I’m in the middle of editing my large and messy bundle of notes on The Elephant in the Brain by Robin Hanson and Kevin Simler into a review of sorts, and one observation grew into something large enough to carve out and make into its own thing.
I noticed a strange effect when reading it. I’ve read plenty of both Hanson’s and Simler’s other work and almost nothing in Elephant was new to me, but I still felt I was learning something. Ideas from the corners of my mind’s eye were pulled together and put front and center, and this had a particular effect on the structure of my mind.
What happened was this: while it was all familiar I’d only encountered it in bits and pieces before, and having the parts brought together in a single volume improved the structural integrity of my worldview. It made the whole edifice sturdier. The partial overlapping with ideas on many topics fixed the system in place, like how papier-maché becomes strong and rigid by layering paper strips partly on top of each other.
Crafting your idea-maché
Any book, article or text communicates one or a closely related set of ideas, and as you read more and more your internal knowledge base will begin to resemble such partly overlapping pieces holding each other in place, like papier-maché. Or a patchwork of exotically shaped lego blocks.
I started to think you could categorize reading materials based on how they interacted with your preexisting “idea-maché”. Elephant was (for me) a patch covering a weak spot where several separate bits were sort of touching each other.
Here’s a picture:
The addition of the orange bit doesn’t add much surface area, but it does make the whole thing stronger and more cohesive.
By interacting with my earlier ideas in this way, this book, to me, right now, had a consolidating function.
Right. Here comes the fun bit with creating a model: exploring its implications. What other functions are there?
I did immediately start thinking of a phenomenon my voracious reading of pop-science and pop-philosophy books has made me familiar (and frustrated) with: after a few entry-level books on economics, psychology, consciousness, evolution, cognitive biases, The Universe, mathematics, the history of philosophy etc. you start coming across the same ideas over and over again.
At that point the books or articles tell you nothing new and does nothing to your idea-maché except adding another layer to something already thick enough.
This is duplicating function.
Reading duplicating material isn’t completely worthless, but it’s like watching yet another episode of an ok sitcom. Relaxing and mildly entertaining, but it doesn’t do anything new for you. It doesn’t make you a different person. It’s a part that could be left on the cutting room floor without The Story of Your Life losing anything at all.
Reading for new insights or personal development means to read things that either reinforce the wobbly parts of your idea-maché, Elephant-style, or extend it outwards.
This, extending function, elaborates on and adds more to something you already know. It’s like reclaiming land from the sea.
Strictly speaking almost everything you read has extending function — there’s always something there you already know even if only very basic facts, and always something new even if tiny and inconsequential. But in this context I mean by extending, text that has significant overlap with what you know and at the same times adds substantial new material.
The proportion of overlap can vary, of course. While a book about the french revolution almost certainly would’ve clearly extending function to me because I have a decent if basic understanding of it, one about (for example) the Taiping rebellion which I know hardly anything about besides that it was a thing, would be better thought of in a different way. I’d have almost nothing except basic stuff like dates and geography to anchor the new information, so it’d be less like reclaiming land and more like establishing an offshore colony.
How about incepting function for this? Exactly when extending becomes incepting isn’t well defined and doesn’t have to be. Just consider the functions ideal types. If I for some reason were to pick up a book about professional dog grooming in the Netherlands in the 1970’s it could reasonably be described as “opening up a new world” to me. That’s what incepting function is.
It goes without saying that looking for incepting (or just highly extending) material is a high-risk strategy. It’ll be topics you’re not already interested in so you won’t know if you could be, and you won’t know how to separate the wheat from the chaff either. If you go for something philosophical or theoretical it’s far from certain you’ll understand it at all, both because of unfamiliar ideas and alien unstated assumptions.
Tools to help us navigate such “uncharted waters” would be helpful, and complementary to existing recommendation systems of the “people who bought X also bought Y” variety that tends to be quite conservative and give us extending (at best) and duplicating (at worst) suggestions.
A recommendation engine for reading that could reliably supply enjoyable inceptive experiences would be a marvel. Free research project idea everyone!
Making connections, finding patterns
The possibility of incepting functions suggest that you can have several different “islands of knowledge” — an archipelago of idea-machés rather than one single structure, which is what you’d have if you stuck to strictly extending (and duplicating) learning your entire life. We don’t do it that way. In school, art class isn’t taught as an extension of mathematics or vice versa. Nor does pop culture trivia, gardening and sports history have anything to do with each other in your mind. This is the norm, I trust. Weirdos like me, obsessed with unification, integration and the Big Picture to such a degree that they would read (let alone write) an article like this one, are probably rare. Archipelagos of idea-machés (ideapelagos?) are standard.
This suggests yet another possible function. If Elephant had a consolidating function on me by solidifying a loose area, there should also be books with connective function that builds bridges between two or more previously unrelated islands.
Good writing with connective function is hard to pull off, because it requires cross-disciplinary expertise. I wrote about that in relation to another Robin Hanson book, The Age of EM which likely has connective function for most with some knowledge of computing, brains and economics. Hanson uses the intersection of the three to speculate about the far future, and in the process creates links and interfaces between them.
One of my favorite examples is Curry: A Tale of Cooks and Conquerors that my wife had me read a few years ago. Not something I would have ever thought to pick out myself, it was fascinating in that it brought together two things I knew a little but not a lot about, namely Indian cuisine and the history of British colonial rule in India. The first is apparently to large extent a result of the second.
There’s a sliding scale between connective and consolidating functions, just like there is between duplicating and extending, and extending and incepting. When there is significant overlap with existing knowledge, incepting becomes extending, and when that overlap approaches 100%, we end up with duplicating. Those functions work by adding information to an existing structure, not by finding patterns across structures.
That’s instead what the connective and consolidating functions do. Connective means making new connections (and is therefore the “intertopical” version of inceptive), while consolidating means reinforcing flimsy connections between related subclusters (the intertopical version of extending).
Here are the functions in a nice two-dimensional model:
Something’s obvious now: we’re missing one.
What corresponds to making connections but with very high overlap? It would be something light on new information per se, but heavy on models and abstraction. It would make you see things in a new way by giving you new mental tools and schemas.
Hopefully, that describes this article. It definitely describes one of my favorite websites, Ribbonfarm. Ribbonfarm’s tagline is “Experiments in refactored perception”, where the word “refactored” refers to a software development practice of rearranging the structure of your code to make it clearer and make new functions possible. To refactor your perception, then, means to acquire new concepts with which to interpret your experience. In line with that I want to call the last element of this six-part model refactoring function.
Using the type of illustration from above, refactoring would come out something like this:
The limitations of using a 2D representation of a multidimensional idea space become apparent now. It might be better to think of refactoring not as a snake slithering around haphazardly in your knowledge structure, but as a process changing the shape of it by making unexpected connections and building cross-topical concepts. The space itself is warped and stretched; distant ideas now next to each other as if by wormhole, and counterintuitive, circuitous paths straightened into high-traffic autobahns. For those familiar with the phrase “insight porn“: this is the stuff in its purest form.
Coda: Functions of fiction
While the model is based on non-fiction books and articles, I don’t see any reason why it couldn’t be extended to fiction. Fiction construct a system of ideas as well, a system very different from those representing the material world, yes, but still a reality of sorts. Its building blocks are themes, tropes, emotions, perspectives, interpretations, personal journeys and destinies, relationships and hypotheticals. An idea-maché built from such pieces can be created, extended or transformed just as well.
Fiction with incepting function would be something exposing you to new tropes and themes. If you only read romance novels, diving into high fantasy or cosmic horror would count.
Extending function fiction would use familiar story elements and explore from there. If your cup of tea is run-of-the-mill murder mysteries, a mystery heavy on psychological themes, social criticism or magical realism would be extending for you.
Reading duplicative stuff is perhaps what we most often do: tackle yet another Tom Clancy high-stakes global politics brick, pulp SF story or brightly colored standard portion of chick-lit. While fine for relaxation, duplicative reading material risks becoming masturbatory and meaningless if it doesn’t leave your mind as something it wasn’t before. It’s the cultural equivalent of “empty calories”.
Connective fiction would bring together disparate tropes and themes. Historical dramas with SF elements is likely to be connective for most. A Song of Ice and Fire was likely many people’s first encounter with political intrigue combined with dragons and ice-zombies. The book Handling the Undead explores how a bureaucratic state deals with the sudden and unexpected reanimation of thousands of recently deceased. Et cetera.
When thinking about what consolidating fiction would be, what comes to mind is Infinite Jest that I reviewed almost a year ago. It takes many vaguely similar themes, like single minded pursuit, addiction, wireheading, personal isolation, short-sightedness, need for belonging, commitment and depression, and turns them into a tapestry depicting motivation as a hunt for rewards: its nature, function and most importantly, its dysfunction when not existing in a context of healthy personal relationships.
For fiction to be refactoring, it’d have to deal with familiar events, themes and tropes but organized so we see them in a new way. I think this is what’s called “literary fiction” (or avant-garde art in general) most often tries to do. It takes known, understandable emotions and processes and twists things around to make us look at our lives in a different way. James Joyces Ulysses is a typical example of taking mundane events and showing them through a strange filter, changing our understanding of them. Even when literary fiction does deal with weird, non-mundane events those events are often best understood as symbolic of the ordinary (e.g. the works of Franz Kafka).
Maybe. My knowledge of literary fiction is weak and scattered. I need some good consolidating material on it.
• • •
I’d call it “pensectomy” (thought removal) if it didn’t sound too much like “penectomy” (penis removal).
I’m being a little unfair here. You can become an expert by reading a lot on the same topic if each iteration adds more detail to previously covered terrain. This isn’t well captured by this model, which makes sense considering my own attitude towards knowledge. I’m biased towards dilettanteism and prefer getting a low-fidelity big picture view rather than a detailed representation of a small area. But it takes all sorts and I don’t forget that for someone like me to get the gist, it requires that a specialist spends the necessary time and effort working out the details first. A superior cousin of duplicating could be called “refining” or “developing” function.
In theory nothing would be purely incepting — even if I were to read a book about the economic history of the hngQ’rp civilization of Phraghalon IV it’d connect to something I already knew, if only the fact than the hngQ’rpians lived on a planet and were subject to the normal rules of chemistry, energy, metabolism and the mechanics of biological evolution.
This is perhaps another way to describe the tendency or non-tendency to compartmentalize.
Another example is perennial nerd-favorite Gödel, Escher, Bach which brings together mathematics, visual art and baroque music using the theme of self-reference. However, note that GEB is particularly prone to having different functions depending on what you bring to it. It was connective to me when I read it in 2009, but if I’d read it at this age it’d more likely be consolidating. For genius mathematician-composer-artists it might be refactoring or possibly even duplicating.
It’s worth reiterating something about this model: A text can have more than one function. This isn’t a system of mutually exclusive categories where we sort texts into buckets. It’s a tool for thinking about reading and help you figure out what you want to read. Text function is dependent on the relationship between a readers preexisting knowledge structure and the text. It isn’t a feature of the text itself (although of course, some texts are more likely to have certain kinds of functions).
One brand of particularly insight-pornographic refactoring is reduction (or compression, see Schmidhuber), where sections of the “ideapelago” are explained in terms of simple underlying principles — in the most extreme case principles that if memorized allow the whole body of knowledge to be reconstructed (the clearest example is math: you don’t need to work as hard to memorize formulas perfectly if you learn how to derive them). Such refactoring strengthens the construct as a whole by adding deep structural reinforcements. It also frees up more mental resources by reducing the load of keeping everything in your head.