Apple developer statistics, or the lack thereof, and their implications for user privacy
Listening to the most recent episode of ATP, I learned a surprising fact about how Apple developers see their user statistics: the number of people who opt out of sharing is not available, and the only statistics developers get is about the users who opted in. This may make naïve sense — hey, they don’t want you to know about them so we’ll erase them from existence Soviet-style — but is in fact statistical malpractice. The numbers Apple does share are only a sample of the total user base, which is fine as long as you know or can estimate the size of the population you are sampling. Without a way to estimate the denominator, the numbers are meaningless.
The bare minimum that Apple could and should do, without breaking any privacy rules, is to share the number of people who opted out. With this number in hand, a developer would be able to project worst-case scenarios and likely ranges for each statistic of interest. If Apple dedicated just a tiny bit more resources to developer relations, they could automate this step and build in worst-case numbers into the interface. A more sophisticated company that cares about the developer ecosystem could even create complex predictions models that would give both worst-case numbers and the 95% confidence intervals for each statistic in question.
Why should Apple care, other than doing a solid to their developer ecosystem with a minimal investment? Well, by making their default user statistics useless they are in fact incentivizing developers large and small to track their users by other means. Large developers may do that anyway since they are likely to have a could component to their app. But an indie developer may turn to third party services: if I am a teenager working on my first big app I will probably turn to the fastest and cheapest way to track, and not knowing anything about anything I am going to guess some of the offerings in that part of the price-value spectrum are going to be less than scrupulous.
So this is how Apple’s privacy-minded view combined with thriftiness and lack of care towards developer experience can lead directly to worse overall privacy for their users. What a surprise.
The enigma of Tyler Cowen
The most recent episode of Conversations with Tyler featured Ezra Klein as a guest. I won’t delve into their discussion, but I will note two quirks in Tyler’s thinking that were somewhat off-brand.
When Ezra mentioned mass firings and re-hirings in the federal government, Tyler expressed uncertainty about that actually being the case — essentially a denial of what is widely known. As a counter, Ezra mentioned knowing people who were fired. I, too, know people who were fired simply because they got promoted and transitioned from contractors to full-time employees within the last two years, despite having deep expertise in the subject matter they were hired for. The “I’m not sure that’s happening” sounds what one would hear on the streets of Berlin in the late 1930s, motivated reasoning par excellence.
Another example is Tyler’s tendency to extrapolate into the future and then use that hypothetical future as a benchmark for current policy. Think “Sam Altman says it will be possible to have billion-dollar companies run by one person” at some unspecific time in the future as a context for massive federal government downsizing now. We do not, in fact, have billion-dollar companies run by a single person; that is a VC pitch at best, and if we are being less generous pure vaporware. Government hiring policy by vaporware sounds bad.
I think Tyler was aiming for provocation. He didn’t mention in his show notes that he “tried to “push him further from a libertarian point of view”, but then what is he really thinking? To come back to the 1930s analogy, I picture this interview’s version of Tyler in the FDR-era US and I see him praising Charles Lindbergh (they both speak German) and Henry Ford — progress über alles — and not really being a fan of FDR. Yikes!
Much has been written and said about the faults of peer review but one thing I think hasn’t been emphasized enough so I’ll state it here: journal editors need to grow a spine. And they need to grow it in two ways, first by not sending obviously flawed studies out for peer review no matter where they come from, then by saying no to reviewers' unreasonable demands, not taking their comments at face value, and sometimes just not waiting 6+ months for a review to come back before making a decision.
Voices in my head, 2024
Podcast-wise, I am in my 2016 mood. There were five prospects for 2024. I became a regular listener of exactly zero. The true regulars continue to be ATP, EconTalk, Conversations with Tyler, The Talk Show and Dithering, but even there I skip through more episodes than I complete.
Still, moods shift and if I listen to more of anything next year it may be one or more of these:
- The VPZD Show, which is back, intermittently, with more cursing than ever, but also some interesting viewpoints that differ enough from my own to make things interesting.
- The Marginal Revolution Podcast with Tyler Cowen and Alex Tabarrok. Kyla Scanlon would have been a better co-host to either, but this will have to do.
- Plain English with Derek Thompson is very much outside of my preferred format, being over-produced and a bit too highfalutin' with its choice of guests, but some of those guests do seem interesting so it may become one of those skip-more-than-I-listen-to shows.
And here are years past: 2023 — 2022 — 2021 — 2020 — 2019 — 2018 — The one where I took a break from podcasts — The very first one
I have been reading Order without Design for the last few weeks — it’s a dense book! — so it is complete coincidence that its author Alain Bertaud was on EconTalk as a guest the same week that I finished it. Bertaud talked to to Russ Roberts once before, back in June 2019. I haven’t listened to either episode yet but they’re on the list.
I have linked to a Conversation with Tyler (Cowen) in a while because most of them have to date been bland (Nate Silver? Seriously?) but the most recent one with Kyla Scanlon was compelling. She is a 27-year-old book author and… popularizer of economics (?) who writes on Substack and makes videos.
Last week’s EconTalk with Marty Makary featured several topics relevant to zombie medicine. One was a zombie’s return to the world of the living, with hormone replacement therapy for women not being as bad as we thought, particularly for preventing hot flashes in early menopause (before age 60). The other was the emergence of a new zombie: removing ovaries to prevent ovarian cancer when it is now thought that most ovarian cancers arise from the Fallopian tubes, not the ovaries themselves. I wouldn’t call it a full blown zombie just yet as there is an ongoing randomized controlled trial comparing the two approaches and who knows, its results may kill the old practice outright.
📚 Finished reading: The Friction Project by Robert I. Sutton, one of those books that should have been a blog post. The fluff is a rapid-fire succession of case studies that are too brief to be memorable but too detailed to be quickly filed away. Just listen to the podcast episode and skip the book.
Today in teaching birds how to sing
From the Institute For Progress-supported newsletter, Macroscience:
Last year, IFP brought together some of our closest friends and collaborators to put together a podcast series that would serve as a beginner-friendly introduction to metascience.
The result? “Metascience 101” – a nine-episode set of interviews that doubles as a crash course in the debates, issues, and ideas driving the modern metascience movement. We investigate why building a genuine “science of science” matters, and how research in metascience is translating into real-world policy changes.
So far so good. First guests?
Journalist Dylan Matthews sits down with economist Heidi Williams and IFP co-founder Caleb Watney to set the scene.
Bah-rump. Episode two?
OpenPhil CEO Alexander Berger interviews economist Matt Clancy and Stripe co-founder Patrick Collison to talk about whether science itself is slowing down, one of the key motivating concerns in metascience.
A journalist, an entrepreneur, two economists and a policy wonk gather around the fireplace to talk science. What seems to be missing is actual scientists. Stop me if you’ve heard this one before.
And if your retort is that few if any scientists have metascience as there full-time field of study, well, are any of the above doing it full time? I am sure the discussions will be brilliant — I will write up updates as I listen along — but the start looks a helluva lot like an echo chamber. Hope I’m wrong!
(↬Tyler Cowen, because who else. He will also be a guest in a future episode.)
ChatGPT, the blog expert
The latest episode of The Talk Show was with Taegan Goddard, who all the way back in 1999 founded the blog Political Wire which is apparently a continuous intravenous drip for people interested in US politics. Now, I’ve had other preocupations back then and not being an American citizen still have little to no interest, so this blog wasn’t even on my radar until listening to the episode. But now I wonder: are there any more relevant blogs I’ve missed out on, about medicine and biotechnology in particular?
ChatGPT’s first pass was mediocre. I’ll save you the verbalist padding, but here are its suggestions in response to my prompt: “Is there a website/blog like politicalwire.com or daringfireball.net but for biotechnology?”
- Endpoints News
- Fierce Biotech
- STAT News
- In the Pipeline
- Xconomy
It’s a 20% hit rate: only Derek Lowe’s In the Pipeline comes close to what I asked for. The others are all medium to big news outlets that yes, focus on biotech, but that’s not what I asked for. The second try, after I asked for more like Lowe’s, was a tad better:
- Timmerman Report
- The Niche
- The Biotech Strategy Blog
- Leonid Schneider’s For Better Science
- Science Translational Medicine Blog
That’s more like it! 80% now, and if I were feeling generous I’d give it a full 100% since In the Pipeline is, in fact, a Sci Trans Med blog. But then I asked for too much, and it hallucinated 3 more, two of which were hallucinations (BioPunk and BiotechBits, which were at least plausible names) and one was a sub-blog of Endpoints that also didn’t exist.
So, now I have two new blogs to follow (Timmerman Report and The Niche; Biotech Strategy is behind a paywall and I’ve already been following the others), and an ever-increasing urge to update the Blogroll, which has been under construction for the past five months with no end in sight.