Is there any ideal way to navigate the semantic shape of any idea, interaction, dialogue, hypothesis or question? What do I mean by semantic shape, let’s take an example question I was thinking about yesterday

  1. “what function could resentment serve?” — before I could answer this, I was caught up in the reeds of the previous semantic layer that is either taken as axiomatic or to be established.
  2. “emotions serve a function” — this started off as axiomatic given Antonio Damasio’s work talking about the equilibrium of the body and how an emotion is a departure from the equilibrium and intended to cause action that moves the body back towards that equilibrium. But it is not self-evident if even this can be taken as axiomatic given that I haven’t actually read those books. At no point are these debates about whether these claims are objectively axiomatic, they are simply an exploration into whether they are axiomatic to me, given what I already know, because I discover that almost all axioms I take for granted are very thinly papering over the cracks of imperfect understanding. So we hit the next semantic level now —
  3. “only that which has function can exist”

At this point I notice that each succeeding semantic layer is that much more intellectually rewarding to navigate. So if the semantic layers are graphed on the Y axis, and their level of intellectual stimulation as width on the X axis, is there a coherent shape that semantic layers create?

Starting with semantic layers going deeper (down the Y axis). I think of really obscure analytical philosophers like Wittgenstein or Russell, it would appear that the interesting-ness increases with depth of semantic layer. There is a huge payoff when un-learning assumptions that have been taken for granted and revealing their inaccuracy, and the payoff is bigger the Then follows that anything farther downstream of original question would then see reduction in interesting-ness linearly in that direction?

Not so, farther downstream also seems quite interesting, an intense zoom-in that reveals subtleties and textures to what seemed a boring surface. So then it’s a big X shape, with increasing interesting above and below surface? But not really, the immediate semantic layer of original question also tends to get everyone in the room groaning and eye-rolling and calling someone pedantic. Like Bill Clinton answering ‘depends on what the meaning of the word ‘is’ is’. It’s weak because it is easy, a knee-jerk reaction akin to ‘your question is stupid’, the game theory angle is I have two options when posed with a question, I answer it or I ask another question. If I’m smart enough to come up with an answer that is intellectually stimulating, then I will settle on that, if not I will save myself the effort and just ask another question.

The really obnoxious versions of this response are just turning the tables and saying ‘you tell me first’, which is so transparently bad that I’m happy to report it never even arises as a legitimate option. The less obnoxious is to add the smallest bit of value and go down 1 semantic layer, asking about the nature of the question itself. It’s still easier than the effort needed to come up with an intellectually stimulating answer to the original question, hence it gets eye-rolls and accusations of being pedantic. But beyond this semantic layer, the deeper ones still hold large payoffs, and it is impossible to get to them without first navigating this one. Further, even this layer can often be objectively interesting, so the best strategy seems to be to ignore the instinctive eye-roll and explore the layer while being excited for the layers to come thanks to this one.

This problem is mirrored at the immediate semantic layer above the original question. The instinctive eye-roll here is because of frustration that people immediately dive down into specifics instead of being capable of abstracting an issue to a higher level of analysis for a more pure intellectual exercise. This is often the layer at which I bail, being biased towards higher levels of abstraction as if abstraction were the ultimate ideal. But that’s only true if the idea stalls at this immediate +ve layer instead of breaking through it to the more interesting deeper +ve layers. A good analogy is a fantasy-writer engaging in detailed world-building to create a more 3-dimensional universe. Attention to detail can have a huge intellectual payoff, and this is something Robin Hanson’s blog does really well with the simplest ideas. Downstream impacts that are separated by more than 1–2 degrees are as abstract as upstream semantic layers.

So that gives us a base question that tapers a little at the immediate semantic layer below and above, before expanding in payoff at both ends continuously, the higher the order of the level, the bigger the payoff. Before we close out the shape of semantic orders, one last questions is whether the width of the base question (interesting value) matter? If everything is relative?

More concretely, does the initial payoff of the question impact the relationship between payoffs at higher orders? Theoretically yes. This question, for example, is novel, and therefore past a certain threshold for payoff. But at a higher level of analysis, one that separates what is novel from what is actually interesting, must point to the number of connections that the base question has, and therefore the number of different spin-offs and potential for semantic layers that it contains. Connection-poor questions, though novel, will be much less interesting. These could be questions that could easily be verified, like what is the currency of Botswana. Questions with a larger number of variables, and ideally unanswerable variables, have larger number of connections and therefore not only have a larger payoff (width), but also spawn the most number of higher orders and the higher orders with the largest payoffs.

Back to original of shape, it is diverging at both ends. What is the most intellectually rewarding way to navigate these levels, given the original question. One answer taken for granted it, it has to be in ascending steps, so no deeper/higher semantic level before the preceding step is done. But what about direction? Up or down? Jump up and down? This sounds like a Jeopardy problem, where I either go horizontally across categories but with same payoff, or vertically with same category and increasing payoffs before coming back to next category and dropping the payoff suddenly. It seems like the vertical strategy is inferior, given there is a risk of boredom and exit when I experience the steep fall in payoff as I come back to the first +ve semantic layer after the final -ve one. But it’s also the most natural, requiring the least resources to rip the mind away from a problem it is grappling with.

This is the strategy I tried last night, because it was path of least resistance. A really thorough follow-up to this exercise will be to navigate the horizontal strategy next time and compare, having normalized for initial question payoff.

A novel insightful exercise to determine the pragmatic difference in intellectual payoff between a novel insight and an obvious fact mistaken for novel insight.