Walking the paths of culture
Why we can never design out the necessity of failure, for better and worse
I’m about a month behind reading all the Substacks I follow, so I just this morning got around to reading Visakan V’s post branching paths. It’s excellent, of course, and you should go read it, but for the purpose of setting up my riff on it, here’s a brief summary.
Life is full of possibilities. There’s more possibility-space than we can fathom. There are lots of possible paths we can take at any given moment.
In navigating the world, we face choices with implications that extend far beyond the horizon of our perception. It’s impossible to model the world well enough to choose rationally; there’s too much uncertainty. Instead, the best we can do is keep our pattern-matching tendency in check enough that we don’t chase a mirage into the desert.
we are pattern-matching, pattern-seeking beings, and one of the tricky things about being human is that we will, again, “see what our eyes want us to see” – which is to say, we will see patterns that aren’t “actually there”. But that’s just a feature of the human mind, and once we understand it, we can work with it. We can double-check our observations by testing them, and corroborating them with other people (also a tricky business).
The solution to this problem of taming your pattern-matching instincts is to follow well-trodden “desire paths” through the web of information. Connections that are made again and again are retained; those that are not, are forgotten.
The crucial insights here are twofold; both, to my mind, are fundamental tenets of cultural evolution (like much of Visa’s work, this post points directly at cultural evolution without naming it as such). The first is that humans solve complex problems by applying thousands of subjective decisions independently and combining their results.
A big reason why information architecture is hard is because it’s not enough to simply know how the information relates to itself. Architecture is a people-centric activity.
Desire pathing recruits human users as sensory organs to feel out the edges of reality. It uses subjective selection to guarantee that edifices too grand to fit in a single mind are nonetheless equipped with handles, signposts, and landmarks that make them navigable by millions of people, even those who might only pass through once.
The second key insight is that this process is fundamentally, if often obscurely, evolutionary. In other words, it relies on arbitrary individual variation and the iterative elimination of failed variants. Visa analogizes human cultural paths with ant pathfinding.
The ants deposit pheromones as they go, and the shorter path has a stronger pheromone scent. The weaker trail gets “forgotten”. Forgetting aids remembering! It’s a principle of contrast! [. . .]
I find it important to reflect on this: The ants that went the long way around the obstacle also contributed to the formation of the efficient path. Their contribution was just as important.
From an evolutionary perspective, this is undeniable. Natural selection—forgetting the failed variants and repeating the successes—is the only mechanism by which adaptation can occur. In the context of culture, though, it can be extremely hard to swallow, because we are so enamored with the notion that intelligent individuals can simply design adaptive structures. It is tempting to think that we can put the landmarks on the map and draw the desire paths in ourselves. After all, most desire paths are just diagonals between popular destinations. Do we really have to let people take the long paths, the bad paths, the forgotten paths? Can’t we just make a plan informed by what we already know about human habits? Can’t we spare ourselves the trouble and the pain?
Visa’s answer is of course: no.
I was stressing out for months because i was trying to find the efficient desire paths without the bad stuff. But now it’s becoming clearer to me that I should just make my peace with taking the bad paths too. There is no way of knowing what the right paths are in advance. You can’t order desire paths to emerge instantly. Desire meanders. It takes its time. You have to let it flow where it wants to go. As long as I don’t die, even my “inefficient” Crackle contribute to the eventual formation of the “final” Boom.
This is what makes this piece especially interesting to me. It doesn’t just identify this evolutionary mechanism descriptively, as cultural evolution theory does. It grapples with the paradoxical experience of living in a world full of tools that instantly design solutions to extremely complex problems, while still relying on the most primitive, tedious, agonizing form of iteration to accomplish anything new. It correctly identifies that, no matter how much we build with evolution, we never get any farther away from it.
The only thing we can do to change our relationship with evolution is to evolve new codes that let us confer new semiotic degrees of freedom on the repetitive, frustrating, dead-end probing of iteration. This is, in cultural evolutionary terms, what it means to “pave” a desire path. We name the nodes and share our routes and light the way so more and more people can follow. And of course, you don’t pave a path unless you want people to keep walking it. The goal isn’t to stop travelling. It’s to make it faster, easier, safer, and perhaps above all more fun, to explore further from the old routes.
What does this mean for the designers?
While I was writing a comment on Visa’s post, it occurred to me that this seemingly trite observation (culture is good because it lets us make more culture) accorded naturally with an extremely contentious conclusion: AI creative tools are a tremendous opportunity. After all, AI is an inflection point in the rate of code creation. It lowers the costs (and thus risks) of creative expression in styles and aesthetics that previously required climbing steep learning curves. The result is that parts of the creative landscape that were minimally explored by intrepid auteurs are now swarming with thousands of “ants.” AI artists are going to make themselves at home in these distant realms, fill with desire paths, pave them with tools and style vocabularies, and they will become, to some degree, accessible to anyone who wants to show up and poke around.
It’s easy to get wrapped up in the romance of collective creation and discovery and forget that this idea is contentious specifically because some people want to be designers. Architects and planners aren’t necessarily trying to find the most efficient paths. They’re trying to create something that pleases their own tastes, something photogenic and elegant and expressive. Individual design still involves finding desire paths—creation is still inherently evolutionary—but you make every trip yourself (or with your trusted collaborators). That’s arduous and onerous and involves tremendous investment of personal time and energy. But until now, the social contract was that in return, you’d end up with a landscape drawn entirely by your own subjectivity.
Importantly, you could trade access to that landscape for wealth and social prestige, and it’s quite obvious that many stridently anti-AI creators are paranoid about losing their privilege to control both. But I suspect the thing many creators find most invasive is simply the loss of the privacy of their domain. Image synthesis programs confer new degrees of freedom on text by associating keywords with particular aesthetic outputs. Among the most useful of these keywords are simply the names of artists. The idiosyncratic style an artist painstakingly defines over the course of their entire career becomes accessible as one in a palette of thousands. There was a time when Wes Anderson was the only person who controlled the images made with his personal visual language. Now, thousands of grubby strangers have used his name to tastelessly connect it with every major IP. It’s not hard to imagine why someone might find that annoying.
Still, it’s important to keep our eyes on the horizon. It’s easy to get parochial about the yard you’ve spent your life gardening, but the space of artistic possibility is just unfathomably vast. AI tools enable us to create fundamentally new kinds of art—sprawling collaborations with high frequency, small-scale iteration and influence and no fixed endpoint—we have no current model for. Older creators, whose notion of art was formed in a paradigm of lone auteurs toiling alone to periodically release single polished works, may be the least able to perceive the potential in this totally new paradigm. They are so intimately acquainted with the emotional ups and downs of evolutionary iteration in their own medium, so attuned to the subjectivities at play as they trace their established routes across the contours of their work, that they struggle to imagine that the underlying process is universal, inevitable, fundamental and therefore fully transferable to new modes of creation.
Which is frustrating, because that same depth of experience means these artists are in the best position to apply their taste and instincts to use AI to make work that stands alongside or above the best traditional creations. I can’t help but hope that many of them will glimpse the possibilities and stake an ambitious claim in these new realms. We are, after all, the first people to have the chance.