• The best way to predict the future is to [combine multiple forecasting methods].

• Be the [role model of rationality] you want to see in the world.

• Believe you can and you’re [<1% of the] way there [, but if you don’t believe you can, then you’ve already lost].

• Change your thoughts and you change [your experience of] the world.

• Don’t live the same year 75 times and call it a life [unless you’ve already optimized it to be an awesome year].

• It wasn’t raining when Noah built the ark [but God had pretty much explained to him what was going to happen – you’ll need to be shrewder since you won’t get that handy advance warning before bad things happen].

• The successful warrior is the average [person], with laser-like focus [and a training regimen involving deliberate practice with tight feedback loops].

• Early to bed and early to rise [helps make] a [person] healthy [,productive], and wise, [except with evening chronotypes, where the opposite applies].

This piece was first written on November 17, 2020, and first appeared on my website on September 15, 2026.

Here are a bunch of examples:

(1) Appeal to authority: some authorities ARE reliable. For example, ask a physicist about a well-established idea in physics, and you’ll likely get the right answer.

(2) Bandwagon effect: lots of people believing something often is evidence for it being true. For example, lots of people saying they heard a loud bang means it probably happened, even if for some reason you didn’t hear it.

(3) Ad hominem: sometimes a person’s character being bad SHOULD make us doubt the evidence they provide. For example, if a sadistic person with a history of harming people is giving evidence that you should trust them, you probably shouldn’t.

(4) Slippery slope: sometimes making one change DOES make further changes more likely. For example, if you let your kids stay up late and they have a great time, they might push harder for it next time.

(5) Gambler’s fallacy: events often really are correlated (negatively or positively), not statistically independent. For example, if you are driving a rental car and the car has one problem, it may signal negligence at the rental company, making other issues more likely. On the other hand, if you are drawing marbles from a bag that contains both red and blue marbles, every blue drawn makes the chance that the next is red higher.

(6) Affirming the consequent [if A, then B. B occurs. Therefore you assume A]: if a cause commonly leads to a certain result, then seeing that result DOES provide some evidence of that cause. For example, if older age causes slower walking, then someone walking slowly IS evidence of older age.

(7) Correlation does not imply causation: it’s true that very often things are correlated without being caused by each other, but correlation is still EVIDENCE of causation. For example, if nearly every day you drink coffee you have trouble sleeping, it doesn’t guarantee coffee is the cause of your sleep problems, but it IS evidence.

(8) Ecological fallacy: sometimes a group has such unusual characteristics that group membership tells you A LOT about one of that person’s attributes. For example, knowing someone is a professional runner tells you they can almost certainly outrun you.

(9) Cherry-picking: sometimes one anecdotal (but real) example is so extreme that it provides a lot of evidence. For example, a story of how someone cheated a former business partner is a lot more important than five examples where they didn’t cheat a partner.

Note that, despite the exceptions I give above, people DO often use invalid arguments. This is extremely common, in fact. But it’s worth being aware that apparent “logical fallacies” are sometimes valid (or at least, not fully invalid) forms of reasoning in real-world situations.

In fact, I think part of the reason why people use these logical structures in an INVALID way is precisely that they actually are sometimes valid – if they were never valid, it would be a lot easier to tell they are types of bad thinking, and so would probably be used less.

For each logical fallacy, it benefits you to know not just WHAT it is, but WHEN it is invalid—knowing that REALLY CAN help you be a better thinker and reasoner.

Some might say, “Well, by definition, it’s only a fallacy if it’s used fallaciously.” For example, “it’s only the appeal to authority fallacy if it’s an invalid appeal to authority.” Otherwise it’s a valid appeal to authority, not a fallacy. But my problem with that is that while some uses are clearly valid, and others clearly invalid, there’s a large range in between, where an appeal to authority provides SOME evidence. The question is, how much? Or, put mathematically, how much more likely is this claim to be true, given that you’ve just found out this particular authority supports the claim, than if you hadn’t learned about that authority’s support of the claim?

At a deeper level, I think these considerations expose some of the weaknesses of “logical deduction” as a thinking tool.

Take an example where a person wearing a white lab coat makes a scientific claim, and people are more likely to believe them because of the lab coat.

The fact that someone in a white lab coat is presenting scientific information PROBABLY IS evidence that the position is correct, in a technical sense. Sure, anyone can put on a white lab coat, but on average white lab coats are worn by scientists, and on average scientists are more often right about science topics than non-scientists are. I’m not saying it’s strong evidence in that case (charlatans could adopt white lab coats to trick people, and they do), but it technically is evidence in the sense that:

Probability(True given White Lab Coat) > Probability(True given Lack of White Lab Coat)

I think that probabilistic thinking is closer to the right model for thinking about the world than deductive logical thinking is. Or, viewed another way, logic is a degenerate case of probability theory where all your probabilities are 1. If there is probability of 1 that Plato is a man, and a probability of 1 that all men are mortal, then deductive logic and probability theory BOTH tell you that Plato is definitely mortal. But deduction starts to fall on its face when non-negligible uncertainty is introduced. In contrast, probability theory handles uncertainty gracefully, while still handling the cases (or almost all of them) that deduction does as well.

So what I think is happening with a lot of the formal fallacies is that in a deductive framework they are perfectly correct (If A implies B, then that truly doesn’t mean that B implies A, deductively speaking), but this is misleading in a probabilistic framework (where, if A implies B, then observing B is, in fact, evidence for A)!

This piece was first written on July 4, 2020, and first appeared on my website on October 16, 2026.