Imagine a sheet of paper where the bottom line — the conclusion — is written first, and then arguments are filled in above it. No matter how compelling those arguments look, they carry no evidential weight because they were selected to support a predetermined conclusion. This thought experiment from the Sequences illustrates why the order of operations in reasoning matters: evidence must be evaluated before reaching a conclusion, not gathered in service of one. Once you have written the bottom line, any further "reasoning" is rationalization. The practical test: if no possible evidence could change your conclusion, you are not reasoning — you are performing reasoning.
Before evaluating an argument, ask yourself: have I already decided what I think? Practice with political or social topics where you have strong priors. Try to specify in advance what evidence would change your mind — if you cannot, you may have already written the bottom line.
From motivated reasoning, you know that desires and identity can steer reasoning toward predetermined conclusions without the reasoner's awareness. The Bottom Line, a thought experiment from Eliezer Yudkowsky's Sequences, crystallizes this insight into a single vivid image that serves as a diagnostic for broken reasoning.
Imagine a sheet of paper with a conclusion written at the bottom -- the "bottom line." Above it, someone has filled in arguments, evidence, and reasoning that support the conclusion. The arguments may be individually valid. The evidence may be genuine. But none of it carries evidential weight, because it was selected specifically to support a conclusion that was already determined before the reasoning began. The order of operations is reversed: instead of evidence leading to a conclusion, the conclusion determined which evidence would be gathered. Any argument selected to support a fixed conclusion tells you nothing about whether the conclusion is true -- it tells you only that the person is skilled at finding supporting arguments, which is a different thing entirely.
The practical test for whether you have written the bottom line is deceptively simple: what evidence would change your mind? If you can specify, even roughly, what the world would have to look like for you to abandon your current conclusion, your belief is held in a way that can respond to reality. You have a prior, not a bottom line. But if you find that no conceivable evidence would change your conclusion -- or if every piece of contrary evidence gets reinterpreted as either flawed or secretly confirming -- you have written the bottom line, and all the reasoning above it is rationalization. The key distinction is between having a strong prior (legitimate -- Bayesian reasoning starts with priors and updates) and having a fixed conclusion (illegitimate -- the "updating" is performative).
Consider two people who both conclude that a controversial policy is correct. Alex decided the policy was correct first, then gathered five supporting arguments and ignored three counterarguments. Beth examined all eight arguments with genuine openness and concluded the policy was correct. The arguments Beth found are evidence about the policy; the arguments Alex found are evidence about Alex's ability to construct rationalizations. The output looks identical -- five arguments for the same conclusion -- but the epistemic value is completely different. This is why the order of operations matters so much in reasoning: the same argument can carry genuine evidential weight when encountered honestly and zero weight when cherry-picked. Detecting which process generated the arguments requires examining how they were found, not what they say -- which is exactly why motivated reasoning is so hard to catch from the outside and nearly impossible to catch from the inside without deliberate practices like specifying what would change your mind.
Topics in reflective domains aren't scored by quiz answers. Read, reflect, and mark when you've thought it through.