The chaos of complexity
How many more decisions do we need to make?
Art: Portrait of Marguerite Sleeping, Henri Matisse, 1920
Since being promoted to CEO in March, I’ve made a lot of decisions with partially-complete information and uncertain outcomes. For example, in order to figure out whether my daughter should start kindergarten in-person, I had to read a (very good) newsletter of crowdsourced information from parents across the country, anecdotal reports in the Working Moms Subreddit, participate in a Whatsapp chat with two other moms from daycare, and ask my husband what his coworkers in the area were doing. This was still enough data to make a decision. Ultimately, I had to go on gut feel.
This process took forever, was very emotionally laden, and competed for attention in my brain with all the other chaos COVID has caused. As a result, when it came time to make all the other millions of decisions: should we buy disposable masks or cloth masks? Are there things I need to be stocking up on in the fall to prepare for winter surges? Should I still be wiping down groceries, all of my decision-making heuristics were overwhelmed and I operated on coffee and fear. I am, as Ed Yong writes, imprisoned in wall by my own instincts.
The biggest decision like that I’ve recently had to make was, how do I get tested for COVID?
One of the very few people I saw in-person recently tested positive, and, although I didn’t have any symptoms, the clock was now ticking for me to get tested and either isolate before I passed it onto anyone else or know for sure that I didn’t have it.
If you’ve had the unique pleasure of going in for a COVID test in the United States recently, you know that the system is still a complete mess. Case in point, this thread:
First of all, there are no standard places to get tests. It varies very much on a state-by-state basis, and even within a state, by county. Upon a search, it looked like there were a couple options for our county; either a pharmacy like CVS, an Urgent Care Center, or a county-wide testing center. In our county, for example, there used to also be drive-up centers, but they closed.
Each of these locations has different hours, and different criteria. Some make you get a prescription from a doctor. Some are by appointment. Some are walk-in. Some have the PCR test, some have something known as the 15-minute test. Some test children, some do not. Some are paid, some are reimbursable by insurance, some are not. I had to figure out which one of these testing centers to go to, which test to get, and how to do all of this quickly enough to get results and not to pass it on to anyone around us.
I initially tried to schedule an appointment through CVS’s website, but it said that no testing times were available anytime over the next three days because they had a very low volume of tests. A phone call to the store was not picked up. I tried Walgreens and got the same result. I then looked up the testing centers at our county, but it looked like some were closed, and the information had not been updated for weeks. The other suggested testing centers on the county website were at CVS.
I decided to go with the urgent care clinic, which offered two tests, the PCR and a 15-minute test. I didn’t know which results would be faster, so I scheduled both. I could take the PCR test the next day. The 15-minute test was only available 2 days from now. I tried calling to get additional information, but no one picked up. At the urgent care clinic, when I came in for the PCR, I asked again which test I should get. I was told at the front desk that, since I didn’t have symptoms, I should get the PCR test. When I asked yet again, this time from the technician who did my swab, she said that I could get the 15-minute test, and it would be totally fine. I did the 15-minute test the very next day. I tested negative.
It took 8 days to receive my PCR result, also negative. The clinician told me that PCR results were currently taking anywhere from 3-10 days. They couldn’t give me an average time, and they said sometimes it took as many as 15 days because basically two labs, Quest and LabCorp, were processing the tests. If there is a flare-up in any part of the country, the labs focus their efforts there, and tests from other areas are deprioritized. It’s all part of a very vague, opaque, slow-moving black box, and we are, again, not in the room where it happens.
What are we, as a society, supposed to do when we have to constantly make decisions like this? I feel like I read a ton on coronavirus on a daily basis, but, when go time arrived, I was still fairly lost on what exactly to do and where to go.
The fact that we still, six months into this thing, do not have a standardized quick-testing infrastructure is extremely worrying, and infuriating. Atul Gawande writes,
To get out of this pandemic, we need fast, easy coronavirus testing that’s accessible to everyone. From the way people often talk, you might think we need a technological breakthrough to achieve this. In fact, we don’t have a technological problem; we’ve got an implementation problem. We could have the testing capacity we need within weeks. The reason we don’t is not simply that our national leadership is unfit but also that our health-care system is dysfunctional.
…
Appointments can take days, results days more. Most testing in the United States is done by four companies—Quest Diagnostics, LabCorp, BioReference Laboratories, and Sonic Healthcare. Through early August, results routinely took four days or more, making the tests essentially useless. Times improved only when testing volumes declined, because many people gave up on getting tested. The vast majority of infected Americans, including those with symptoms, never get tested. And we have not even reached the fall, when flu season will hit and coronavirus-testing needs and demand are expected to rise substantially. As the saying goes, it’s as messed up as a pile of coat hangers.
By the way, the 15-minute test costs $75 per person, payable by credit card immediately upon test. I am fortunate enough that the cost didn’t factor into my decision.
After all of that, I don’t have any room in my brain to make any additional kinds of decisions. I am decisioned out. I’m tired of thinking about this system and why it’s failed us in this way, tired of trying to think through all the steps that have failed. I started to read this book, on our healthcare system, An American Sickness, and had to put it down.
This is why, recently, when there is tech in my life that makes decisions for me, I am very grateful. (We’re finally getting to the part of the tech newsletter post where I talk about tech)
I usually am pretty cautious about algorithms, but 2020 has got me so down that I’m reconsidering.
The first thing that I love that makes decisions for me is the Covid Tracking Visual Dashboard. I generally very much trust the data that the team puts together, but before they only had the API and the daily number reporting. (Which was still a TON of work. I can’t imagine how hard it is to collect and clean all that data.) But the charts are so good.
All I do is look at my pictures every day, and I get a sense for where we are as a country and whether my life today feels riskier than it did yesterday . And since I hit the site on a daily basis, I have a good sense for its rhythms, its inconsistencies, and where I think it might be wrong. For example, there is definitely not a dip as it shows, because numbers were extremely low over Labor Day, probably because A) People weren’t getting enough tests, and B) reporting took time off, too. Deaths, the gold standard metric, are pretty much lower but steady,which means we’re still hovering at around 40k new cases a day.
I am familiar with this chart now, I love it, I use it. It makes, at least this part of my life, easy for me.
The second thing is the new portrait mode on my iPhone. I recently got an iPhone SE. (This decision was another choice in itself. What’s the difference between SE and XR and 11 Pro? Why is it so much less expensive than all the other phones? What is Apple trying to pull over on me?) and it has the new Portrait Mode, which takes beautiful pictures of my kids without me having to know photography even a little bit.
Because a smartphone’s sensor is so small and the field of view is so wide, most of the time everything in a normal photo taken with the phone’s camera will be in focus. Portrait mode simulates a shallow depth of field by using edge detection and/or depth mapping to differentiate between the foreground and background. It then blurs the background, simulating that shallow depth of field and making the foreground pop.
Google published a paper detailing how its portrait mode worked (although for a single-camera phone, unlike Apple’s original iteration), if you’re curious, and it’s really interesting. But the great thing is that, if you’re not curious, you don’t need to understand neural nets, dual-pixel autofocus hardware, people segmentation, training data from Flickr, or any of that. It doesn’t matter because it all works. All you have to do is point your phone at your baby.
When we get to the point where we can say, go here and take a 15-minute COVID test that’s 99% accurate, for free, we will have solved this thing, and life can go back to normal, and we can go back to worrying about stupid stuff like Facebook launching Facebook on Facebook.
Until then, we live in the chaos of complexity.
What I’m reading lately: Just a bunch of Twitter threads.
I asked, “What underrated tech is the tech news not covering?” The replies are great.
How to make virtual watercoolers work
Good secret tools in the ML toolchain
Rachel by the Bay on return values
The Newsletter:
This newsletter’s M.O. is takes on tech news that are rooted in humanism, nuance, context, rationality, and a little fun. It goes out once or twice a week. If you like it, forward it to friends and tell them to subscribe!
The Author:
I’m a machine learning engineer. Most of my free time is spent wrangling a kindergartner and a baby, reading, and writing bad tweets. Find out more here or follow me on Twitter.
If people are giving up on get testing because "it's too hard to do" and, the symptoms and illness are generally mild for 80%, of the population, are the COVID data numbers totally skewed as to be unreliable? Many just go through the COVID sickness episode uncounted in the stats. Unless they require hospitalization, they are not recorded.