Bayesian
An approach to reasoning that updates beliefs probabilistically based on evidence; Cowen urges being a 'strict Bayesian' rather than swayed by mood.
As defined in Patrick Collison
Where it comes up · 20
- 335 | Andrew Jaffe on Models, Probability, and the Universe13 claims
- 285 | Nate Silver on Prediction, Risk, and Rationality5 claims
- Tyler Cowen on Life and Fate2 claims
- Adam Frank2 claims
- David Bentley Hart1 claim
- Patrick Collison1 claim
- Philip E. Tetlock1 claim
- William MacAskill1 claim
and 12 more
Most important claims · 10 of 38Search all claims for “Bayesian” →
- A Bayesian and a frequentist can print the same '67 plus or minus 3' error bar but mean different things: the Bayesian states a credence about the actual value given the data, while the frequentist makes a statement about how often a procedure would bracket the truth over repeated experiments, so they are answering different questions even when the numbers coincide.335 | Andrew Jaffe on Models, Probability, and the Universe
- The main practical advantage of Bayesian methods over traditional statistics is the ability to marginalize over quantities you do not understand about your instrument or physical process, rather than bracketing your ignorance and pretending those quantities have a known value, allowing you to still learn what you care about.335 | Andrew Jaffe on Models, Probability, and the Universe
- The 1987 supernova neutrino detection was a one-off event that could not be coherently analyzed with frequentist statistics; Tom Loredo's Bayesian analysis showed the extra neutrino flux was very consistent with coming from the supernova and even implied one instrument had a timing error of a few seconds.335 | Andrew Jaffe on Models, Probability, and the Universe
- The slogan 'follow the science' is unscientific because it obscures the question 'which scientist?'—scientific consensus is a reasonable Bayesian prior but not infallible, and treating dissent as prima facie evidence of lunacy contradicts the scientific method; the proper response to an open question is a careful evidence-weighing deep dive, not demanding that society accept claims on faith because of scientific credentials.Megan McArdle on the Oedipus Trap
- The problem of induction has no deductive solution, but Bayesian inference is its answer: you start from meager prior information (itself a model), and systematically update with data to become more and more certain, though you can never reach 100% certainty.335 | Andrew Jaffe on Models, Probability, and the Universe
- A Bayesian approach to knowledge increases confidence that one is approaching truth without ever reaching certain truth statements, but Popperian epistemology argues Bayesian processes ultimately rest on an initial guess or conjecture forming the prior, so conjecture and refutation may be the more fundamental truth-seeking process—or alternatively, Bayesian updating could itself be construed as part of the critical testing demanded by conjecture and refutation.AMA | August 2025
- Bayesian inference is subjective only in the limited sense that different models or priors yield different answers; two people who approach the same problem with the same model and prior and the same data will get the same answer, and the Bayesian holds there is no meaningful notion of complete ignorance—every starting point is some model.335 | Andrew Jaffe on Models, Probability, and the Universe
- There is no answer to 'what is the probability that something will happen' in the abstract; there is only the probability given all the information you bring to bear, so all probabilities are conditional on a model, which is both the appealing and the troubling feature of probability.335 | Andrew Jaffe on Models, Probability, and the Universe
- Rather than hunkering down in philosophical positions (Bayesian vs frequentist) about whether one-time events can be assigned probabilities, the pragmatic approach is to run forecasting tournaments and empirically discover where the real limits of precision lie.Philip Tetlock on Superforecasting
- Moral philosophy is a 'train to crazy town': starting from commonsense intuitions, careful theorizing builds simpler, more explanatorily powerful theories that move us away from those intuitions, raising the methodological question of how far we are willing to go; MacAskill gets off the train before accepting Pascalian wagers about infinite value from a possible God.William MacAskill