Here are my thoughts on specific technologies, companies, privacy, and larger trends as it relates to recommendations technology. Below are some questions that came up from the previous blog post Recommender Systems and You. Ok, let's dive in!
Question: "I know many people (myself included!) who often wonder how on earth these recommendations manage to appear so… magically. I completely relate to the red-shoes anecdote." [The red-shoes anecdote refers to the situation where you tell someone in your house about a novel phrase you didn't search for, like "red shoes" and then you start getting recommendations online for red shoes.]
My answer: In the red shoes situation, there's one specific and simple way this can be achieved. There are other ways, but I think this one is the most straightforward. You talk to someone in your house about red shoes. The person you talk to then does a, for example, Google search for "red shoes." The next time you bring up a page controlled by Google such as youtube.com, you see a video near the top that's about red shoes. How did that get there?! Simple. Google knows that you and the person you talked to about red shoes live in the same place. How does it know that? You both share the same IP address. The IP address is a numerical address like 25.112.33.201 that identifies your house on the internet. At your house you have a wi-fi modem. That modem projects that address, for example, 25.112.33.201 to the world. That modem also must allow for multiple devices to access the internet from your house. Your modem uses what's called Network Address Translation to translate a single public IP like 25.112.33.201 to multiple private IP addresses, one for each internet-accessible device in your home. Google cannot see those private addresses. They are private to your home. Google can see the public address of 25.112.33.201. If somebody from that IP address (your housemate) searches for red shoes, Google will happily send a recommendation for a red shoes video to someone else from the same public address (you). Why? Because Google knows people in a household talk to each other and they reckon the person who didn't search for red shoes might want to learn more about them based on a probable conversation that Google guesses might've happened.
This is a good example of how a first instinct might be to think that Google is listening to conversations via microphone and using that information to send ads or other recommended content. In my opinion, that would be a LOT of processing power Google would need to do just to figure out what kind of ad to send to you and a couple billion other people! I think they prefer much simpler ways to determine what ads to show you.
Q: "I’ve always found these suggestions intrusive and have often wondered how they align with privacy, both from an ethical and a legal perspective."
A: Ok, ethical privacy and legal privacy. Big topics! The purpose of good / scary suggestions is to try to show you something you'll like enough to click on or otherwise engage with. Ethically, the big players will use anything at their disposal to help show you better content that they hope you'll engage with. They for sure invent new tech and ways of tracking you across the internet to help achieve their goals. I'm not sure that ethics is a consideration here, for the most part. I do remember a time in the late 90's I think when Amazon would show you recommendations on other companies' websites and at the time that was considered very creepy by a lot of people. They ended up stopping that particular practice at least for a number of years. That's the only situation I know of where ethics changed the behavior of one of the big players. These kinds of things change over time, too. People get more used to the idea that when they purchase something from Amazon, Amazon is going to use all of that data to try to sell you more stuff. And more people know now that if you're not paying to get a service, then you and your behaviour are the product. That's all part of the value proposition of using various services on the internet for free.
From a legal perspective, companies absolutely push right up to, and over, the edge of legality all the time. GDPR in Europe is one attempt to legally require the site owners to give consumers more control over personal data. As far as I know, there isn't much else in terms of legal avenues to protect the consumer. The sites themselves have Terms & Conditions that often outline in vague terms how they'll use your data and what you legally agree to by using their site. Of course, the legalese used in those T&C's is quite broad and heavily favours the site. One trap is the "Necessary cookie" that you're often given the choice to accept when you first visit a site. This a required cookie in order to use their site. However, the site determines what is "necessary" and can define necessary to be whatever they want in terms of info they can store about you in a cookie on your machine. Sooo, yeah, not much there from a legal standpoint in terms of intrusive recommendations, afaik.
Q: "I’m curious about your view: how do you see the tension playing out between the major players — big tech platforms who rely on targeted recommendations to drive ad revenue, the(ir) lobbyists, and legislators who are navigating complex questions of technology and extraterritoriality."
A: Legislators can't keep up, lobbyists are the real power, and targeted recommendations are only going to get more targeted in the future. The tech moves too fast and too broadly for the law to keep up. This is only going to speed up and perhaps come to a head when robots compete with humans for old-age pension.
Lobbyists have too much power. It's lobbyists that push for and get special treatment for the big tech companies, including OpenAI now suggesting they get guarantees from the US federal government in case OpenAI goes under. OpenAI is now trying to position themselves as "too big to fail" with the federal government and they are paying a lot of money to Congress to help get that point across.
Targeted recommendations are going to be using more and more data. Now that LLMs are here, that's a treasure trove of personal information! OpenAI recently released a web browser called Atlas that merges your ChatGPT conversations with web browsing. I think they're trying to make it easier for people to get information in general AND they want to grab all kinds of web browsing data. When (not if) OpenAI starts selling ads, they'll be able to command top dollar because of all the personal info they've been collecting. Maybe that info will be anonymized before it gets to the advertiser. Maybe.
Q: "And in the background, the citizens, split into three groups: those who increasingly feel overwhelmed or frustrated, those who genuinely enjoy the shopping experience and the red shoes, and the passive group who may not fully grasp what is at stake or has got bigger fish to fry."
A: I'm going to cheat here and say that all three groups are rolled into each of us as individuals, to varying degrees. On any given day, I may be at 10% frustrated, 50% enjoying, and 40% not understanding / don't care. Most days recently that's where I would be. In the future, I'll probably be higher on the "not understanding / don't care" percentage. These topics and the technology behind them are huge. Twenty-five years ago I felt that cookie technology was huge and creepy. And today, cookies are the absolute least of our worries! Why? The tech has progressed so much that getting worked up over cookies is like a tempest in a teacup.
In the end, many of these issues come down to the value proposition for the consumer: "As a consumer, am I getting good value for giving up some info?" Very few people actually think that, or could quantify it even if they do think it. But people do have a strong gut feeling for whether or not something is "worth it" whatever it may be. If enough people don't like something, business will drop it and move on to something more interesting. So I guess that's my overall take. I believe in the market dynamics for adoption or rejection of any technology. If a government wants to step in and do something, that's fine, but I think it's rare that that's effective. Often when government steps in the force is too heavy-handed or not effective enough. I'm always open to new ideas, though.