Organizing Rational Discussion

Critical Rationalism (CR) says critical discussion is a major learning tool. But it doesn’t say how to have a critical discussion. What do you do, step by step? Brainstorm and criticize. That’s not much guidance.

CR has other broad tips like tolerance and a viewpoint on culture clash.

Discussion is a creative process. It’s not like following a recipe. But lots of help can be offered. My philosophy, Critical Fallibilism (CF), fills in this large gap with more info about discussion.

CR says we learn by critical thinking. Discussion and thinking use lots of the same methods. So the gap is not only about how to discuss but how to think. CR doesn’t say enough about how to organize your thinking to be effective.

Discussion and thinking both deal with ideas and the interactions between ideas, particularly disagreements (conflicts between ideas). Disagreements can be resolved based on the content of the ideas without considering the source. In other words, whether two ideas are both yours, or one is someone else’s, or they are both other people’s ideas, the disagreement can be resolved with the same arguments. That makes thinking and discussion similar. Also, when you discuss with others, they tell you their ideas. After you listen, versions of those ideas are in your mind, so then there are disagreeing ideas inside of just your mind. (By the way, there are ways to productively discuss the sources of ideas. They just aren’t necessary to resolve any particular disagreement. And there are lots of bad ways to discuss sources such as saying “Joe’s idea is probably bad because Joe is an idiot.”)

Put another way: discussion is just a tool, like reading, where you get some ideas to help your thinking. In discussion, people externalize and share some of their thoughts with you (and you do that for them). Books are a one-sided way of discussing: the author externalizes and shares thoughts but the readers don’t.

Thinking is more fundamental than discussion. But the term “discussion” better describes a process with multiple competing ideas which argue with each other. In this article I’ll be focusing specifically on discussion-like thinking – whether it’s alone (self-discussion) or talking with others – so I’ll call it discussion. Thinking is a broad category; there are other types of thinking too that are less like discussion.

CR Discussion

Next I’ll summarize how discussion works according to CR (which is pretty similar to the conventional view of discussion). It’s a broad, overall view that doesn’t give very specific discussion advice. I agree with it what it says, but I think more detail would be helpful.

Discussion-style thinking starts with one or more problems/goals/questions/purposes. And it happens in a context with background knowledge – people already know a bunch of stuff which they can use in the discussion.

One or more people brainstorm and propose solutions – ideas to solve a problem, achieve a goal, answer a question, or accomplish a purpose.

People debate the solutions. They come up with critical arguments. They try to explain why solutions will or won’t work.

Based on criticism, people modify solutions to try to improve them to no longer be refuted by that criticism. A criticism can point out a weakness of an idea which can then sometimes be fixed with a small change to the idea (or sometimes with a large change, and sometimes not at all). And sometimes people think of brand new solutions; discussion can help inspire new ideas.

CF Discussion

CF says to organize ideas with parent/child relationships. Some ideas are about other ideas. A child idea is about a parent. Children include footnotes, questions, positive arguments and negative arguments. Also, solutions are children of problems.

This organizes discussion in a tree structure. (Technically it’s a graph, but most discussions can be represented well by trees. Trees are simpler. One of the main differences is that trees only allow each idea to have one parent, whereas graphs allow multiple parents.)

The discussion tree is conventionally shown with the highest level ideas on top, and their children below them. The further down you go, the more concrete and detailed the ideas get. The higher up you look, the more ideas deal with abstractions and the big picture.

Basically, you can start with an idea that addresses the whole issue and write that down. The difficulty is that it’s complicated and hard to think about. So you break it into parts. The parts are its children. And then the children can be broken into parts. And you can keep breaking ideas down into smaller parts until things get simple enough to deal with. This helps you find errors, understand them better, and isolate them to specific areas of the tree.

You can read more about how CF uses trees.

CF also says to categorize ideas in certain ways.

Which problem(s) is this idea intended to solve? Which other ideas is this idea intended to support? To criticize?

All critical arguments should be evaluated as decisive or non-decisive. A decisive criticism means it contradicts what it criticizes – so if you accept the criticism (and don’t find an error in your background assumptions), then you must reject the target of criticism. A non-decisive criticism means you could accept both the criticism and idea being criticized, at the same time, because they’re actually compatible. A non-decisive criticism says an idea is bad but not necessarily wrong, whereas a decisive criticism argues that an idea must be wrong.

If an idea has multiple goals/purposes, then a criticism may apply to some goals and not others. An idea can be decisively refuted as a solution to one problem but still work as a solution to a different problem. When criticizing, It’s important to keep track of which goals the criticism applies to or not.

For supporting or positive arguments, if you aren’t taking shortcuts then CF says you should convert/translate them to negative arguments. (All good arguments can be easily converted. If you can’t convert it, the argument was bad. I discuss conversion and decisive criticism here.)

So you end up with four things in a discussion: problems, solutions, decisive negative arguments, and extra information which helps understand stuff.

The negative arguments can apply to each other (and to claims made in a problem or info), not just to the solutions.

An idea with one decisive criticism of it (that isn’t refuted) must be rejected. A decisive criticism is a reason an idea does not work.

Every idea should be judged as refuted or non-refuted (for each problem it’s meant to solve) based on whether it has a non-refuted decisive criticism of it.

People can also bring up categories of ideas rather than individual ideas. Categories are basically like an idea with a few parts left undefined. It’s using variables. E.g., instead of discussing “I want a fish” you could discuss “I want an X” and talk about how to get anything, and you could also limit X to a specific category like physical objects, food or coins.

This has just been a quick overview. Other CF articles have more detailed explanations about how to organize discussions, and have other ideas including Paths Forward and IGC (idea, goal, context) charts. My goal here was to explain the big picture concept about how I think CR is incomplete and CF has ideas to help.

Using decisive criticism and trees can help make discussions more effective. I’ll close by addressing a common objection. It can be a problem with any discussion, but it comes up more when trying to organize discussion with trees and decisive arguments.

Last Word Problem

It’s trivial to construct a decisive criticism of any idea if you don’t care about the quality of your criticism. In other words, you don’t care if someone can easily think of a counter argument.

In a chain of counter arguments, there’s a sense in which the last one wins everything. Suppose I say A and you counter with B. I counter B with C. You counter C with D. And then I counter D with E. That’s a chain of arguments because each argument connects to the next one, like a chain. And there are no branching points to make this a tree-shape instead of a line-shape.

Argument E refutes D, which means C has no refutation. So C refutes B, which means A has no refutation. Due to having the last argument, E, my original argument, A, is non-refuted.

If you make one new argument, it can change everything. You say F to refute E. Then there’s nothing refuting D, so D gets to refute C. That means there’s no outstanding (non-refuted) argument against B, so B refutes A.

With argument chains, the last word is powerful. Each new last word can flip the winner. This is especially true when using decisive arguments. And it’s a larger problem when trying to organize discussions in a tree (which involves chains plus branching points) rather than just throwing a bunch of unstructured organized out there.

So there’s a concern about bad faith actors who propose low quality ideas persistently and keep trying to get the last word (they’ll deny they’re doing that, and claim they’re making serious arguments that aim at high quality).

CF’s solution is to build up a library/archive of criticisms about categories of ideas. This preemptively addresses lots of bad arguments. It makes it harder to think of a last word that isn’t already addressed in advance.

Instead of saying “Hey, no fair, stop making dumb arguments.” or avoiding organized/structured discussion, or avoiding decisive arguments, CF says to learn how to easily, efficiently handle dumb arguments. If an argument is low quality, it shouldn’t be that hard to answer it. We just need a way to do that which doesn’t take too much time or work. And the basic solution to that is reuse arguments and to refute categories of ideas. Refuting categories lets one argument deal with many dumb ideas at once, instead of needing a different argument for every single dumb idea. The point is to criticize a certain type of idea, or ideas which fit a certain pattern. For example, ad hominem arguments are a bad category of argument. People have written down why all ad hominem arguments are bad, as a whole category, rather than trying to write a counter-argument to each individual ad hominem.

In short, write down refutations of all common errors. (And in the many cases where someone else already wrote down a refutation, you can use that instead of writing your own. You just have to take responsibility for what you use.) Then if someone makes an argument, there are two possibilities. Either you can reuse an existing refutation or else it isn’t a common error. If someone makes an argument and no one has ever written down why it’s wrong, then it’s actually worth considering. If it doesn’t fit any known pattern of bad ideas, then it’s something new and might be good.

When you first do this, people will come up with many last words you haven’t covered (either by writing arguments or by finding arguments other people have written). You should add (by finding and/or writing) more generic, powerful criticisms to your library which cover every type of bad argument they come up with.

Document what all the things are that you think are bad arguments, not one by one, but as large groups with shared bad traits. The more you do this, the harder it becomes for someone to come up with a new argument which survives pre-existing criticism from your library.

If someone can come up with a last word which isn’t refuted by your library of criticism, then either they have a good point about the topic or they found a hole in your library of criticism which you can improve. In both cases, it’s productive and worth considering. You get to learn something new about the topic or you get to improve your library. So this addresses the problem of endless “I think that’s wrong!” last words where people think of unhelpful, low-effort disagreements.