New APPS

This time Margaret Mitchell, one of the other authors on the fabulous "Stochastic Parrots" paper (that's my post on it.  The paper is here) on natural language processing.  This was obviously coming, since they'd suspended her email account weeks ago.  In case you haven't read the paper (and you really should!), it's worth mentioning that the paper does not mention Google, and the goal is to point to strategies to make AI better, i.e., less likely to entrench racism, sexism and so forth while destroying the planet in climate catastrophe.  It turns out that those are related, because bigger and bigger datasets are both carbon intensive and likely to pick up on and magnify hegemonic and toxic speech.  Attention to efficiency and curation moves the needle on both.

No commitment to even looking like they care about ethical AI.

Knee pain is common and debilitating, and it’s often caused by osteoarthritis in the knee.  Treatment options range from analgesics (including opioids) to knee-replacement surgery.  If you go to the doctor with arthritic knee pain, you can get an x-ray which can then be interpreted using standard rubrics like the Kellgren–Lawrence Grade (KLG) to quantify damage to your knee and then guide treatment options. The KLG isn’t perfect in that the correlation between pain and objective scores of damage to the knee isn’t perfect.  Some people’s knees are a wreck and they report no pain; others have pain beyond what their KLG score indicate. But here’s the thing: Black patients consistently report more knee pain than white patients.  They also tend to have more knee damage on the KLG – but even when you factor that in, Black patients report much more knee pain than white patients with comparable KLG scores.  What’s going on?

One possibility is that factors external to the knee – stress, for example – explain the higher pain.  If that’s the case, then patients need less knee treatment.  But what if their knees were in worse shape?  To answer that question, you’d have to ask yourself what in an x-ray indicated poor knee condition.

Disease is often measured through indicators, and we know that these indicators can lead to all sorts of complexity.  In the context of Covid, for example, there are all sorts of questions about testing and sensitivity that I’ve talked about before.  Along the way, I referred to a fantastic paper on malaria testing in sub-Saharan Africa – suffice it to say that “cases of malaria” reported to donor organizations is a difficult number to parse for reasons having to do with vagaries in testing and diagnosis.

In a new paper in Nature Medicine, a team led by Emma Pierson makes ingenious use of artificial intelligence to tackle the problem of racial disparities in knee pain.  Since algorithms and data are so often implicated in increasing or magnifying racial disparities (see, for example, Safiya Noble on Google, or Timnit Gebru on facial recognition, or Margaret Hu’s chilling “Algorithmic Jim Crow”), it’s encouraging to learn about machine learning working to undermine racial disparities.  Ordinarily, you train an algorithm to perform like an excellent clinician.  In this case, that would mean training it to look at radiography and determine the correct KLG score.  The trick here was to instead train it to look at pain: to determine what features of the x-ray predicted that the patient would report pain.  It turns out that the algorithm’s diagnoses reduced racial disparities in diagnosis by a jaw-dropping 47%.

(more…)

To review the issue: The Oxford/AstraZeneca vaccine uses a modified adenovirus, as do several other vaccines in development, most notably the Russian Gamaleya Institute one.  Early, puzzling results suggested that the Oxford vaccine was about 70% effective overall, but that the overall number obscured a disparity between two groups: a two-dose group of all ages that showed efficacy in the 62% range but 90% in a group (including no elderly people) that had received a half-strength dose first.  Eh?  Even more strangely, the half-dose group seemed to be have given that dose… by mistake?  What happened?

Reuters does a deep dive.  There's a lot of threads, but it appears that the problem began when the Oxford team didn't trust measurements of the strength of a batch of vaccine from an Italian manufacturer.  The Oxford team then measured it using a different technique, concluded it was more potent than the manufacturer said, and trusted its own measurement.  Guess who was right? [Time to scream at the void: WHY on earth, given discrepancies between two measurements, both supposedly reliable but using different techniques, WOULD YOU NOT MEASURE AGAIN UNTIL YOU FELT VERY GOOD ABOUT THE DISCREPANCY AND HOW IT HAPPENED?]

"Oxford’s measurement showed that the batch was more potent than the Italian manufacturer had found, the documents show. Oxford trusted its own result and wanted to remain consistent with a measuring tool it had used throughout an earlier trial phase. So it asked Britain’s drugs regulator for permission to reduce the volume of vaccine injected into trial participants from the K.0011 batch. Permission was granted …. The documents published in The Lancet confirm that the error lay with the Oxford researchers. A common emulsifier, polysorbate 80, used in vaccines to facilitate mixing, had interfered with the ultraviolet-light meter that measures the quantity of viral material, according to the documents. As a result, the vaccine’s viral concentration was overstated and Oxford ended up administering half doses of vaccine, believing they were full doses."

This vaccine matters because it is one that developing countries are depending on: it is cheap, and it is stable in refrigerator temperatures, which makes the logistics a lot easier.  There is a hypothesis that explains the half-dose/full-dose discrepancy: it's possible that the lower first does primes the immune system better than the higher one; adenovirus variants also circulate in some human populations, and so it's possible that some combination of prior infection and the first dose generated an immune response sufficiently robust that participant's own immune systems destroyed the second dose's virus before it could generate further immune response.   The Russians report a 90% efficacy, and that vaccine uses two different adenovirus vectors across the two doses to avoid precisely this risk.  These questions need more data, and we're still waiting for the full results of ongoing trials.  Adenovirus vaccines matter too: in addition to Oxford, Gamaleya and a Chinese vaccine, the single-dose Johnson & Johnson vaccine, with trial results expected in January, is also adenovirus.  In the meantime, Gamaleya is sharing their adenovirus vector, in order to enable a combined trial: one does of the Russian vaccine, and one of the Oxford/Astrazeneca one.

 

 

By Gordon Hull

In an important recent article, Robin Kar and Margaret Radin propose a way to interpret the volumes of boilerplate that accompany pretty much any electronically-mediated consumer transaction.  Rather, they propose a way to interpret the phenomenon of the deluge of such boilerplate.  We all know the scenario: you decide to buy a song for $0.99 on a site called SketchyFiles.com, and at some point in the process, you click to indicate your acceptance of the “terms and conditions.”  Did you read those terms and conditions?  Of course you did not: if you were to print them, they’d probably run in excess of 30 pages, most of them using language that you don’t understand.  It’s not rational to slog through and try to understand all of that for a 99-cent purchase!

But the flip side is that SketchyFiles is very much going to interpret this as a contract when it suits them.  Say, for example, that you think that your purchase included a virus that destroyed a couple of files on your computer. You spend a few minutes online, and discover that there are 50,000 people to whom this exact thing happened!  You lawyer up and file suit against SketchyFiles, including a request for class certification, since your files aren’t individually worth a lot, but the aggregate of them are, and you think SketchyFiles ought to have to own the problem and scan their songs for viruses.  You will promptly discover in court that the “contract you signed” included a “mandatory arbitration clause,” which says that you agree that any and all disputes involving your purchase are to be settled out-of-court, using an arbitration procedure and a venue chosen by SketchyFiles.  Not only that, you’ve probably also contractually agreed that no class certifications are possible: all users’ claims must be adjudicated one at a time.

(more…)

At least not from the appearance of things.  Google summarily fired Timnit Gebru, one of its lead AI Ethics researchers and one of the few Black women in a leadership position at the company.  Her sin?  Producing academic research critical of biases in AI:

“The email and the firing were the culmination of about a week of wrangling over the company’s request that Gebru retract an AI ethics paper she had co-written with six others, including four Google employees, that was submitted for consideration for an industry conference next year, Gebru said in an interview Thursday. If she wouldn’t retract the paper, Google at least wanted the names of the Google employees removed.  Gebru asked Google Research vice president Megan Kacholia for an explanation and told her that without more discussion on the paper and the way it was handled she would plan to resign after a transition period. She also wanted to make sure she was clear on what would happen with future, similar research projects her team might undertake. …. The paper called out the dangers of using large language models to train algorithms that could, for example, write tweets, answer trivia and translate poetry, according to a copy of the document. The models are essentially trained by analyzing language from the internet, which doesn’t reflect large swaths of the global population not yet online, according to the paper. Gebru highlights the risk that the models will only reflect the worldview of people who have been privileged enough to be a part of the training data”

The company responded by “accepting” her “resignation,” effective immediately (while she was on vacation!).

Let’s be clear: this sort of research is what Gebru does, and she’s very good at it.  Before she got to Google, she co-authored a paper that proved that facial recognition software misrecognized people of color (especially women) at much higher rates than white men.  The paper is influential, and supported efforts like the one that eventually got Amazon to stop selling its facial recognition software to police. Why is Google surprised that she would pursue this topic now?  Also, the link between biased training data and crappy outcomes is a thing.  A thing that a company like Google ought to care about.

Gebru’s unceremonious firing suggests that Google isn’t seriously committed either to actual critical thinking about AI and ethics, or to dealing with Silicon Valley’s glaring diversity problems.  I suppose file under disappointing, but not surprising.

UPDATE 12/6: For more on the mess in Rhode Island, see here.

This week’s SCOTUS opinion overturning New York’s restrictions on religious gatherings is disappointing in many ways. Most obviously, it hamstrings the ability of governors to respond with science to Covid and is part of a conservative backlash to that effort; Justice Breyer makes some of the important points in dissent, followed by more of them in Justice Sotomayor’s dissent. More ominously, it signals the accelerating arrival of the era of having to pander to right wing religious people who think that their religious freedom outweighs the harm principle (see, Thanksgiving is not just about spreading disease. It’s about imposing religion!). This has been coming for a while – recall the Court’s endorsement of Hobby Lobby’s theory that it’s ok to freeload on public services and goods, but that it does not have to extend contraception rights to citizen-employees. But Amy Coney Barrett makes it faster. That is, part of the acceleration is that the Court is illegitimate, by Mitch McConnell’s own standard: the rule is either that you seat a justice before an election or not. Those can’t both be true, except in a shameless right-wing power grab. We’ll get back to Moscow Mitch in a minute.

Also unfortunate here is Governor Cuomo, who seems to have gone out of his way to antagonize the Hassidic community in particular. The order in question also apparently did not index capacity to the size of the building or social distancing plans. The antagonism plus the restrictions lets the Court make disingenuous comparisons between the number of people in a grocery store and a church. It also gives people like Gorsuch the opportunity to say “the Governor is remarkably frank about this: In his judgment laundry and liquor, travel and tools, are all “essential” while traditional religious exercises are not. That is exactly the kind of discrimination the First Amendment forbids” (2). No one should view this as anything other than opportunism; after all, Gorsuch signed onto the Court’s opinion ignoring Trump’s many, many anti-Muslim tweets on the way to upholding his Muslim travel ban.  Politicians’ chatter is only part of the law when you don’t like the politician, it seems.

These are important, but what I haven’t seen discussed is that this opinion intersects with the other news percolating up around Thanksgiving, which is the absolute incoherence of public health restrictions. Folks like Ellie Murray and Julia Marcus have been trying to sound the alarm on Twitter, and Amanda Mull wrote about it in the Atlantic. The public health message is basically: “Thanksgiving is dangerous! It is dangerous to sit around a table with people from outside your household for a long meal! If you want to do that safely, go to Cheesecake Factory!” Ummm… in other words, it is bonkers incoherent that in most of the country, indoor dining at restaurants is still possible when pretty much anything else is closed. Indoor dining is on the same epidemiological list as church for likely spread!

In this particular context, is it any surprise that a group of people who are predisposed to prioritize religious rights noticed something amiss? This is from the per curiam:

“In a red zone, while a synagogue or church may not admit more than 10 persons, businesses categorized as “essential” may admit as many people as they wish. And the list of “essential” businesses includes things such as acupuncture facilities, camp grounds, garages, as well as many whose services are not limited to those that can be regarded as essential, such as all plants manufacturing chemicals and microelectronics and all transportation facilities.” (2)

The invitation for activist judges to ignore science is too much for them to pass up. Here is Kavanaugh: “in a red zone, for example, a church or synagogue must adhere to a 10-person attendance cap, while a grocery store, pet store, or big-box store down the street does not face the same restriction.” (2). Gorsuch piles on: “Nor is the problem an isolated one. In recent months, certain other Governors have issued similar edicts. At the flick of a pen, they have asserted the right to privilege restaurants, marijuana dispensaries, and casinos over churches, mosques, and temples” (3). And he concludes: “It is time—past time—to make plain that, while the pandemic poses many grave challenges, there is no world in which the Constitution tolerates color-coded executive edicts that reopen liquor stores and bike shops but shutter churches, synagogues, and mosques.” (7).  Kavanaugh again: “once a State creates a favored class of businesses, as New York has done in this case, the State must justify why houses of worship are excluded from that favored class” (3).

Nevermind that you do different things in all of those facilities; as Justice Sotomayor acidly notes, “bike repair shops and liquor stores generally do not feature customers gathering inside to sing and speak together for an hour or more at a time” (3). She adds:

“Justices of this Court play a deadly game in second guessing the expert judgment of health officials about the environments in which a contagious virus, now infecting a million Ameri-cans each week, spreads most easily” (3).

So the conservatives are simply wrong on the facts, even though those facts are articulated quite clearly in the decision they overturn.  And that decision will probably cost people their lives.  Freedom!

But: bars and restaurants? Perhaps I am particularly sensitive to this point because schools are closed too (as one meme wondered – should NYC parents send their kids to the bars that are still open, now that schools are closed?), but it’s impossible not to notice that moneymaking endeavors get a different set of rules from civic ones. As Mull puts it:

“With people out of work and small businesses set up to fail en masse, America has landed on its current contradiction: Tell people it’s safe to return to bars and restaurants and spend money inside while following some often useless restrictions, but also tell them it’s unsafe to gather in their home, where nothing is for sale. It’s a woefully inadequate stimulus plan, funded by money extracted little by little from the pockets of people who are mostly just confused about what they’re being compelled to do.”

She cites the example of Rhode Island, whose restrictions are completely absurd:

“Residents are prohibited from gathering with even one person outside their household, even in the open air of a public park. But inside a restaurant? Well, 25 people is fine. Hire a caterer? You’re legally cleared to have up to 75 outdoors. The governor’s executive order merely notes: “The lower attendance at such events, the lower the risk.” (The Rhode Island governor’s office did not respond to a request for comment.).”

And how did we get here? Well, as Mull underlines, closing bars and restaurants (etc) will cause a lot of unemployment. Closing schools and churches won’t.

In other words, it’s not that the plaintiffs in the New York case necessarily raised a compelling First Amendment argument, but they are a canary in the coal mine for a deeply problematic public health policy, one that systematically favors commercial activities over civic ones.  Governors’ hands are tied, even the ones who want to do the right thing, because the proper public health measures will cost money in the form of income guarantees to the people affected by them. And guess who has – perhaps singlehandedly – stood most in the way of that financial assistance? That’s right – Moscow Mitch and his repeated bad-faith refusals to negotiate a second stimulus package. Turns out he was so busy ramming right-wing court justices through the Senate that he made it more dangerous to do stuff that is more important, like school, worship and family Thanksgiving.