Boldist - What the FLoC? How Should We Feel About the Federated Learning of Cohorts?

What the FLoC? How Should We Feel About the Federated Learning of Cohorts?

Google has launched a trial of an experimental tracking feature called the Federated Learning of Cohorts (FLoC), a new system to track users and send relevant advertisements their way. The trial currently only affects 0.5% of users in select regions (including the United States). You can check to see if you are currently part of this trial at amifloced.org.

What Is FLoC, and How Should We Feel About It?

A lot of rapid change has happened in 2021 in terms of data privacy, digital data collection and digital marketing. Apple announced they were changing the way they collected and stored our personal data on iOS devices, starting a data PR battle with Facebook. But as vocal as that back and forth has been, the biggest change proposed so far is coming quick and quiet from Google.

We recently wrote about the ways in which Google collects and uses data, about how website owners can contribute and benefit from this collection, and how consumers can control what data is collected about them as they interact with certain Google products. But in March 2021, Google announced their response to a post-third-party cookie world – the Privacy Sandbox.

What Is Google’s Privacy Sandbox?

The Privacy Sandbox initiative was announced in August, 2019 as a way to “fundamentally enhance privacy on the web.” The Privacy Sandbox is a project by Google intended to develop technologies that protect consumer privacy and personal data, whilst maintaining the advantage Google has in creating a web ecosystem that is valuable to advertisers, with content that is relevant to users.

Their mission statement is to “Create a thriving web ecosystem that is respectful of users and private by default.” The biggest factor to this is their stated intent to make cookies redundant.

What are Cookies?

Cookies are files created by websites you visit that save microdata from your interactions with a website. Cookies can improve your web browsing experience by recognizing who you are and remembering your website preferences and personal data. But cookies are problematic.

Third-party cookies can glean and store more information about you than necessary. And cookies are saved to the machine you are using, so if you are using a public or shared computer, you might run into privacy issues.

The increased push towards consumer data protections has been a death knell to cookies. In fact, Chrome has called for 2022 to be the year that third-party cookies are removed.

Google, knowing that the day of the cookie is over, has developed a new method to maintain its reputation in delivering relevant interest-based advertising to consumers and delivering valuable traffic to the businesses that benefit from Google’s advertising services. It’s called the Federated Learning of Cohorts (FLoC).

What Does FLoC Mean?

With the phase-out of cookies, the Federated Learning of Cohorts (FLoC) is a privacy-preserving advertising solution. Basically, FLoC is a cluster of people with similar interests. Instead of building a data profile on individuals, Google will hide individuals in a crowd of similar people based on activity gleaned from their browsing activity through data layers without any insight from cookies.

What’s My FLoC ID?

In this new system, users will be assigned a FLoC ID. Your browser will process your browsing history and analyze the traffic and behaviors to then group you in a FLoC. As Bennett Cyphers at the Electronic Frontier Foundation puts it, Google says “at least a few thousand users should belong to each cohort (though that’s not a guarantee).”

Users will only have one FLoC ID at a time. Though different cohort IDs might seem very similar on a surface level, you can trust that the data sets Google has the capacity to build will be rich enough to be hyper-specific.

Is FLoC a Good Move?

FLoC is billed as a “privacy-first alternative to third-party cookies.” The intention is to protect individuals by using on-device processing and keeping a person’s web history private on their browser. The alternative has been tested in Google Ads, and Google has stated that FLoC “can provide an effective replacement signal for third party signals” and that advertisers can “expect to see 95% of the conversions per dollar spent when compared to cookie-based advertising.”

Whilst making the environment safer to consumers, Chrome also states it will “aggressively combat the current techniques for non-cookie based cross-site tracking, such as fingerprinting, cache inspection, link decoration, network tracking and Personally Identifying Information (PII) joins.”

What About Retargeting?

The Privacy Sandbox has been working on new privacy-preserving methods to store site visitor information, proposing TURTLEDOVE: where companies can target ads to audiences they have built “without the browser revealing anyone’s browsing habits or ad interests.”

The latest iteration of this concept is FLEDGE, which proposes a trusted server, “as defined by compliance with certain principles and policies,” designed to store and protect certain data instead of storing it on a browser.

What About Conversion Tracking?

Again, there are several proposals to what conversion tracking might look like in this new environment, including privacy-preserving techniques such as adding noise to data and limiting the number of events the API can send at a time. Right now, these concepts are still being prototyped and investigated.

Can We Trust Google With FLoC?

Google has a track record of doing the things they say they’re not doing. For example, there is currently a class-action lawsuit that claims Google surreptitiously collects user data in it’s Incognito Mode. Incognito Mode claims it won’t store your web history or identifying cookies on your browser, but most users of this function likely don’t realize there are many other ways Google will track your behavior on that site, including through Google Ads and Google Analytics tracking codes. As we pointed out in our recent look at the variety of ways Google tracks your behaviors and collects your data, they can still use other identifiers to know who you are, including server and geolocation data through public Wi-Fi networks.

There have been a lot of criticisms about FLoC. It is currently being tested simultaneously with the existence of third-party cookies, so the argument can be made that your FLoC ID (the cohorts you are in) can inform the personal profiles.

There is also the issue of developing “sensitive” cohorts that group people based on information they might consider private and damaging. As Dr. Augustine Fou puts it in his contribution to Forbes:

“Whether a behavior is “sensitive” varies wildly across people. My mom may not find her interest in “women’s clothes” a private part of her identity, but my dad might! Similarly, an adult happily expecting a child might not find their interest in “baby goods” particularly sensitive, but a scared and nervous teenager might.”

How This Serves Google

Google’s main resource is data. By removing third-party cookies from the digital environment, they are effectively limiting competition. I think there is more of an antitrust case against the Federated Learning of Cohorts, and that we should be concerned about the kind of cohorts they are building or that we would become part of.

That’s likely why we’re not seeing Google testing FLoC in Europe right now. Europe has gone after Google for GDPR violations hard in the past few years, as well as for antitrust violations.

FLoC is controversial, and it remains to be seen whether it will become the de facto replacement for cookies or end up an antitrust case in European courts.