[Note: This is the last 7-item listicle for now, promise. I also apologize for the cheems mindset on display in this post.]
In a footnote to my summary of 2021 I complained about how irritating it’s becoming to have a boss.
I struggle to be motivated to do things when I don’t get to do them the way I want to and in my own name. It’s draining to subsume yourself under some employer’s banner and not have your work be your own — not legally, not socially, and not psychologically. It just seems improper for an actual adult. It’s textbook alienation and this is probably the most Marxist I’ve ever sounded. If you live with your parents you’re not fully an adult. And to me it’s looking increasingly like if you’re not self-employed you’re not fully an adult either. Having a boss is like having your parents telling you what to do. Proper adults don’t have bosses. Proper adults don’t get “evaluated” on their “performance”. I realize this is completely unreasonable, by the way, but I’m just feeling it, regardless of how much my superego tells me to stop.
It’s become increasingly clear that the only route forward if I’m ever to be satisfied with my work situation is self-employment. Find something to do. Find something to sell.
Too bad I hate a few things about as much as I hate answering to an employer — things like admin, sales, and taking risks. But I can deal with those, I think, as long as the basic idea is strong. Shouldn’t be too hard. I’ve got a lot of ideas, right?
Unfortunately most projects come up with are small and unprofitable. I have ideas for novels, infographics, pieces of music, puzzles and games, and decorative objects. The ROI on such things tend to be small and you need to focus on building your skills for a long time to get good enough at those thing to even have a slight shot — and I’m terrible at sticking to things without constant external validation.
Some project ideas have a different problem. They’re more creative and interesting but unrealistically ambitious and/or with little economic value. Most of them is some version of “make something that’s way more interesting to make than valuable to use”, and the remainder can be summed up as “not (to me) worth doing unless done so well that the cost becomes absolutely prohibitive”1.
A few of those has stayed with me for years, popping up now and then. “Maybe I should try to do it? Or do some simplified version?” but of course not. Instead I just fantasize a little about having the time, resources and focused resolve to take a sufficiently sustained stab at any of them. That has its advantages, like not having to do any of the difficult or boring parts.
Still, I want to share them, because for some reason that feels rewarding.
In no particular order:
1) The artistic taste landscape
I’ve been fascinated by recommendation engines since I first came across one about 20 years ago. In the mid-00’s I had quite serious plans of building one — well, several. The fascinating thing about finding patterns in what kind of movies, music or books people like is that it’s like finding the structure of thought itself. People who like this also like that? It’s like getting a glimpse of a hidden order, and I desperately want to see more of it.
The big problem with recommendation engines, as they exist, is that the data they have to work with isn’t very good. Good enough to make recommendations, perhaps, but not good enough to map out the true structure of artistic preference. You’d need much more detailed, complete, and reliable information. You’d need to know every movie/book/piece of music every user had ever consumed, when they did, and what they thought of it, along several dimensions. And ideally you’d also know what they’d come across and chose not to partake in, and why.
Then you’d might be able to find the big, hidden structure underneath it all, and not just pull out this or that useful tidbit or crude correlation. The fact that it’s impossible to get this material, or anything near it, saps away about two thirds of my passion for the project.
I did have a pared-down idea I was quite intent on going for around 2009-2010 or so. I tried to “decide” I was going to do it, force myself commit and pull myself out of my lack of focus and motivation. I held on for a few days and made some notes in a pad I still have in a drawer somewhere.
The main idea was to focus on taste in stories, regardless of whether movies, tv series or books was the medium, and to not rely on data so much for extracting a pattern or model, but to work it out from (entirely guessed) first principles.
I hypothesized that people’s taste in stories depended on their personalities, as described by psychologists (the Big 5 model plus Simon Baron-Cohen’s Empathizing-Systemizing spectrum). In an effort to confirm this I looked for academic research on the topic, which was underwhelming. There’s not a lot of it, and what does exist suffers from that well known lack of good, precise data. Results tend to be about as detailed as everyday intuitions and stereotypes (often on the “genre” level). Thanks for nothing.
But you can try to get further, right? You can just make wild assumptions and see where they take you. I started on a model where I’d translate personality traits into qualities of stories, largely determined by whether it made sense to me. Then I would (but I never got that far) mark stories based on whether they had those qualities. In the end you were supposed to take a personality test and then get a list of stories that would fit your personality profile. It could’ve worked if these personality-trait-inspired qualities actually do correspond to what people like or don’t like in stories. I don’t know if they do.
Introversion/extraversion could translate to a taste for stories from one person’s perspective (think Taxi Driver or Crime and Punishment) or about relationships among a group of people (think Pride and Prejudice, or any soap). “Openness to experience” could mean fanciful vs. realistic stories, or (more like Myers-Briggs “sensory vs. intuitive”) abstract or idea- and theme-driven vs. sensory stuff like action, beautiful scenery, , nice costumes/cool spaceships, evocative prose or cinematography, etc.
I had less clear thoughts about other personality traits but agreeableness and neuroticism could be connected to enjoyment of conflict vs. cooperation (naivety vs. cynicism), or enjoying titillation by upsetting themes and events vs. wanting happy endings. High conscientiousness might lead people to enjoy stories where people are competent, composed and serious about their tasks, as opposed to say, a typical teen drama or dysfunctional relationship story.
The empathizing/systemizing axis could be used to distinguish between those who want logical plot progression with clearly established rules and strong quasi-mechanistic sense of cause and effect (think murder mysteries, spy stories or hard SF) and those who are looking for vicarious experiences, who want to identify with characters having strong feelings, follow them on their journeys, and share their sorrows and triumphs second hand (think coming-of-age stories, family chronicles or misery porn)2.
In reality you’d want many more axes and story attributes, and a lot of bright, dedicated people to score stuff accurately.
2) The validity web
The whole of academic research is one giant cluster of partially deferred credibility. A paper argues that perhaps, under certain circumstances, X is true, and backs it up with some evidence the authors have produced. To support their argument, and to set up the whole conceptual system in which the claim makes sense, they also cite papers A through P.
The credibility of any one paper, or textbook, or popularization, therefore relies both on the quality of cited papers and the correctness of the citations themselves — i.e. if cited stuff actually says what it supposedly says. Cited papers in turn depend on its own citations being correct, and so forth. The whole thing becomes impossible for anyone to check.
Peer-review is supposed to deal with this, but it doesn’t and of course can’t3. There’s not nearly enough time nor will to be as thorough and critical and you really need to be to fully assess the value of a piece of research, and by extension everything else it’s later used to support through citation.
Psychology, for example, tries hard to be science but struggles to outperform our already fiercely perceptive social intuitions by using scientific methods that barely work because running sufficiently controlled experiments in sufficiently natural environments is impossible for reasons of practicality, costs and ethics. Its often weak results become the object of speculation far beyond their validity, and escapes into public discourse a lot more than say research on worm metabolism or Minoan script does.
The replication crisis is well-trodden ground by now. Lots of things we used to believe was solid enough put in textbooks and thinkpieces and God knows how many Powerpoint slides turned out to be flimsy bullshit. Wouldn’t we want a way to determine the damage done whenever something turns out to be ~worthless? And I’m not just talking about the fraudulent or non-replicating. I’m talking about shit that’s just weak. Surveys with bad, vague questions (which is most of them, I say partly as a professional survey-wrangler). Words that get redefined enough to lose their normal meaning; Nutrition and lifestyle studies based on self-reports (please); Clear political/ideological bias; Obvious confounders unaccounted for.
Peer-review isn’t necessarily enough to weed out weak research because often the people doing the review are guilty of the same sins, and if weak methodology was actually banished it would take too much of research output with it. The pseudonym Literal Banana (@literalbanana on Twitter) wrote The Extended Sniff Test as an example of how to evaluate research as a nonspecialist. Written in an imitated academic style, it’s an absolute gem.
The Extended Sniff Test (EST), introduced here, is a formalization of the practices the author uses when volunteer-fact-checking scientific claims. It is performed after an initial sniff is taken of the claim to be studied, if the sniff reveals the possible presence of bullshit. The EST contains three parts: Double Contextualization of the claim in question (DC), a Noun Abuse Assessment (NAA), and a lay literature review.
And some lovely one-liners:
General claims have often experienced traumatic amputation of their contexts.
A word is being abused if ordinary intelligent readers would take it to have a particular
meaning, but upon investigation, the meaning is found to be missing or changed.
I imagine an institution full of smart generalists (and specialists when needed), who do Extended Sniff Tests on published research with relevance to society at large (psychology is a good starting point). And I don’t just mean giving every paper a credibility score, but rating it on various dimensions like construct validity, robustness, vulnerability to alternative interpretations, etc. And each edge in the network would have to be judged as well, on several dimensions, like how pivotal a a citation is to the conclusion, and how well represented the cited claim is, etc.
With this, whole clusters or subfields — branches on the tree of Science, if you will — could automatically be identified as rotten if enough corruption spreads through it. It would make it harder to do knowledge laundering by exploiting illegibility and obfuscation, and it would reduce the incentive for weak research.
Citation network projects do exist, and perhaps there are even examples trying to let “badness” found in one paper propagate to other papers that cite it, but this unworkable idea is more ambitious, considering the size of the institution required. I consider it unworkable because of how much research is produced. It’s not just papers but popular articles and books too, and it takes a lot of work to evaluate things4. It’d be very hard and expensive to find enough people with the right skills and sufficient integrity, and then pay them. It’d also be extremely hard to not introduce political or ideological bias into the review process itself — safeguards against that would add another layer of complexity.
3) Modern temple
Perhaps the most ridiculously expensive project I fantasize about is a modern temple — a secular building that nonetheless manages to create a sense of the eternal. It would have to be very large, and built very carefully to evoke and preserve a sense of awe, reverence, and mystique.
It’s not an unusual fantasy. Krishnan Rohit at Strange Loop Canon has defended modern architecture from commentators (like me) who lament its lack of beauty, but even he has to admit that while “the buildings I wrote about are undeniably amazing, one difference from the olden styles is that most modern buildings feel expressly prosaic”. He goes on to say that even though many of today’s skyscrapers are incredible, they are indeed “built for quotidian reasons, as office buildings or homes”.
Today I think we all too often feel unsettled to build large things without a utilitarian purpose. Worship isn’t as favoured as this purpose any longer either. So we hedge. We make museums with lovely gift centres as a sort of compromise. And it doesn’t work.
No it doesn’t. The sublime is fragile. Otherworldiness collapses upon contact with the mundane, the utilitarian, or the administrative. The greatest challenge for a modern temple5 would be to achieve the right mood, the right vibe.
The skills needed to design evocative environments do exist. Game designers and other artists are very good at it. It only took me a few minutes to find these pictures of gorgeous spaces.
Is anything like this easy and cheap to build, in a solid and authentic enough manner to not make it feel plastic and fake? Noooope. And you’d have to erase all visible intrusions of practicality, like electrical wires or light switches, toilets, signs, garbage bins.
However, the physical environment is only part of it. There would have to be rules. Weird and draconian ones. You don’t want it to be like an overcommodified church or museum with hordes of tourists sauntering around, wearing backpacks and snapping photos. In fact, no pictures at all. No cameras or electronics of any kind allowed in. No bags, no coats or jackets. Maybe no talking, even.
No pictures because you don’t want people to know in advance what it looks like inside. There should only be descriptions. The place should be large and complex, with many different spaces — from a glass tower to dark catacombs, labyrinthine systems of rooms with hidden reading nooks — and no maps. You have to explore.
Obviously breaking any rules would get you banned. And time inside would have to be rationed to keep it from getting crowded. You’d have to apply for a slot. Maybe there could be a more everyday, liminal section near the entrance, like a café everyone can visit, and keep restricted areas further inside, maybe even some special places you could only access on your second or third visit, etc.
Unless I become a multibillionaire, and armed with the power to ignore things like building codes that stand in the way of my vision, I guess I’ll have to keep fantasizing.
4) A simulated world
As a teenager I enjoyed playing strategy games like Civilization III and Sid Meier’s Alpha Centauri. Not, however, quite as much as I enjoyed fiddling with and modifying them, changing the names, looks, stats and effects of technologies, policies, buildings and military units6. More than once I spent quite a lot of effort on a new mod design only to find out that playing it once it was done was not the fun part.
I read forums too. I especially enjoyed discussions about what changes people wanted for future iterations of Civilization. More realism was popular, which stood against the mantra that it was “a strategy game, not a simulation”. Maybe, but the fact is that many of us who enjoyed Civilization were more attracted to the simulation aspect than the competitive game aspect. To be honest, some of my own thoughts for the ideal version would, if taken to their logical end, have eliminated the actual players.
No wonder I spent the years in the beginning of my university education thinking I should’ve picked something with more programming in it, so I could complete a massively unworkable project: en entire simulated world. It’d be in the style a strategy game but far more complex, pulling a lot of known mechanisms together into a single system.
The physical world would be created randomly, and simplified but understood models of plate tectonics would form mountains and volcanoes; climate zones would form, including rainfall patterns and with it rivers, lakes and deserts. On that world little societies would emerge, based on where food would be plentiful. They’d grow and shrink based on known models for demographic dynamics, disease, war, etc. Their languages would develop, branch, drift and borrow words from each other, again based on known mechanisms.
Transport infrastructure would determine were cities spring up: rivers, natural harbors, roads and eventually railway junctions. City plans and architecture could be dealt with using some of the naturalistic street plan generators that existed already 15-20 years ago7.
You’d be able to explore your world in many different ways, zoom in through a Google Maps-like interface and study roads and cities. You’d see how trains, planes and ships cris-crossed the planet, you’d get stats on population, wealth, military might, education, and so on. You could see the flags of countries, perhaps hear their national anthems, written by music-producing AI:s. And so on.
The hardest would be capturing cultural values, social roles and the resulting politics and their effects on economics and technology (including feedback loops), but existing research could be used to create at least serviceable (if basic) models, I thought. The idea wasn’t to copy the real world as such, but to just have an artificial world — a toy — that felt dynamic, interesting, unpredictable and “alive” in the way strategy games never can because they have to work as games (i.e. they have to be fair and sufficiently under the players’ control).
What’s that? You’d want to know what the point would be? The use case? Well there’s the problem I have a lot of the time… its grotesque effort-to-value ratio is like that of a Rube Goldberg Machine and its raison d’être would be the same: the mere joy of building it and setting it off. Putting down the rules for a complicated simulation, pushing the start button and seeing spring alive and sit right on that sweet spot of being just recognizable as what you made but also so much more… it’s exhilarating. One of life’s great pleasures.
Could it be done? Well enough to be an interesting and inspiring work of art, theoretically yes. Could it be done well enough to produce valuable insight, and at a cost not orders of magnitude more than it’d be worth? No.
5) Exploring networks as an art medium
Speaking of the joy of seeing a system you’ve created “come alive” and surprise you in interesting ways:
Early in my career I worked on building a custom analysis application for the small company I worked for. Because there were advantages to getting exactly what we wanted and nothing but, and also because I had engineer’s disease (“I’ll build this myself instead of adapting some off-the-shelf solution because that’s totally better, and more fun for me”) with no one to call me on it, I coded it all by hand. One of the things I built, and in the process learned to experiment with, was a network simulator using force-directed layout algorithms.
You said what? Pretty pictures of complex networks tend to be the result of simulations, where every node in the network repel every other node, while edges act like springs or rubber bands, pulling nodes towards each other. Implementing such an algorithm means going through each node in turn and adding up all the forces acting on it: every other node pushes a little (in most versions they push harder when closer) and every attached edge pulls a little. You get a resulting force for each node and move them a little bit in that direction. Repeat until everything stabilizes.
The end result is often pretty but the process itself is perhaps even more mesmerizing, as clusters of densely interconnected nodes clump together and unconnected strands are blown apart. Intricate structure is gradually revealed and perfected in something that reminds you of embryogenesis.
It’s mostly (~exclusively) used as data visualization, but I’ve always thought it has great potential as an artistic medium. Think of the core mechanism as a raw material. Construct artificial datasets that create effects you likely wouldn’t see with anything natural. Add, remove or alter nodes or edges on a preprogrammed — fixed or probabilistic — schedule, or make them respond to what happens to them or near them (or to user input!). The possibilities are nearly endless. With the right programming node clusters can get molecule-like properties, producing a quasi-chemistry.
I think of it as fertile ground because the possibility space seems too vast to get an idea of from the start. And that’s the case for incipient art forms. They take a long time to mature, as practitioners slowly learn what works and what doesn’t, build upon what they’ve learnt and what they’ve seen others do. They acquire a lexicon or toolbox of techniques and mental shorthand that makes it possible to push forward, far beyond what anyone was capable of in the beginning. Only our own brainpower and accumulated experience stand in the way of creating truly hypnotic visual marvels out of this.
Conway’s Game of Life is dead simple: it’s just a grid of cells that can be either living or dead and whose status is determined by three short rules. Given that, the vast potential inherent in it is just astounding. It spawns a menagerie of phenomena, explored over decades, and by carefully selecting initial conditions you can create almost anything, including a computer that can implement the Game of Life inside itself.
Force-directed layout algorithms can do even more, and with more visual flair. The greatest problem might be that the possibility space for such networks is so vast that it gets completely unworkable to explore in a methodical way. It has so many degrees of freedom that unless restricted in some way, it risks blending seamlessly into “generally pretty patterns”.
You could set them to music, make’m pulsate to complex beats, with movement and color. That’d be cool.
6) Name translator
When my kids were watching PJ Masks I noticed that the character names had been translated in the dub. The green kid, called “Greg” in the original, had been renamed “Jens”. I snickered a little and thought the translator had done a good job — “Jens” does have a very similar feel in Swedish as “Greg” does in English.
It made me think about translating names. Sure there are versions of the same name in different languages — Charles, Karl, Carlos or Karol etc. — but that’s an overly literal way of looking at names. I remember a comic strip making the observation that “Kurt” is a much cooler name in English than it is in Swedish, and it showed that the associations people have with names are much richer and thicker than what you could figure out by just looking at etymology.
People don’t give their kids random names. There are so many factors that go into it, and this has an effect on what names “mean””. They suggest age, since names are subject to fashions. A few explode onto the scene and burn out in a few years (hello “Khaleesi”), and others stay put on the top-100 list for generations (“James”).
Some names are plain, others make you take notice. Some are posh and others are trashy. Some are stodgy, some are pretentious, and others smell of incense. There are male and female names, of course, plus a small numbers of unisex ones (a few change over time), but even within those categories there are spectra of stronger or weaker masculinity/femininity. Other spectra are cool vs. dorky, attractive vs. unattractive, etc.
If you want to know what name would match yours the best in another language and in another country, you’d need to take all this into account. You’d have to collect information on many, many names from many countries, through long and boring surveys where people tell you their feelings about various names. You’d complement this with public statistics that may or may not be available and/or comparable across countries (as a data analyst this gives me anxiety). And you’d have to do it again every few years to keep up with changes.
And for what? You’d have a toy that could be a bit of fun. Maybe it could even be somewhat useful for marketing and artistic purposes to know the connotations of names in different countries and languages. But the workload and cost to ensure decent data quality would be prohibitive, although perhaps not as severely as other items on this list.
7) The mind graph
I thought about this 20 years ago after walking home after a History of Ideas seminar (don’t know why I remember that). It’s in the same vein as most others on the list, in that it combines personality, meaning, and visualizing complex information. It would be a big, near-complete picture of a person’s mind, as revealed by a very detailed questionnaire.
It’d be a huge collection of little dots. Artfully connected, they’d look like a giant bundle of grapes. Each blot would be either present or missing based on whether a thing/interest/preference etc. had been selected on the questionnaire. Do you like peanuts? A little spot right here. Symphonic doom metal listener? That’s right over here. Oh, you play guitar and speak Polish and Greek? Little spots for those too. Hundreds, or more likely thousands, of little dots in one very big, and very personal, work of art.
It was before I knew how to program so in my mind I’d make it by hand. I’d customize based on the answers, but there’d be a basic template determining the placement and connection between dots. Basically, large differences between people would correspond to large differences in how the picture looked, and once you’d learn how you could get a feel for what kind of person someone was by just looking at it.
This was also before I’d learned to abandon ideas immediately (the “fail fast” strategy of self-motivation), so I actually started. I planned the overall layout: music taste here, food likes over there, sports and activities on this side (sorted by individuality/team-orientedness, physicality etc.). I tried to make areas next to each other line up by having something in common, for example making physicality in sport line up with danceability in music, and having cerebral interests like science and art line up with music like classical and jazz.
Starting with food and music preferences made it clear right away that I needed varying levels of resolution. Maybe you just like “pop” or “science” or “meat” and are satisfied with large, low-def blobs showing this — or maybe you want to specify “eastern European bubblegum pop”, “biophysics” and “venison” because that’s specifically a big part of your identity somehow. It couldn’t just be a set of equally sized, present-or-absent dots but also show what areas of personality space that you personally had mapped out in high resolution.
At this point I began to realize that this wasn’t going to work. Not as a little side project, anyway. The task of building and organizing a library of options to choose from, that captures everything that anyone thinks is important about themselves, and keeping it up to date, is daunting enough to make the project unworkable — at least sufficiently unworkable that it clearly wasn’t the best option to spend effort on, among a host of other similarly ambitious ideas. I would still love to make something like this, but I’m not optimistic about being able to do it. I really envy monomaniacs who can spend their entire life on a single obsession.
- Perhaps if I’d been a trained, credentialed artist I could’ve called it art and been taken seriously. ↩︎
- Lost was and remains so controversial because it initially appealed to fans across this whole spectrum, but ended up satisfying only the “empathizer” standard of storytelling and giving the “systemizer” view a giant middle finger. ↩︎
- The more I hear about peer-review the worse it looks. I feel I should justify this more but for the moment I won’t. It’s just an impression I’ve picked up from reading a lot about how the scientific process works and hearing stories from friends in academia. ↩︎
- For example, blogger Alexey Guzey spent a year on a very critical review of claims in Matthew Walker’s Why We Sleep, and AFAIK they have not been addressed. I don’t know how correct they are, I haven’t read the book and I haven’t read all the cited research (obviously). It’s an excellent example of how essential 1) skill, 2) time, and 3) trustworthiness would be for a project like this. ↩︎
- Modern as dedicated to individual contemplation rather than communal worship. ↩︎
- Same with the classic Liero that I played to bits in high school. Once I had the tools I spent more time than I’d like to admit building new levels and designing new weapons. ↩︎
- Even more software like that exists now and I’m little sad I haven’t been in a position to be working on that kind of stuff. I see Townscaper and I’m mesmerized. I don’t actually play it (because I always prefer making to using), but just the fact that it exists makes me happy. ↩︎