evolutionary cousins, the great apes. (Girls born with this condition secrete unusually large amounts of testosterone, and exhibit more male-typical play.) Ritchie notes that she elides significant criticism:
This fits into a pattern: evidence contrary to Fine’s position is often cited, but it’s not mentioned in the text, instead being relegated to endnotes where it can’t cause too much trouble. Witness, for instance, Fine’s mention of “stereotype threat”, where a single supporting study is discussed in the text but a contrary meta-analysis is only mentioned in the endnote. Or her discussion of a 2015 paper on how males’ and females’ brains aren’t essentially different, but are a mosaic of features: you wouldn’t know that four strong scientific critiques of the study had been published (with a response) unless you flick to the back of the book. This allows Fine to use the main text to critique only the most overblown claims about sex differences, and avoid having to deal at length with more reasonable arguments. of "stereotype threat": "a single supporting study is discussed in the text but a contrary meta-analysis is only mentioned in the endnote".
It seems foreordained that this award will serve as a significant data point for the notion that politics has corrupted the academy. Or will it? The panel issuing the award consisted of
  • Richard Fortey (paleontologist, committee chair)
  • Naomi Alderman (novelist)
  • Claudia Hammond (broadcaster, writer, social psychologist)
  • Dr. Sam Gilbert (brain researcher)
  • Shaminder Nahal (broadcast TV editor)
That is to say, it is predominated by non-scientists. How this might have happened would prove an interesting story, but it is not the same thing as a vote by the bulk membership (as with the Hugo Awards). Yet it points in a disturbing direction, cloaking badly-executed science with the imprimatur of the Royal Society. I hope this draws a strong reaction from actual biologists.