A number of popular arguments predicting seeming paradoxes rely on dangerous self-sampling assumptions. For example, the Doomsday argument predicts the end of humanity in the near future. Meanwhile, the Boltzmann brain model predicts the nonexistence of the reality we observe. The cosmic inflation model predicts an extremely low age for the observable universe, and so on. A common trend in this style of argument is the self-sampling assumption: “I should imagine myself as being a typical observer in my reference class of observers.”
But this assumption makes the direct claim that I’m “typical” - that is, to say, I didn’t “get lucky” by being on some extreme end of the bell curve. One only needs to look into the past to see how the Doomsday argument, for example, breaks down. Our ancestors, respectively, would have used the exact same line of reasoning to conclude our own nonexistence. The factor they forgot to account for is how extremely special a place they occupied in time - being among the first 1%, say, of humans to exist. Subjectively, they have no way of measuring how “lucky” they got.
It’s this missing luckiness factor that can logically defeat every possible self-sampling argument. In essence, it’s a question of how big to make our “reference class”. Is our reference class the class of observers that got just as lucky as us (e.g. “humans”)? Or is our reference class the class of all self-conscious observers? (Whatever that means.) Pinning down the precise “reference class” essentially requires deciding on how much “luck” we want to assign to ourselves.