Episode Overview

Podcast: New Books Network
Host: Greg McNiff
Guest: Andrew H. Jaffe, cosmologist and author
Book: The Random Universe: How Models and Probability Help Us Make Sense of the Cosmos (Yale UP, 2025)
Date: November 25, 2025

In this episode, Greg McNiff interviews Andrew Jaffe about his new book, The Random Universe, which explores how probability, models, and inference underpin our understanding of the cosmos. The discussion weaves together physics, philosophy, mathematics, and scientific method, providing a thoughtful yet accessible peek behind the curtain of modern cosmology and how scientists build and judge their models amid uncertainty and randomness.

Why Write The Random Universe? (03:49)

• Jaffe wanted to write for non-scientists to demystify how science—especially cosmology—works, focusing on “the tools of model building, the tools of probability and...interpersonal communication.”
• The title came as a revelation from a friend (“The Random Universe”), which quickly crystallized the book’s structure and themes.
• Target reader: “People who want to know not only about the science of cosmology...but also how we do cosmology and...science in general.” (04:41)

Scientific Theories: Models, Data, and Which Comes First (05:45 - 10:14)

The Eddington Quote

“Observation is not sufficient. We do not believe our eyes unless we are first convinced that what they appear to tell us is credible.” (05:45, Jaffe)

• Which comes first, data or theory?
  • We interpret the world through our internal models—from infancy, we develop models to make sensory data meaningful.
  • The raw data is not meaningful until fit into a model; in the history of astronomy, models evolved to fit beautiful simplicity or empirical accuracy, but ultimately both model and observation are intertwined.
  • “You can't interpret the world without the model. Right. It doesn't work the other way around. And that's the same thing in science.” (06:33, Jaffe)

Evolution of Cosmological Models

• From the geocentric models and cumbersome epicycles to Copernicus, Kepler, and eventually Newton’s elegant unification.
• Emphasis on how even “wrong” or superseded models (e.g., Ptolemy’s epicycles) can provide excellent fits to data.
• The beauty of models sometimes leads science astray, but also hints at deeper truths.

What Is Science? Induction, Inference, and Naturalism (09:52 - 16:46)

• Jaffe defines science as “the process of making inferences around the world from data.” (09:52)
• Inference: combining data with models/hypotheses to generalize or predict.
• The Problem of Induction (John Stuart Mill, Hume):
  • Science extrapolates from a finite set of observations, but logically, this extrapolation can’t be "proved."
  • “You can't prove that having seen the sun rise a million days...we should...see it again tomorrow. You need a model to prove that.” (12:25, Jaffe)
• Naturalism:
  • The assumption that the universe is, at bottom, intelligible and lawful—a precondition for science itself.
  • “If you believed that the world was unintelligible...then, you know, literally all bets are off. You therefore believe that anything could happen.” (14:17, Jaffe)

Mathematics and the Universe

• Is math the universe’s language?
  • Jaffe argues mathematics works because we design it to capture the regularities we observe, not because the universe ‘knows’ math.
  • Beautiful mathematical coincidences (like π turning up in varied contexts) can usually be traced to deeper connections, but are not necessarily magical or ontologically fundamental.
• “I think the universe is...just one damn thing after another. And...we are the ones who choose to describe that by differential equations and integrals...it really is a good description. But it’s a good description because...once you have things changing with time, then derivatives come in.” (20:29, Jaffe)

Beauty in Scientific Theories—and Its Risks (21:59 - 23:24)

• Simplicity and elegance often guide theoretical physics (Newton’s gravity, Einstein’s relativity), but beauty can be a double-edged sword.
• “You can...go down rabbit holes where you...pursue a beautiful theory that just isn't right.” (21:59, Jaffe)
• Debate exists over whether some modern theories (e.g. string theory) are pursued more for elegance than empirical adequacy.

Is There a Final Theory? (24:36 - 25:40)

• Jaffe is optimistic about a more complete theory uniting gravity and quantum mechanics but cautions it may not be as ‘unified’ or elegant as we imagine.

Probability: From Everyday Reasoning to Scientific Modeling (25:57 - 32:36)

Nature and Foundations of Probability

• Probability is about “how likely we think something happened,” not a property of things in the world (à la Bruno de Finetti: “Probability does not exist.”).
• “All probabilities are conditional.”
  • Probabilities depend on models and information: “The probability of heads is a half, given my model for your flipping the coin.” (31:30, Jaffe)
  • Changing the model or learning new information revises probabilities—a core Bayesian principle.

Principle of Indifference

• When no reason to favor one outcome, assign equal probability (e.g., 50:50 for a fair coin, 1:52 for a random card).

Frequentist vs. Bayesian Probability (33:10 - 39:49)

• Frequentists: Probability as the long-run frequency in repeated experiments (idealized infinity).
• Bayesians: Probability as a subjective (but principled) assignment given state of knowledge and model (“a model for learning”).
• In cosmology and many real-world cases, Bayesian reasoning is more transparent and flexible—especially when experiments can’t be repeated infinitely.
• “You start out with some prior information...It transforms your prior into your posterior...it’s a model for learning.” (39:53, Jaffe)
• Bayes' theorem is the mathematical foundation for updating beliefs with new evidence.

Bayesian Answer to Hume’s Problem of Induction

• It “dissolves” rather than “solves” the problem: “You can’t derive what we ought to think about the world from the way the world actually is. You have to put other information in.” (42:08, Jaffe)
• Bayesian inference formalizes this: beliefs are updated in light of evidence, but never with absolute logical certainty.

The Messy Realities of Scientific Practice (49:44 - 55:52)

The Scientific Method: Popper, Lakatos, Feyerabend

• Popper: Theories are falsified, not proven.
• Lakatos: Scientific programs, nested ‘rings’ of theory and practice, distinguish ‘progressive’ from ‘degenerating’ programs.
• Danger of degenerating programs: theories that survive by endless ad hoc fixes rather than being meaningfully tested.

Models Sometimes Grow by Adding “Dark” Ingredients

• In cosmology, challenges to the model (e.g., dark matter, dark energy) push scientists to continually re-examine and sometimes patch the framework—a risk but also a necessity in a probabilistic universe.

Entropy, Information, and Conditional Probability (55:52 - 62:39)

• Entropy: Originally a measure of “disorder” and inaccessible energy in thermodynamics, generalized by Boltzmann and Shannon (in information theory) to quantify randomness or uncertainty in any system.
• Conditional probabilities underpin all these formulations: entropy increases as our uncertainty (given what we know) increases.
• Shannon’s work: provides mathematical description of information transmission under probabilistic uncertainty.

On Boltzmann’s Equation

• Jaffe calls it “remarkable” for relating the microscopic states of a system to macroscopic entropy without detailed knowledge of atoms, connecting statistical randomness to practical outcomes.

Quantum Mechanics: Copenhagen, Many Worlds, and Probability (64:22 - 73:12)

Copenhagen Interpretation

• Quantum mechanics fundamentally predicts probabilities, not certainties; the act of measurement collapses possible states to a specific outcome.
• In practice, most physicists simply “shut up and calculate,” regardless of philosophical discomfort.

Many Worlds Interpretation

“Theoretically parsimonious but ontologically extravagant.” (64:51, Jaffe)

• Parsimonious: The wave function evolves deterministically, without “collapse.”

• Ontologically extravagant: All possible outcomes are realized in some branch; the universe “splits” with every quantum event.

• For Jaffe, all interpretations struggle with how to understand probability—he favors viewing probability as conditional on our knowledge and models, including which “branch” we might find ourselves in.

Cosmology and the Random Universe (73:12 – 86:36)

Sakharov Conditions

• Explains why the universe contains more matter than antimatter.
• Three conditions: (1) Some law must favor matter over antimatter, (2) processes must avoid reverting back to equilibrium, and (3) the universe must experience periods out of equilibrium (e.g., through expansion and cooling).
• Observing the preponderance of matter provides a rare chance for cosmology to inform fundamental particle physics.

Mapping the Cosmic Microwave Background (CMB)

• Jaffe recounts the progression from COBE to WMAP to Planck satellites and balloon experiments (Boomerang, MAXIMA) mapping the CMB’s tiny irregularities—“the baby picture of the universe.”
• Each successive project improved precision and resolution, confirming the Big Bang and models like inflation, providing more constraints and insights into cosmological structure and history.

The Next Frontier: Simons Observatory

• The Simons Observatory, with vastly more detectors and sensitivity, aims to detect subtle polarization signals (including potential gravitational waves from inflation) in the CMB.
• This may offer radical new insights into the very early universe and the unification of quantum mechanics, gravity, and cosmology.

Memorable Quotes & Timestamps

• On the necessity of models:
  “You can't interpret the world without the model. Right. It doesn't work the other way around. And that's the same thing in science.” (06:33, Jaffe)

• On induction and naturalism:
  “If you believed that the world was unintelligible...then, you know, literally all bets are off. You therefore believe that anything could happen.” (14:17, Jaffe)

• On mathematics and the universe:
  “I think the universe is...just one damn thing after another. And...we are the ones who choose to describe that by differential equations and integrals...” (20:29, Jaffe)

• On Bayesian inference:
  “It’s a model for learning. You start out with some prior information...It transforms your prior into your posterior...that in that sense, it's a model for learning.” (39:53, Jaffe)

• On the many worlds interpretation:
  “Theoretically parsimonious and ontologically extravagant.” (64:51, Jaffe)

• On conservation of humility in science:
  “People have done this forever...maybe things like string theory get a lot of press because they seem like, oh, it's a lot of effort on this one thing and maybe it's not right because it seems so beautiful. But this isn't new and it's not. And they're not being foolish. They might be wrong.” (22:46, Jaffe)

Key Takeaways

• All scientific reasoning is inherently probabilistic and conditional—always based on the best (but imperfect) models and data available.
• Mathematical structures in science are immensely powerful, but are models—tools we invented to capture empirical regularities, not intrinsic features of reality.
• Probabilities are not “out there” in nature but reflect our state of knowledge and the structure of our models—a thoroughly Bayesian perspective.
• Scientific progress involves not just finding better data, but developing deeper, more fruitful models for explaining and predicting it—accompanied by a necessary humility about the persistent mess of uncertainty and possibility.
• The quest to unify the apparently random (quantum) and regular (cosmic) aspects of the universe continues, with future discoveries likely to challenge and revise our current models.

For listeners seeking an accessible yet rigorous exploration of the scientific method—as it's genuinely practiced—and the probabilistic foundations of cosmology, this conversation with Andrew Jaffe is an invaluable guide.