Spend Writing Weirdness Points Well
N/A
[see ‘novelty’ tag] While reading a psychology methodology paper, I ran into the following comment:
Unfamiliar things are distrusted and hard to process, overly familiar things are boring, and the perfect object of beauty lies somewhere in between (Sluckin et al 1983). The familiar comes as standard equipment in every empirical paper: scientific report structure, well-known statistical techniques, established methods. In fact, the form of a research article is so standardized that it is in danger of becoming deathly dull.
So the burden is on the author to provide content and ideas that will knock the reader’s socks off—at least if the reader is one of the dozen or so potential reviewers in that sub-sub-specialty.
Besides the obvious connection to Jürgen Schmidhuber’s esthetics, it occurred to me that this has considerable relevance to LessWrong & Overcoming Bias (and life in general).
Robin Hanson in the past has counseled contrarians like us to pick our battles and conform in most ways, while not conforming in a few carefully chosen ones (eg. “Dear Young Eccentric”, “Against Free Thinkers”, “Even When Contrarians Win, They Lose”).
This strikes me as obviously correct, and that one could think of oneself as having a “idiosyncrasy budget” where non-conforming on both dress and language and ideas blows one’s credit with people / discredits oneself.
(See also “You have a set amount of ‘weirdness points’. Spend them wisely.”, “A Non-Conformist’s Guide to Success in a Conformist World”, Schank’s Law, “The economy of weirdness”.)
This idea about familiarity suggests a different way to think of it is in terms of novelty vs familiarity: ideas like existential risk are highly novel compared to regular politics or charities.
But if these ideas are highly novel, then they are likely “distrusted and hard to process” (which certainly describes well many people’s reaction to things on LW/OB), and any additional novelty like that of vocabulary or formatting or style, is more likely to damage reception or perhaps push readers past some critical limit than if applied to some standard familiar boring thing like evolution, where due to sufficient familiarity, idiosyncratic or novel aspects will not damage reception but instead improve reception.
Consider the different reactions to Nick Bostrom and Eliezer Yudkowsky, who write about many of the same exact ideas and problems—but no one put on off-Broadway plays or YouTube videos mocking Bostrom or accusing him in novels of being a sinister billionaire’s tool in a plot against all that is good and just—while on the other hand, Hofstadter’s GEB is dearly beloved for its diversity of novel forms and expressions, even if it’s all directed toward exposition on pretty standard unshocking topics like Gödel’s theorems or GOFAI.
This line of reasoning suggests a simple strategy for writing: the novelty of a story or essay’s content should be inverse to the novelty of its form.
If one has highly novel, perhaps even outright frightening ideas, about the true nature of the multiverse or the future of humanity, the format should be as standard and dry as possible. Conversely, if one is discussing settled science like genetics, one should spice it up with little parables, stories, unexpected topics and applications, etc.
One might also try to ‘buy’ weirdness points to spend by putting extra effort into polish & quality (trying for a perfection premium), avoiding risky countersignalling, and making an effort to affiliate with prestigious entities (eg. Bostrom has benefited enormously from his association with the University of Oxford).
Does this predict success of existing writings?
Well, let’s take Eliezer as an example, since he has a very particular style of writing. 3 of his longest fictions so far are the “Ultra Mega Crossover”, “Three Worlds Collide”, and MoR.
Keeping in mind that the former were targeted at OB readers and the last at a general audience on FanFiction.net, they seem to fit well:
the “Crossover” was confusing in format, on a low-status website (FanFiction.net) introduced many obscure characters or allusions, in service of a computationally-oriented Tegmark multiverse that only really made sense if you had already read Permutation City, and so is highly novel in form & content, so naturally no one ever mentions it or recommends it to other people;
“Three Worlds Collide” took a standard SF opera short-story style with stock archetypes like “the Captain”, and saved its novelty for its meta-ethical content and world-building, and accordingly, I see it linked and discussed both on LW and off;
MoR, as fanfiction, adapts a world wholesale, reducing its novelty considerably for millions of people, and inside this almost-“boring” framework introduces its audience to a panoply of cognitive biases, transhuman tropes like anti-deathism, existential risks, the scientific method, Bayesian-style informal reasoning, etc., moved to a slick custom website eventually, and MoR has been tremendously successful on and off LW. (I saw someone recommend it just yesterday on HN!)
Of course this is just 3 examples, but it does match the vibe I get reading why people dislike Eliezer or LW: they seem to have little trouble with his casual informal style when it’s being applied to topics like cognitive biases or evolution (where the topic is familiar to relatively large numbers of people), but then are horribly put off by the same style or novel forms when applied to obscurer topics like subjective Bayesianism (like the Bayesian Conspiracy short stories—actually, especially the Conspiracy-verse stories) or cryonics.
Of course, I suppose this could just reflect that more popular topics tend to be less controversial and what I’m actually noticing is people disliking marginal minority theories, but things like global warming are quite controversial and I suspect Eliezer blogging about global warming would not trigger the same reaction as to, say, his “you’re a bad parent if you don’t sign kids up for cryonics” post that a lot of people love to hate.
Speaking of Jürgen Schmidhuber, he serves as another good example: he spends weirdness points like they’re Venezuelan bolivars.
Despite him and his lab laying more of the groundwork for the deep learning revolution than perhaps anyone and being right about many things decades before everyone else, he is probably the single most disliked researcher in DL. Not only is he not unfathomably rich or in charge of a giant lab like DeepMind, he is the only researcher I know of who regularly gets articles in major media outlets written in large part about how he has alienated people: eg. 2016 or 2018. And this is solely because of his personal choices and conduct!
It’s difficult to think of an example of a technologist inventing so much important stuff and then missing out on the gains because of being so entirely unnecessarily unpleasant and hard to bear. (William Shockley & the Traitorous Eight come to mind as an example; maybe David Chaum & DigiCash too. But other than that…)
Have I seen this “golden mean” effect in my own writing?
I’m not sure. Unfortunately, my stuff seems to generally adopt a vaguely academic format or tone in proportion to how mainstream a topic is, and a great deal of traffic is driven by interest in the topic and not my work specifically; so for example, my Silk Road 1 page is not in any particularly boring format but interest in the topic is too high for that to matter either way. It’s certainly something for me to keep in mind, though, when I write about stranger topics.