Magic Cookies

Can we get better at anticipating the unintended consequences of new technology?

As some of you know from earlier editions of this newsletter, I’ve been working on a series of profiles of under-appreciated pioneers from the history of software that I’ve been collaborating on with the folks at NetGuru. The latest installment is a story about Lou Montulli who happens to be one of the most fascinating figures from the early days of the Web for several reasons, in part for the key innovations that he added to the Web experience during that period, and in part for the way one of those innovations ended up triggering a chain of events that directly undermined what Montulli was trying to do with the innovation in the first place.

I’ve had a long interest in unintended consequences, dating back at least a decade to some of the stories in How We Got To Now. There’s something narratively compelling about chains of unlikely causation; when I talk to people today about HWGTN (either the book or the TV series) they almost always bring up the stories about how the invention of air conditioning helped elect Ronald Reagan president, by triggering a mass migration of older, more conservative voters to Sun Belt states that had previously been too hot and humid for many people to tolerate; or how the printing press led to the invention of the microscope and telescope, because the new practice of reading revealed to a meaningful subset of the population that they were farsighted and thus needed spectacles, which created a new industry of lensmakers, who then started experimenting with other uses for glass lenses.

Those stories are fun because by definition there’s an element of surprise to them; they cut nicely across the usual topical categories that govern a lot of historical storytelling, jumping from an engineer in Brooklyn trying to deal with the humidity problem at a local printing press to large demographic shifts in American settlement patterns, all the way to presidential politics. But of course, they’re also sobering in that they remind us that most significant innovations end up generating second- or third-order effects that are difficult to predict in advance, that in many cases are not aligned with the values of the original innovator who set it all in motion.

Now, I was already interested in Lou Montulli’s story because he was an integral figure in a period that I lived through in my early twenties — that stretch between roughly 1990 and 1994 where it was clear that some kind of global information network was going to become a mass market phenomenon, but what that network would actually look like—who was going to own it, what the dominant user interface was going to be—was very much up for grabs. As I discuss in the profile, Montulli was the co-creator of one of the very first hypertext browsers, Lynx, and played a key role in the early development and dissemination of the HTML standard. But it was as an early employee at Netscape that Montulli unleashed his most controversial innovation onto the world: the Web cookie.

It’s important to remember that in the early days of the Web, the medium had an almost comically short attention span. You’d request a page, and a server would send you back the information you’d requested, and then immediately forget about you. The idea of doing something as basic as flipping through a few pages at a Web-based commerce site and adding items to your shopping cart was unthinkable, because each time you clicked on a link, the web server at the other end would have no record of your previous activity on the site. As Montulli later described it, “[It was] a bit like talking to someone with Alzheimer's disease. Each interaction would result in having to introduce yourself again, and again, and again.”

Montulli’s cookie was designed specifically to give Web servers something like a short-term memory that would enable user “sessions” where you would have some kind of persistent identity as you interacted with a single web site. I go into more of the technical details of all this in the profile, but the key point here is that the cookie architecture was designed specifically to protect user privacy. Whatever you did on one site was not going to be visible to any other site on the Web. But because of another, unrelated attribute of HTML—the way it handles image requests—it turned out to be possible to build a system that could use cookies to track user behavior and identity across a broad network of sites, a technique first pioneered by the online ad company DoubleClick, later acquired by Google. In the end, a tool designed explicitly to maintain user privacy ended up being a crucial ingredient in the targeted, cross-web advertising and user profiling that became dominant in the social media age.

As it happens, I’m working on a number of projects right now that explore how new technologies trigger unintended consequences, and how we can get better at anticipating those downstream effects, and hopefully at preventing some of the most deleterious ones. (Some of this came up in the discussion of pre-mortems and scenario planning in my book Farsighted from 2018.) One of the things that I thought was striking in reporting the Times Magazine piece on large language models was just how many smart people—both inside Big Tech companies and also in various academic labs—are actively thinking about the secondary effects of AI now, effectively running pre-mortems to anticipate the way these technologies might be exploited to work against the best intentions of their creators. (There was very little work being done in that vein in the early days of social media, by contrast.) We are getting better at thinking about unintended consequences of new technology—that’s the good news. The bad news is that the technology is advancing—and being disseminated—faster and faster, and it’s not clear if our forecasting skills can keep up. So there’s much more thinking to be done on this front, some of which will no doubt appear in future editions of Adjacent Possible.

---

Speaking of forecasting, my first official episode as host of the TED Interview featured the futurist and game designer, Jane McGonigal, who has written a terrific new book all about future-thinking called Imaginable. The whole conversation is a lot of fun and expands on a number of the tools for long-term forecasting that I explored a bit in Farsighted, but maybe my favorite bit is when Jane asked me to imagine some element of my life ten years from now—and because I am a podcast host, what immediately jumped into my mind was the future of mattress technology. And while you’re listening, be sure to check out the interviews with Chandra Prescod-Weinstein and Garry Kasparov, who turned out to be quite prophetic himself in his warnings about Vladimir Putin over the past decade or so.

Finally, some of you might have seen the post yesterday from Evan Williams—co-creator of Blogger and Twitter—announcing that he was stepping down from his role as CEO of Medium, which he founded about a decade ago. I’ve known Ev for more than twenty years, and helped out a little in the early days of Medium when they were first thinking about what the product would be. It’s going to be fascinating to see what he does next given his track record, but for our purposes, I thought I would just share the closing lines of Ev’s announcement:

I also plan to start a new holding company/research lab to facilitate this learning, to be helpful to Medium and other companies I believe in, and to keep doing what I’ve always found most interesting — opening doors to the adjacent possible.

That sounds promising!