You're never wrong injecting complexity, but rarely you're right

post by MrMind · 2018-10-03T14:20:15.177Z · score: 38 (15 votes) · LW · GW · 6 comments

I just want to put this idea in the form of a post, to gather your impressions. I think it's my main rationalist failure mode.

Recently in a Facebook group, some poster has proposed this synthesis of Harari's book '21 lessons':

In the 21st century, three narratives were used to explain the past and predict the future: the fascist, the communist and the liberal narrative. During the century, the latter has prevailed, although in recent times it has started to crack, due to events like the election of Trump, the Brexit, and so on.

Then, the same user asked: what could be a new narrative that would help us in the future?

I was tempted to reply as I always do: criticize the simplification. I was about to write that the concept of narrative itself is a narrative, that Harari is seeing the past with the eyes of the present, but not necessarily this lens will help with navigating the future, that also a better concept would be that of a memeplex, which is less internally coherent than a story, and thus more complex to pinpoint.

Then a reflection occured to me: I always end up doing this, in almost all discussions I partecipate. People simplify too much and come to the wrong conclusion, they consider only the extremes of a spectrum, they use words as rigid classifiers and debate endlessly about them, they do not have internally coherent point of views, etc. I almost invariable end up 'winning' (i.e. appear wise) by injecting some complexity: usually in the form of a new parameter that was buried in the presuppositions.

Then, I was struck by another insight: it is too easy to win this way. From a mathematical point of view, a system with a bigger state space is more flexible than a smaller system, and so an 'optimization' that increases the space state is always correct. Is that possible? How probable it is that I've discovered a universal optimization of every human debate? I reasoned that it's very low, and indeed I think I've always failed to consider the downside: a more complex system is more difficult to use. When I add a parameter to the problem space, I'm multiplying the costs to use said idea in practice.

A classic example: was Martin Luther King a criminal? Some argue from the letter of the law, some argue from a moral point of view. If I 'win' the debate by injecting the concept of 'moral inertia' (speed at which a legal system reacts to an evolving moral landscape), then I implicitly add a parameter to every closed case, that could be potentially re-examined.

So I've decided I will refrain to add complexity, to formally but sterilely win debates, unless I have also a good 'reduction', that forgets some complexity in favor of being more useful. Doing so however is not trivial, it adds real work: this is due to the fallacy of gray, the fact that there are shades of gray doesn't mean that perfect middle gray is the correct answer.

6 comments

Comments sorted by top scores.

comment by Lou Nowell (lou-nowell) · 2018-10-03T15:01:36.129Z · score: 4 (5 votes) · LW · GW

From a purely pragmatic perspective; conversation is a tool for getting what you want. There exists a spectrum from simplicity to complexity; and favouring one over the other is really just a case of the grass seeming greener on the other side of the fence, imo.

There are situations where it pays to simplify your point, and situations where it pays to get deep and dirty. I wouldn't hesitate to guess which type of situation is more commonly encountered, so it seems to me that there's no readily applicable heuristic for this problem - you just have to take each conversation as it comes.

comment by habryka (habryka4) · 2018-10-03T18:28:22.561Z · score: 9 (5 votes) · LW · GW

Is this comment intentionally trying to do the exact thing the OP was talking about for comedic effect? Or is it just an accident? I think it's actually a pretty decent comment if it tried to do this intentionally.

comment by Pattern · 2018-10-04T02:33:47.158Z · score: 2 (2 votes) · LW · GW

I understand what the OP is saying - adding complexity to the model helps win arguments. On the other hand I might also accept the strong statement 'people tend to 'oversimplify' things', or the weak statement 'people do not always overcomplicate things', and from the second statement it follows that adding complexity isn't always a bad thing, and so whether or not it's right might very well be a judgement call. (This is my understanding of Lou's comment.)

Sometimes people start conversation around weird ideas whose merits (aside from being interesting to talk about) aren't clear, and you might not too many of your conversations hijacked by memes that you find to be weird premises... On the other hand, it's also polite/sometimes useful to interact with other people's model/s.

comment by Hazard · 2018-10-04T01:06:39.359Z · score: 2 (1 votes) · LW · GW

More simple might help.

comment by Elo · 2018-10-03T20:21:52.450Z · score: 2 (1 votes) · LW · GW

Steelman already exists.

If you find a situation more simple than it is, you need to evaluate the outcome for all cases.

If you find adding a variable makes better outcomes, re-evaluate each closed case.

But whichever you do, address the core of the argument, not the edges.

For example - "narrative", it's a simplification. A future narrative might be taking this information into account when carving itself.

comment by romeostevensit · 2018-10-15T02:05:44.358Z · score: 1 (1 votes) · LW · GW

Consider open and closed mode for inspiration (one of if not the best talk I've ever seen):

https://www.youtube.com/watch?v=Pb5oIIPO62g