The reform that would really break Facebook would be strict limits on behavioral data collection, and a ban on using customer data in third-party ad targeting. This would bring advertising back to the pre-2000 status quo and demonetize some of the worst behaviors on social media.https://twitter.com/justinamash/status/1445462709588291586 …
-
Show this thread
-
Replying to @Pinboard @bergmayer
The noyb cases may effectively lead to the second outcome in the EU if they prevail, allowing Facebook users to refuse to share their data with advertisers as a condition of usehttps://noyb.eu/en/noybeu-filed-complaints-over-forced-consent-against-google-instagram-whatsapp-and-facebook …
2 replies 3 retweets 6 likes -
Replying to @henryfarrell @bergmayer
How would that lead to the second outcome? Having no data on a particular user within a surveillance architecture is very different from having a framework where you can only target advertising to content.
3 replies 0 retweets 1 like -
Replying to @Pinboard @bergmayer
Sorry for delay. The logic is that if the EU rules against, Facebook won’t be able to refuse service to people who decline to share data with advertisers. That in turn may make their current business model unworkable in the E.U.
1 reply 0 retweets 0 likes -
Replying to @henryfarrell @bergmayer
My point is that Facebook can use data from the majority of people who don't opt out to make fairly accurate inferences about the remainder who do. The model survives intact unless you can get a large majority opting out, which Facebook will find creative ways to make difficult
3 replies 0 retweets 1 like -
Replying to @Pinboard @bergmayer
This is likely my stupidity, but are they going to be able to make good inferences about person x if they are not allowed to use that individual’s personal data at all? Agree that Facebook will do all it can to make things as hard as possible.
1 reply 0 retweets 0 likes -
Replying to @henryfarrell @bergmayer
There's kind of a three-part answer here. The first part is that ML is really creative at using proxies to get to the same conclusions if you ban it from using certain data as input. We see this come up when you try to eliminate (for example) racial bias from machine learning.
2 replies 0 retweets 1 like -
Replying to @Pinboard @bergmayer
Is this the federated approach that Google has been talking up? My query - and this really depends on the details of regulation and how it is applied - is how much can be done with no personally identifiable information at all.
2 replies 0 retweets 0 likes -
Replying to @henryfarrell @bergmayer
It depends a lot on the allegedly "non-identifiable" information that is passed. If you collect a ton of data on a subset of users, it takes very little information to make accurate guesses about the rest. Age, ZIP code and sex (all 'non-identifiable') would be more than enough
1 reply 0 retweets 0 likes -
Replying to @Pinboard @bergmayer
Yep. This is the question I don’t know the answer to. I can remember not being able to get basic firm level data below a broad level of aggregation in Germany in the 1990s, thanks to data protection law and risk of figuring out which firm had submitted which figures.
2 replies 0 retweets 0 likes
What Facebook has proven very adept at, and one of its activities where we have the least visibility, is finding alternate data sources and inventing ways to correlate them (like mapping mobile numbers to users). Regulation makes this harder, but FB lawyers are paid for a reason
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.