6 ms·

Link without all those ads. https://mobile.twitter.com/doctorow/status/1296248167319482375 https://mobile.twitter.com/doctorow/status/12962481673194823... Don

Link without all those ads. https://mobile.twitter.com/doctorow/status/1296248167319482375 https://mobile.twitter.com/doctorow/status/12962481673194823... Don't really follow his logic there. Contact tracing is a manual process, thus digital contact tracing apps bad because of that one Michigan app that's actively stupid and a few other failed attempts? That whole argument seems to make major leaps and ignores more sensible approaches to the degree that I'm unable to contact trace whatever point he was originally trying to make. Sure, some people took the "just throw tech at it without thinking" approach but that happens everywhere these days.

dotcoma 6y ago
To me, the important point is #6 6/ The thing is, contact tracing is high-touch/low-tech because it is a social science intervention. Social scientists have always understood that if you only gather the data that's easy to reach, you'll come to bad conclusions skewed by defects in your collection.
throwanem 6y ago
The point he's making is that "just throw tech at it without thinking" is this industry's only approach, and it's an approach that, in a situation like this, not only can't help but can't even help but be actively harmful. I have some relevant experience here. Back almost a decade ago, while I was working as a staff engineer for a local medical school's genomics institute, I was briefly involved with some tooling work for a project that aimed to characterize the spread and prevalence of various STDs among Baltimore's at-risk youth. Contact tracing was of course a major part of that, but where today's (or 2012's!) tech industry would just shrug and stick New Project.xcodeproj in front of an unsecured Mongo store, the doctors and public health professionals running this project built free clinics. Yes, there was an app involved - a tablet app for doctors to use, in documenting what their patients told them about whom they'd been with, when, and so forth. But, unlike some, they knew better than to assume that any app could by itself do the job for them, rather than they themselves needing to go out and serve the community they wanted the chance to learn more about. You know. Contact tracing. As opposed to, slapping together a pile of React Native and throwing it on an app store in hopes of grabbing some VC bucks and making a get-rich-quick exit - or even Apple and Google getting together and actually engineering something that's technically very sound and still totally useless because it can only ever be purely quantitative in a single dimension, physical proximity, and - as I learned when I was asked to help out with building data models for it - contact tracing is not and never can be a one-dimensional problem. That's the point Doctorow is making here. And for all that I'm not a big fan of the guy, in this case he's nailed it dead square in the bullseye.
- roenxi 6y ago
  > The point he's making is that "just throw tech at it without thinking" is this industry's only approach... More subtly, his point is that tech is choosing to do what is easy, not what is necessary. Tech has had some great successes doing this in the past (see also Worse is Better). It isn't a very good approach though when the relationships are mandatory and the data sensitive.
  - throwanem 6y ago
    Sure. Looking at it in the most generous fashion possible, it's an example of what happens when the low-hanging fruit lasts so long that people forget what ladders are for. Looking at it with the experience of two decades spent in and around this industry, it's the latest in a long and depressing series of reasons why unicorn factory farming needs to die.
- usrusr 6y ago
  > The point he's making is that "just throw tech at it without thinking" is this industry's only approach Maybe that's the point he intended to make, what he actually wow is really just a collection of anecdotes about troubles with location based tracking. That's like realizing that it hurts if you hit yourself with a hammer and thus vowing to never again use a screw. The closest thing be has to an argument against the decentralized proximity approach is a hand-wavy mention of the idea that there might be a possibility of false positives in a traffic jam. The thing is that the conflict between manual tracing vs electronic exposure notification is a complete fantasy because the resources required to do one have zero overlap with the resources to do the other. Sure, automated exposure notification is wildly imprecise. But so is viral spread, it's a game of probabilities with or without the uncertainties of Bluetooth beacons. But beacon tracing scales where manual tracing doesn't. In a high incidence situation manual tracing will quickly run out of call center agents and in a low incidence situation social distancing and manual location logging (e.g. documentation of restaurant table allocation) will quickly erode to the point where a very, very large fraction of contacts cannot be manually reconstructed. And even in the moderate incidence band where manual tracing works best it has blind spots where beacon logging can fill in, not only anonymous exposures like public transit but also the entire field of "awkward" contacts that Doctorow spent an entire paragraph on. "And then there's this survival sex worker I occasionally meet, here's his number" surely isn't a line that comes up often in a manual tracing conversation. But yeah, it must be fun to drum up negativity about exposure notification apps, it's a self-fulfilling prophecy when nobody is using them.
SideburnsOfDoom 6y ago
> thus digital contact tracing apps bad because of that one Michigan app that's actively stupid and a few other failed attempts? Well, what successful attempts should we balance these "few" data points against? What app can compensate for the difference between a) I have been a few feet from this person for hours because they slept in the same bed as me and b) I have been a few feet from this person for hours because they live next door, behind a brick wall.
- blisterpeanuts 6y ago
  A campus map including blueprints solves "b". It's pretty easy to tell whether two Bluetooth broadcasters are in the same room or have a wall between them. Beacon-based asset tracking systems do this all the time.
  - SideburnsOfDoom 6y ago
    Campus? I live in a city not a campus.
- usrusr 6y ago
  It's not an issue at all. Including the brick false positive (or drywall) you still get a much better test allocation than blindly guessing. Same for distancing allocation: occasional false positive seclusion would have much better cost/benefit than prolonged blanket closure. You don't need to be able to perfectly predict the roulette wheel, you'll already find yourself in an interesting conversation with security if you consistently beat the odds between black and red.

dotcoma 6y ago

To me, the important point is #6 6/ The thing is, contact tracing is high-touch/low-tech because it is a social science intervention. Social scientists have always understood that if you only gather the data that's easy to reach, you'll come to bad conclusions skewed by defects in your collection.

throwanem 6y ago

The point he's making is that "just throw tech at it without thinking" is this industry's only approach, and it's an approach that, in a situation like this, not only can't help but can't even help but be actively harmful. I have some relevant experience here. Back almost a decade ago, while I was working as a staff engineer for a local medical school's genomics institute, I was briefly involved with some tooling work for a project that aimed to characterize the spread and prevalence of various STDs among Baltimore's at-risk youth. Contact tracing was of course a major part of that, but where today's (or 2012's!) tech industry would just shrug and stick New Project.xcodeproj in front of an unsecured Mongo store, the doctors and public health professionals running this project built free clinics. Yes, there was an app involved - a tablet app for doctors to use, in documenting what their patients told them about whom they'd been with, when, and so forth. But, unlike some, they knew better than to assume that any app could by itself do the job for them, rather than they themselves needing to go out and serve the community they wanted the chance to learn more about. You know. Contact tracing. As opposed to, slapping together a pile of React Native and throwing it on an app store in hopes of grabbing some VC bucks and making a get-rich-quick exit - or even Apple and Google getting together and actually engineering something that's technically very sound and still totally useless because it can only ever be purely quantitative in a single dimension, physical proximity, and - as I learned when I was asked to help out with building data models for it - contact tracing is not and never can be a one-dimensional problem. That's the point Doctorow is making here. And for all that I'm not a big fan of the guy, in this case he's nailed it dead square in the bullseye.

roenxi 6y ago
> The point he's making is that "just throw tech at it without thinking" is this industry's only approach... More subtly, his point is that tech is choosing to do what is easy, not what is necessary. Tech has had some great successes doing this in the past (see also Worse is Better). It isn't a very good approach though when the relationships are mandatory and the data sensitive.
- throwanem 6y ago
  Sure. Looking at it in the most generous fashion possible, it's an example of what happens when the low-hanging fruit lasts so long that people forget what ladders are for. Looking at it with the experience of two decades spent in and around this industry, it's the latest in a long and depressing series of reasons why unicorn factory farming needs to die.
usrusr 6y ago
> The point he's making is that "just throw tech at it without thinking" is this industry's only approach Maybe that's the point he intended to make, what he actually wow is really just a collection of anecdotes about troubles with location based tracking. That's like realizing that it hurts if you hit yourself with a hammer and thus vowing to never again use a screw. The closest thing be has to an argument against the decentralized proximity approach is a hand-wavy mention of the idea that there might be a possibility of false positives in a traffic jam. The thing is that the conflict between manual tracing vs electronic exposure notification is a complete fantasy because the resources required to do one have zero overlap with the resources to do the other. Sure, automated exposure notification is wildly imprecise. But so is viral spread, it's a game of probabilities with or without the uncertainties of Bluetooth beacons. But beacon tracing scales where manual tracing doesn't. In a high incidence situation manual tracing will quickly run out of call center agents and in a low incidence situation social distancing and manual location logging (e.g. documentation of restaurant table allocation) will quickly erode to the point where a very, very large fraction of contacts cannot be manually reconstructed. And even in the moderate incidence band where manual tracing works best it has blind spots where beacon logging can fill in, not only anonymous exposures like public transit but also the entire field of "awkward" contacts that Doctorow spent an entire paragraph on. "And then there's this survival sex worker I occasionally meet, here's his number" surely isn't a line that comes up often in a manual tracing conversation. But yeah, it must be fun to drum up negativity about exposure notification apps, it's a self-fulfilling prophecy when nobody is using them.

SideburnsOfDoom 6y ago

> thus digital contact tracing apps bad because of that one Michigan app that's actively stupid and a few other failed attempts? Well, what successful attempts should we balance these "few" data points against? What app can compensate for the difference between a) I have been a few feet from this person for hours because they slept in the same bed as me and b) I have been a few feet from this person for hours because they live next door, behind a brick wall.

blisterpeanuts 6y ago
A campus map including blueprints solves "b". It's pretty easy to tell whether two Bluetooth broadcasters are in the same room or have a wall between them. Beacon-based asset tracking systems do this all the time.
- SideburnsOfDoom 6y ago
  Campus? I live in a city not a campus.
usrusr 6y ago
It's not an issue at all. Including the brick false positive (or drywall) you still get a much better test allocation than blindly guessing. Same for distancing allocation: occasional false positive seclusion would have much better cost/benefit than prolonged blanket closure. You don't need to be able to perfectly predict the roulette wheel, you'll already find yourself in an interesting conversation with security if you consistently beat the odds between black and red.