Studies on behavioral advertising that probably already exist

previously: an easy experiment to support behavioral advertising (or not)

In Economic Rationales for Regulating Behavioral Ads (PDF), Pegah Moradi, Cristobal Cheyre, and Alessandro Acquisti ask,

Ultimately, what should regulators do in the absence of macro-level empirical studies that credibly establish how behaviorally targeted advertising affects all stakeholders? Discarding well-founded legal and philosophical justifications for regulation in favor of poorly substantiated claims from the ad-tech industry would be misguided.

The fun part is that these macro-level empirical studies…already exist. They just happen to be behind the firewall at the Big Tech companies and the data brokers. Maybe they’re not fully written up as PDFs with properly formatted references and everything, but a lot of the questions about the harms and/or benefits of cross-context behavioral advertising (CCBA) have answers in data that has already been collected, and it would be hard to believe that nobody has run the query.

The experiment has already been running for decades, because people’s exposure to CCBA varies. Some people, whether because of their general habits or deliberate use of privacy tools and settings, are either exposed to fewer impressions of CCBA, targeted less accurately, or both. But it’s hard to escape surveillance completely.

• Anyone in the surveillance advertising business is going to have a database containing records on both low-CCBA (less data) and high-CCBA (more data) people.

• Big platform companies have “conversion tracking” data that gives them a good idea of people’s shopping habits.

• Platforms and data brokers have data on household consumption habits, income and assets.

So a couple of straightforward studies to do would be:

• Are high-CCBA or low-CCBA people buying more or less from small local businesses? This would be an easy query to run, and it would be the missing piece to back up a lot of the claims about the benefits of CCBA.

• Do low-CCBA or high-CCBA people build higher net worths over time? The Big Tech companies know how much data they can collect on you, and they know roughly how much money you have. Are the people who have had more data collected on them over the past five or ten years richer today? Pro-CCBA papers already point out that wealthier people have higher privacy preferences, but which is the cause and which is the effect? More: building wealth the privacy way?)

The data exists, and turning the data into publishable research is not that much work for those with access to it. If the answer to either question favored the high-CCBA answer, it would be a slick report in every state legislator’s inbox by now. The lack of a report on either one of these is more informative than the actual report would be.

So why aren’t we seeing empirical studies on benefits of CCBA to legit small businesses or for consumers?

Maybe deceptive advertisers, who don’t have the money and time costs of delivering a legitimate product, are able to put more effort into working “the algorithms,” and benefit from an asymmetric flow of customer data away from legit businesses? Maybe as people use surveillance apps more, and are exposed to more CCBA, they put more of their money into drop-shipped crap and in-app purchases, and lose more to scams and gambling, so have less left over for legit shopping? Because every transaction has two sides, an outcome that is worse for legit businesses—fewer sales—is also worse for consumers—fewer beneficial purchases. More: accounting help needed

Bonus links

The Future Of Brand Creativity Belongs To The Small And Reckless by Adel Borky. Loss aversion explains almost everything. The perceived downside of not doing AI feels immediate and existential: angry boards, nervous investors, headlines asking if you’re “behind.” The downside of doing it badly is delayed, distributed, and someone else’s problem. So holding companies overreact. They centralize. They standardize. They automate. They turn a flexible tool into a rigid system and call it transformation.

The one problem with Russia’s shadow fleet Europe still hasn’t addressed by Ben Harris. These countries shirk their oversight duties by enabling barely seaworthy ships on their registries to carry inadequate insurance from unreliable and under-capitalized Russian insurers. The result is a costly environmental tragedy in waiting….

The Measurement Caddies by Brian Jacobs. First, today’s biggest players don’t care about the rules. Question the principles, change the language, keep the data hidden. Do what’s best for them regardless of any higher good. Move fast and break things, as Facebook had it.

The radical confidence of Generative AI and MAGA (Spilling Ink #14) by Dan Brown. They lie and they double-down on the lie when pushed. They never admit the need to investigate, much less admit that they don’t know. (Related: fix Google Search)

Sports betting reshaped newsrooms, and it’s “a little gross.” Now, here come the prediction markets by Sarah Scire. Many journalists told me they view gambling partnerships as a necessary evil, since their companies are fighting for survival and it’s difficult in that position to stand on principle and turn away a new, massive revenue stream. That said, there is definitely concern that journalists are being used to make gambling seem innocent and harmless, especially for young people.

With ‘KPop Demon Hunters,’ Korean women hold the sword, the microphone — and possibly an Oscar by Hyounjeong Yoo. As “cultural diplomats” both on and off the screen, Huntrix carry not only entertainment value but also the symbolic labour of representing a nation to a global audience.

Your Search Button Powers my Smart Home by Tom Casavant. So, I started exploring the Wide Wide World of Customer Support Chatbots. A tool probably used primarily because it’s far cheaper to have a robot sometimes make stuff up about your company than to have customers talk directly to real people. The first thing I discovered was that there are a lot of customer support LLMs deployed around the web…. So I started collecting them all. Anywhere I could find a Chat with AI button I scooped it up and built a wrapper for it. Nearly all of these APIs had no hard limit (or at least had a very high limit) on how much context you could provide. I am not sure why Substack or Shopify need to be able to handle a 2 page essay to provide customer support. But they were able to. This environment made it incredibly easy prompt inject the LLM and get it to do what you want.