Modular Argument

Even in supposedly serious and noble-minded web conversation, an awful lot of it seems to add little to understanding.  Yes, at times, even at Overcoming Bias.  I have long wondered what the core problem is, and how to make it clear to these authors.  Here is my best shot: good arguments try to be modular

When I moved from physics to software long ago, I learned the overwhelming importance of modularity; you must write complex software in a way that lets you test and change one small part with minimal attention to other parts.  When I became an academic I learned that academia gains its power similarly.  Academics specialize, and write so related specialists can understand each contribution with a minimal understanding of other contributions.

A good arguer considers what his audience knows, and what sort of evidence or analysis they find persuasive.  A good arguer may have some big final conclusions he wants his audience to reach, but he will usually not try to argue for them in one step.   Instead, he breaks his argument into small modular parts, each of which is pretty likely to convince most of an open-minded audience of one small new conclusion in a relatively short time.  He takes care to explain what he means in terms they understand, and to summarize his new small conclusion so that it can be ready to build on in future arguments. 

Of course sometimes he will over or under estimate such tasks, his audience will find flaws in his argument, or he will probe to see what future arguments may require.   And sometimes the reward will not seem worth the effort. 

In useless arguments, people often just state strong claims, daring others to prove them wrong.  Such a position taker often does not even bother to make his claim clear, daring others to prove that nothing he could mean could be true.  Such bad arguers are more often unwilling to respond to questions asking them to clarify their claim, or to outline what argument path could support their claim.

Of course not every audience is worth the effort to craft an argument to convince them, and there can be a point to just making clear to some audience that there are people who disagree with them.  But if you bother to talk a lot to some audience not yet convinced of your views, yet you do not pursue a path of modular arguments to convince them, ask yourself: just why are you talking to them?

GD Star Rating
loading...
Tagged as:
Trackback URL:
  • http://profile.typekey.com/sentience/ Eliezer Yudkowsky

    I dearly hope that the length limit and fast feedback here at Overcoming Bias will finally teach me to be modular.

  • http://www.econ.canterbury.ac.nz/eric Eric Crampton

    Have you been arguing with the LaRouchists on campus again, Robin?

  • michael vassar

    Robin, I’m skeptical that this is the only approach to academia.

    It actually seems to me that much of the Continental/Analytical divide in philosophy can be understood in terms of a preference for holistic understanding vs. reductionistic modeling/description. The latter sounds very similar to modular description.

    As someone who is generally sympathetic to the Analytic side in general, it seems to me, at first, that this argument makes a lot of sense. Since libertarian inclining social thinkers tend towards the analytical and reductionistic in style, it even explains my early libertarian tendencies and those of most people I get along well with. However, it seems to me that a broader historical view would tell us that the modular/holistic divide is also present in the distinction between science and math. In this case, the observational side of the experimental method is the relatively non-modular analytical style, and it seems to me that math and natural philosophy inspired by math existed for millennia before the addition of careful observation brought it’s efficacy to a new level.

    My general conclusion is that relatively analytical arguments do work better in those cases where the problem chosen is one where they work at all, but often they only take you so far. The best thinkers usually pursue a variety of analytical pursuits when they are young and their high fluid g allows them to develop expertise, and then, when they are older use the expertise so gained to enable broader more synthetic insights.

    I’m also not convinced that typical people are more convinced by the sort of argument you are saying a good arguer makes than by the sort of argument a bad arguer makes. They are certainly much less likely to claim to be convinced and then demonstrate when challenged that they don’t understand your position, but my observations suggest that they are also less likely to claim to be convinced period. How many people have been convinced by analytical economic arguments from comparative advantage? How many by holistic ‘economic’ arguments about exploitation or alienation?

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Michael, most academic experiments and the papers that describe them are very modular. I agree many “bad” arguments convince many people, but they don’t do so for good reasons.

    Even in programming one comes across hard “knots” of problems that are harder to decompose into problems. If one doesn’t want to avoid these problems, one must pay the much higher price of a less modular approach.

  • http://profile.typekey.com/halfinney/ Hal Finney

    I have two points. First, these considerations apply even if your goal in offering an argument is to receive criticism rather than to persuade your listeners. In either case, presenting the argument in terms they can understand is a prerequisite.

    Second, one issue I find in arguments here is that many of them refer to specialized knowledge that not everyone may be familiar with. We often reference results from evolutionary psychology, philosophy, or economics that may be rather esoteric and technical. This presents a burden to wider understanding and appreciation of the issues.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Hal, the larger your organized library of previous modules, the easier it is to write any given new system, by reusing that library. So we with specialized knowledge face a choice when posting, whether to make use of that library. We face a tradeoff between the size of our audience, and how easily we can talk.

  • http://profile.typekey.com/halfinney/ Hal Finney

    Something like that happens with software libraries too. Sometimes I’ll download some interesting-sounding software, only to find that it depends on a library I don’t have; I go to get that library, and discover that it depends on yet another library I have to find. There’s even a name for this: dependency hell. There is often a tradeoff between ease of software development and the size of the audience that will be able to download and use the software.

  • TGGP

    Holism just plain rubs me the wrong way and has for as long as I can remember. I don’t know if that’s a bias or not, but I always just get the impression from holistic arguments that they aren’t terribly concerned about the underlying details (and hence, in my view, reality) and prefer their intuition. I don’t know if that has anything to do with my libertarianism, but I think holistic libertarian arguments can certainly be made, primarily of the Hayek/Burke type that we shouldn’t fiddle with some aspect of society because we don’t know how it works, although I think one could also make the same point with a less holistic approach.

  • michael vassar

    Robin: That’s why I talked about the observational sciences. Also, standards differ. Compared to analytic philosophy or math, the sciences are VERY holistic. All sorts of unexamined assumptions and vague implicit knowledge. Very expertise driven.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Hal, it would be very interesting to explore the analogy between dependency hell in software and who talks to who about what in ideas.

  • http://www.optimizelife.com Gustavo Lacerda

    This thread wouldn’t be complete without a reference to argument mapping: http://en.wikipedia.org/wiki/Argument_map