Arguments in parallel vs arguments in series
When arguments are delivered in parallel, their combined strength is no less than the strength of the strongest argument. For example, you might encounter two different people arguing in favor of the same position, one giving an argument that you can easily see doesn’t work, and one giving an argument that’s harder to refute. Or somebody might give five examples of a phenomenon they say exists, two of which are stretches but three of which work out all right. In both cases, the situation should have at least as much “persuasion power” as it would if the worse arguments did not exist.
When arguments are delivered in series, their combined strength is no greater than the strength of the weakest link. Somebody might begin with premise A, use it to draw intermediate conclusion B, and then use that to draw final conclusion C. If the link from B to C is suspect, then the entire argument is suspect, even if the link from A to B is irrefutable.
I’m guessing most people who will read this post already understand what I’m saying, but I’ve never seen it presented with the “in parallel” and “in series” phrasing before. I thought it was cute, and maybe useful, so I decided to post it. As usual, be careful not to apply this principle selectively—your own in-series arguments need complete inspection to be accepted, and others’ in-parallel arguments need complete inspection to be rejected.
Caveat: When I was getting feedback on this post, somebody pointed out that the presence of two bad examples actually can weaken the case made for the theory if your reasoning process involves evaluating how competent the person presenting it is. If they were 3 for 3 on giving examples, you might trust them more and therefore give more credence to their claims, whereas if they were only 3 for 5, you might trust them less and therefore give less credence. I agree with this. I think the reason it doesn’t quite work out perfectly is that, if you’re trying to evaluate the competence of the other person, then the five examples they give aren’t actually quite parallel arguments, since they also contribute to the shared pool of “how trustworthy is this person?”.
But really, the important point about parallel arguments is that you can’t pick one of the two bad examples, argue that it’s bad, and then ride the momentum from that rejection into a rejection of the whole theory. By the same token, you can’t reject a position on the grounds that you thought of a reason why somebody might believe it, and the reason was obviously stupid. The problem with doing that is, that reason exists in parallel alongside an infinitude of other possible reasons to believe the position, and arguments in parallel can’t be rejected collectively just because you found a weak argument among them.
This is a recipe for Gish Gallops. I’ve seen lists of 100 arguments against evolution. By your reasoning, I would not be able to say “that person opposes evolution for poor reasons” unless I’ve addressed all 100 arguments.
Or to put it another way, treating parallel arguments as strong creates incentives that lead people to act in ways (producing Gish Gallops) that produce weaker parallel arguments.
While I agree that the original post does invite Gish Gallops, I think it’s still reasonable to say “that person opposes evolution for poor reasons” if you’ve checked the first two of their hundred reasons and they’re both terrible. Even if there was a fantastically good argument against evolution in that list somewhere, that person is still endorsing many poor reasons and appears unable to distinguish the good from the bad.
It is true that you can’t reasonably say “none of these are good arguments” until you’ve investigated all of them, but you can reasonably argue that you have strong enough evidence that this person is so bad at evaluating arguments that it’s not worth your time to investigate the rest.
Unfortunately, this is sub-optimal from a game theory perspective. It results in information hiding, and can result in arbitrarily-incorrect views as a result.
Suppose that everyone knows that you stop reading after the first two wrong arguments you see.
Now suppose that there are two common and reasonably well-regarded arguments that every knowledgeable person on the subject knows, and as well each person knows two novel arguments of their own devising that they know are likely incorrect.
The optimal for each person trying to convince someone of the premise, knowing only that the person stops reading after two wrong arguments, is to give them the two common arguments and nothing else.
So you see the same two common arguments repeated a thousand times, instead of the thousand novel arguments. And you miss that ten of the novel arguments are actually correct.
(This is very akin to strategic voting in some ways.)
(This still “works” even if, say, 50% of the population has heard the common arguments before...)
An agent wanting to search for truth with bounded resources is better off looking somewhere that isn’t a known cesspit of bad evidence presented as if it were good.
That aside, I still don’t think your hypothetical world captures the essence of the proposal. The Gish Gallop is primarily an asymmetric resource exhaustion strategy. Evaluating the first(*) couple of arguments to judge the quality of the evidence is a defense against that strategy. If they’re only presenting their own two arguments then this isn’t a Gish Gallop, and the defense is inapplicable.
It seems to me that a better strategy would be to first quickly check that they do in fact already know the standard not-obviously-wrong arguments or other evidence, and then proceed with your own information. If they don’t already know it, then presenting the well known information first is a good thing. If they do know it, then you now both know that you’re working from a common base before starting on the new information of unknown value. In both cases your situation is stronger.
(*) They need not be the first strictly, but shouldn’t be cherry-picked. Reliably choosing the first few makes it obvious that they’re not cherry-picked, and establishes a useful norm of communicating strongest information first.
Right. The 100 arguments the person gives aren’t 100 parallel arguments in favor of them having good reasons to believe evolution is false, for exactly the reason you give. So my reasoning doesn’t stop you from concluding that they have no good reason to disbelieve.
And, they are still 100 arguments in parallel that evolution is false, and my reasoning in the post correctly implies that you can’t read five of them, see that they aren’t good arguments, and conclude that evolution is true. (That conclusion requires a good argument in favor of evolution, not a bad one against it.)
But that isn’t the conclusion that I’m trying to make.
That one is the conclusion that I’m trying to make.
But if those don’t count as 100 parallel arguments because each bad argument speaks to the guy’s incompetence at evaluating arguments, well… someone’s bad argument always speaks to his incompetence at evaluating arguments. Every single list of arguments is like that. So the exception then swallows the rule and there is no such thing as a true list of parallel arguments.
In strict logical terms you are correct. In practice this is leaving the door open to someone addressing the weakest argument only and declaring your entire post invalid as a result. (There is a fallacy for this; I cannot remember the name.)
Yeah. I should have made it clear that this post is prescribing a way to evaluate incoming arguments, rather than describing how outgoing arguments will be received by your audience.