Al Gidari is a smart, good faith interlocutor. Otherwise I would respond to this rather amazing tweet... differently. But I think some response is indicated. https://twitter.com/agidari/status/1263837397776650246

I’m starting at the end so you see just where the line of thinking takes you: people are flawed so we shouldn’t trust them over technology. It’s a variation on a theme you see in many contexts, from sentencing (judges are flawed) to disability benefits (nurses are flawed).

I’m going to focus on the poor assumptions embedded in this trope and skip over other of Giadari’s claims such as traditional contact tracing is hard (okay) and has never been used at scale in the U.S. (verifiably incorrect). If anyone wants to tackle those, Godspeed.

The first thing to understand is that automated contact tracing *is* contact tracing by people called “developers.” They too possess a certain amount of training in epidemiology (nil) and a particular worldview (which precipitated the techlash, an entire societal phenomenon).

Maybe you’re thinking, but Ryan, they can talk to experts! If this is your impulse, I encourage you to read @AINowInstitute’s Litigating Algorithms series detailing the massive failures of software developed alongside state authorities.

Specific to contact tracing, I encourage you to look at the experiences of the UK, or North Dakota, or the TraceTogether lead. Yet despite overwhelming evidence and, you know, the existence of @daniellecitron, hope in technology springs eternal.

Tech does offer new affordances: to Giadari, only apps can identify unknown exposures. Now you could figure out who was on a bus or in a restaurant at a particular time without an app, so the idea must be random contacts like the park bench illustrated by Google in its protocol.

There is significant and mounting evidence that apps will be over and underinclusive. It’s also pretty intuitive. You’d have to set a location and duration threshold that is robust across an arbitrary number of contexts (parks, cars, office buildings, pogo sticks, corn mazes).

Ok, so what’s the harm? Why not have both? Well, because states won’t invest enough in widely accepted, if expensive approaches to containing the virus if they think there’s an app for that. And that’s setting aside all the other ways the apps could backfire.

This conversation isn’t new. It’s a conversation we’ve been having for decades. And yet, every time tech doesn’t deliver, every time IBM or Google quietly shutter some failure, every big @JuliaAngwin or @kashhill story, the Matrix resets and we get to debate once again.