Thread by @jben0

I see many folks referring to this Brookings article as justification for opposing contact tracing apps

The article does not provide a good justification for that position

Below I will go through the article paragraph-by-paragraph with my comments 1/n

brookings.edu/techstream/ina…

Note: I am not disinterested.

My colleague David Byrne(@dprbyrne) and I are working on how to help increase app adoption (and use) within Australia (and Canada)

We both believe that the app has promise, and it is worth giving a shot (or two!) 2/n

If you want a reason for seeing promise, see this paper: science.sciencemag.org/content/early/…

Note: the app is not a panacea, it is a compliment to traditional tools. This point was been made by Jason Bay (@jasonbay), who was involved in Singapore’s development: blog.gds-gov.tech/automated-cont… 3/n

back to the Brookings piece..

If your take-away message is that the app may help but:
- it won't replace traditional tools
- false positives need to be minimized
- the app won’t catch all infections
- we need be vigilant about privacy protections
then good, you can stop here 4/n

That was not my take-away message

It does not appear to be the take-away message of many folks, judging by how they cite this article on Twitter and in the news 5/n

Here it goes

(if you want to skip ahead to the critical parts, please begin at paragraph P6 below) 6/n

P1 This paragraph characterizes the push for the tracing app as panic-driven. Perhaps this is fair in the US, not so much in Australia and other countries, see here: theguardian.com/australia-news…
7/n

P2 This paragraph sets the tone. By writing “fragile”, “pinning their hopes” and “Coronavirus? There’s an app for that”, the reader is being given the sense that this app unlikely to work. Fine if that is the conclusion of an analysis, I didn’t see that case made 8/n

P3 This paragraph establishes the article as an across-discipline authoritative take. It mentions the risks of over-hyping the app and using it as justification for opening up too soon. These are important points. I agree completely for the US context 9/n

P4 This paragraph urges that the limitations of the app be acknowledged. This is important and I agree with it completely 10/n

P5 This paragraph describes traditional contact tracing
11/n

P6 This paragraph says there is little promise in the apps. I see three problems... 12/n

1. it gives the reader a false impression that contact tracing apps are intended as a substitute, rather than a compliment to traditional measures. Serious advocates say otherwise, e.g. researchers: science.sciencemag.org/content/early/… developers: blog.gds-gov.tech/automated-cont… 13/n

2. The claim that *automating* contract tracing hasn’t been demonstrated reliably “despite numerous concurrent attempts” is not an accurate description.

Again, the serious people advocating for this app do not see it as a substitute for traditional methods... 14/n

…Further, the failure is primarily with adoption. For example, Singapore has had issues with adoption, but has not abandoned its use of the app. Outside of their marginalized communities, they appear to be doing a good job tracking (too early to know how much app helped) 15/n

3. The description that the impact is at best marginal does not gel with this paper: science.sciencemag.org/content/early/…. I see no countervailing analysis/reference. The promise may not pan out, but one can push back against hype without throwing cold water on the entire project 16/n

P7 This paragraph describes the Apple/Google API and that it was proposed because of pressure from the T administration. Is this troubling association a fact? A T claim? Germany and many other countries have adopted the Apple/Google approach: reuters.com/article/us-hea… 17/n

P8 This paragraph discusses false positives for barriers (walls/masks). But masked folks can opt-out if comfortable. Walls won’t be registered unless contact is of extended duration. If so, likely office co-worker or neighbor, so won’t be frequent & you probably want to know 18/n

… note the fact that apps don’t “account for when individuals take precautions” is not unique to the app.

This gives a false impression that this is somehow a limitation relative to traditional measures 19/n

P9 This paragraph discusses false positives due to fleeting interactions. I see three problems:

1. This is overstated due to a basic error in understanding basic facts regarding how the app uses Bluetooth LE, as I mentioned in this tweet:

https://twitter.com/jben0/status/1257517195648577536

20/n

2. It is odd to claim that if the app has a false negative issue for fleeting interactions, then this is a limitation. This is shared with traditional tracing, so not sure why this is mentioned as a limitation

21/n

3.What is that evidence for your prediction that cooperation will decay because of many false positives? This will depend on the base rate of positives, the specificity (understated), how communication is handled (noted by authors). I see no analysis 22/n

P10 This paragraph discusses the problem of smartphone penetration & smart phone use. Why throw up roadblocks and not think of creative solutions? Buy phones for people? Dedicated blue tooth apps for seniors? RFID alternatives? Requirements for entry into stores, etc.? 23/n

Fear of reporting can potentially be addressed.

The undetected positives not reporting is an issue without the app

You aren't arguing against the best version of how app tracing can be implemented. Why not steelman it? 24/n

P11 This paragraph brings up the problem of motivating use because the app doesn’t benefit the individual. First, lack of benefit isn’t necessarily true if antivirals are shown to have more benefit the earlier disease is detected... 25/n

Second, yes, motivating use will have to be thought through. Why not do that rather just flag it as an unsolved problem? Maybe people can get a free test to allay their concerns? Maybe there can be other benefits attached? Public health authorities can be involved here 26/n

P12 This paragraph discusses potential privacy issues. It overstates the concerns. Is this example the most severe one? The correlation exploit would not work in a crowded area. What is the motive to set up a camera for the tiny chance of finding a random COVID-positive? 27/n

P13 This paragraph on the possibility of malicious use reflects a false understanding of the kind of behavior that the app cannot prevent. The first white paper by BlueTrace discusses how to prevent this (p. 6): bluetrace.io/static/bluetra… 28/n

P14 This paragraph makes the point that the limitations of apps need to be acknowledged, and that they should never be used in isolation. Important points for Apple/Google messaging. My previous links indicate that many advocates and govt policy folks get this 29/n

P15 This paragraph makes the point that there is a danger that apps will become effectively compulsory. It is a value-judgement if we want to regard this *temporary* state-of-affairs as a “danger.” There are trade-offs. Some loss of privacy may increase freedom in aggregate 30/n

P16 This paragraph reiterates the earlier false positive & griefing concerns. It suggests that these concerns will lead people to be social pariahs. As noted, griefing can be prevented, OTOH... 31/n

I share the false positive concern (note: it is not as much as claimed). Can we analyze the extent of the problem? Can we think of a solution? Perhaps there is a way to detect neighbors? 32/n

P17 This paragraph accepts the reality that these tracing apps are coming. It suggests good ideas for assuring privacy and security.
Given this reality, shouldn’t we be trying to improve the apps, and encourage adoption, rather than find fault, and discourage? 33/n

P18 This paragraph discusses important concerns if apps are not made temporary. I agree completely. 34/n

P19 This paragraph discusses other privacy issues, including the importance of protecting vulnerable communities. I agree this is important, and it should be monitored as this is rolled out 35/n

P20 This paragraph makes the important point that the app must do no harm. I agree, but I fear that this article may have done harm to the cause of motivating app adoption and use. If the app helpful for public health, that would not be good 36/n

I will end with this:

"If you need to be right before you move, you will never win."

https://twitter.com/AbraarKaran/status/1257526025245855747?s=20

37/n

@CT_Bergstrom @rcalo @ashk4n @enscand @AlecStapp @joejerome @EricTopol @doctorow @monteskw
@gregggonsalves @AntonioCasilli