🔬 physics

A formal theory of experimentation

This paper presents a formal theory that establishes the necessary and sufficient conditions a world must satisfy to permit the application of the scientific method.

Original authors: Eric Tesse

Published 2026-02-02

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Eric Tesse

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Building a Rulebook for Reality

Imagine you are an architect trying to design a building. Before you lay a single brick, you need to know the laws of physics: gravity must pull down, materials must hold weight, and space must exist.

Eric Tesse is doing the same thing, but for science itself. He isn't studying a specific experiment (like dropping a ball or measuring a star); he is trying to write the "Constitution" for what a universe must look like for the Scientific Method to even work.

His goal is to answer: What are the absolute minimum rules a world must follow so that we can run experiments, gather data, and learn things?

Part 1: The Stage and the Actors (Dynamic Systems)

To understand an experiment, you first need to understand the "stage" it happens on. Tesse calls this a Dynamic System.

The Metaphor: Think of a movie. The "system" is the movie itself. The "state" is a single frame of the film. The "parameter" is time.
The Rules: Tesse defines time not just as a clock, but as a mathematical structure that can be added up (like 1 hour + 1 hour = 2 hours). He creates a rigorous definition of a "path" through time.
The "Knowledge" Concept: This is a crucial idea. Tesse suggests that a system's current state is like a memory bank.
- If you look at a system's current state, it "knows" its past because that past is encoded in its current form (like a scar on your skin tells you where you were injured).
- However, the current state usually doesn't know the future. Just as a movie frame doesn't know what happens in the next scene, the present doesn't dictate the future. This explains why we can learn from history (the past is recorded) but struggle to predict the future (it hasn't been written yet).

Part 2: The Experiment as a Machine (The Shell)

Now, how do we formalize an experiment? Tesse introduces the idea of an Experimental Shell.

The Metaphor: Imagine a black box with a start button and a stop light.
- The Shell: The box contains both the thing you are testing (the system) and the machine testing it (the environment).
- The Start (Initial State): You press the button. The machine is ready.
- The Run: The system changes over time inside the box.
- The Stop (Final State): The machine turns on a light. The experiment is done.
The Rules of the Shell:
1. Re-runnable: You must be able to reset the box and run it again.
2. Clear Start/Stop: You can't be halfway through an experiment. It either hasn't started, or it has finished.
3. No "Ghost" Starts: You can't accidentally start the experiment twice before it finishes.
4. The "Knowing" Rule: When the light turns on (the final state), the machine must "know" that an experiment just happened. It can't just be in a random state; it must be a state that signifies "I just measured something."

Part 3: The Recorder (The Observer)

Inside the shell, we have the System (what we are studying) and the Environment (the measuring device). Tesse calls the environment a Recorder.

The Problem: Sometimes, a recorder might get confused. Maybe it records that a particle is "Spin Up," but later forgets that detail and only remembers "Spin Up or Down." This creates ambiguity. Did the particle actually have a specific spin (intrinsic uncertainty), or did we just lose the data (extrinsic uncertainty)?
The Solution: Ideal Recorders. Tesse defines a perfect recorder with two superpowers:
1. All-Retentive (All-reet): It never forgets. If it saw a detail at the start, that detail is still part of the final result. It's like a camera that takes a photo and never deletes the pixels, even if you zoom out later.
2. Boolean: It gives clear, distinct answers. It doesn't give you a blurry "maybe." It says "Yes" or "No," and these answers never overlap in a confusing way.

When a recorder is both "All-reet" and "Boolean," it is an Ideal Recorder. This is the gold standard. If a universe cannot support an Ideal Recorder, Tesse argues that the Scientific Method cannot function there because you can't systematically handle experimental errors or uncertainty.

Part 4: The Branching Tree (Ideal Partitions)

How do these experiments split reality into different outcomes?

The Metaphor: Imagine a tree.
- The trunk is the start of the experiment.
- The branches are the different possible outcomes.
- Tesse proves that for an Ideal Recorder, the tree doesn't just split randomly. It splits in a very specific, orderly way called Ideal Partitions.
- Think of it like a "Choose Your Own Adventure" book. At every page (moment in time), the story splits into distinct paths. An Ideal Partition ensures that once you take a path, you can't accidentally slip into another one, and the book knows exactly which path you are on.

Part 5: The Probability (The Odds)

Finally, Tesse looks at the math of chance.

The Rules: He shows that for a single experiment, the rules of probability are the standard ones you know (if you flip a coin, Heads + Tails = 100%).
The Twist: When you look at a collection of different experiments (like measuring position vs. measuring speed), the rules get tricky.
- In our everyday world, if you add up the odds of all possibilities, they equal 1.
- In quantum mechanics (which Tesse's theory covers), this isn't always true for groups of experiments. You can have a situation where the probability of "A or B" is not simply the sum of "A" plus "B."
The Construction: Tesse builds a mathematical structure called a Dynamic Probability Space. Think of this as a giant filing cabinet.
- Each drawer is a different experiment.
- The papers inside are the outcomes.
- He proves that even if the drawers don't perfectly line up (which happens in quantum physics), you can still organize the whole cabinet into a consistent system called a Simple Generalized Probability Space. This allows scientists to calculate odds even when the experiments are weird or don't fit together perfectly.

The Conclusion: What Makes a World "Scientific"?

Tesse concludes with a profound statement:

If a universe allows for Dynamic Systems (things that change), Ideal Recorders (machines that measure without forgetting or confusing), and Consistent Probabilities (ways to calculate odds), then that universe is scientifically comprehensible.

In other words, if you can build a machine that runs an experiment, remembers the result perfectly, and lets you calculate the odds of what happened, then that universe obeys the rules of science. If a universe cannot support these things, then the Scientific Method simply cannot exist there.

In short: Tesse didn't just describe how we do science; he built the mathematical blueprint for what a universe must look like to allow us to do science at all.

Technical Summary: A Formal Theory of Experimentation

Problem Statement

The paper addresses the need for a rigorous formal framework that defines the necessary and sufficient conditions a physical world must satisfy to admit the application of the scientific method. While scientific theories often describe experiments, they rarely formalize the structure of experimentation itself. The author seeks to develop a theory that is sufficiently general to encompass both classical and quantum systems, avoiding assumptions specific to any single physical theory (such as the requirement for state spaces to be manifolds). The core challenge is to formalize how dynamic systems change over time, how experiments record information about these changes, and how probabilities can be consistently assigned to experimental outcomes across different experimental setups.

Methodology

The paper employs a constructive, axiomatic approach rooted in set theory and the theory of dynamic systems. The methodology proceeds in four distinct stages:

Formalization of Parameters and Dynamics: The author defines "parameters" (time-like structures) via axioms governing ordering, additivity, and density, categorizing them into four classes isomorphic to standard number systems ( $\mathbb{N}, \mathbb{Z}, \mathbb{R}, \mathbb{R}^+$ ). Dynamic systems are then defined as sets of "paths" (functions from a parameter space to a state space). The paper introduces operations for slicing and concatenating paths, defining "dynamic spaces" as systems closed under concatenation, where the current state contains all necessary information to determine future paths.
Formalization of Experiments (Shells): Experiments are modeled as "shells," triples $(D, I, F)$ consisting of a dynamic space $D$ , a set of initial states $I$ , and a set of final states $F$ . The author imposes axioms to ensure experiments are reproducible, have a defined start and end, and that the final state "knows" the experiment occurred. This framework is refined into "environmental shells," which decompose the system into a "system" component and an "environment" (apparatus) component.
Refinement of Recorders: To address the ambiguity between intrinsic uncertainty (system dynamics) and extrinsic uncertainty (environmental state), the author introduces "recorders." These are refined through two properties:
- All-reet: The recorder never forgets information gathered about the system.
- Boolean: The recorder decides a partition of paths rather than just a covering, ensuring distinct outcomes are mutually exclusive.
  The intersection of these properties yields "ideal recorders," which provide a clean formalization of experimental outcomes.
Formalization of Probability: The paper extends the deterministic framework to probabilities by defining "Experimental Probability Spaces" (eps) and "Dynamic Probability Spaces" (dps). It introduces "Generalized Probability Spaces" (gps) and a construction $(X_\omega, T_\omega, P_\omega)$ to handle the algebraic structure of probabilities, particularly in cases where traditional additivity fails (as in quantum mechanics). The paper investigates conditions for "convergence" (countable additivity) within this generalized framework.

Key Contributions

1. The Concept of the Ideal Recorder

The paper's central contribution is the definition of the ideal recorder. By requiring an experimental setup to be "weakly unbiased" (the environment does not unduly influence system dynamics for undifferentiated paths), "all-reet" (no loss of information), and "Boolean" (outcomes form a partition), the author creates a formal object that perfectly captures the requirements of the scientific method.

Result: The paper proves that a covering of a dynamic set is decided by an ideal recorder if and only if it is an ideal partition.
Structure of Ideal Partitions: A major theorem establishes that ideal partitions are precisely those that unfold as "branching sequences of moment partitions." This means any valid experimental outcome can be decomposed into a sequence of instantaneous measurements (moments) that split the set of possible paths.

2. Formalization of Knowledge and Asymmetry

The paper formalizes "knowledge" as the set of paths consistent with a current state. It demonstrates a fundamental asymmetry: a system's current state encodes information about its past (what has happened) but not necessarily its future. This is formalized through the notation $S \to X$ (what is known about the past given state $X$ ) versus $S X \to$ (what is known about the future). The theory posits that experiments are reliable precisely because they record past events, aligning with the scientific method's reliance on recorded data.

3. Generalized Probability Spaces

The paper moves beyond standard Kolmogorov probability to accommodate systems where probabilities may not be additive over arbitrary unions (a feature of quantum mechanics).

Dynamic Probability Spaces (dps): Defined as collections of consistent experimental probability spaces.
Generalized Probability Spaces (gps): A broader structure $(X, T, P)$ where $X$ is a collection of partitions and $T$ is a collection of subsets.
The $(X_\omega, T_\omega, P_\omega)$ Construction: The author constructs a "simple" generalized probability space that is the unique extension of any initial probability assignment. This structure allows for the handling of non-additive probabilities while maintaining consistency across different experimental contexts.

4. Unbiased Conditions

The paper distinguishes between strongly unbiased conditions (where the environment never influences the system) and weakly unbiased conditions (where the environment may influence the system only after paths have been differentiated into separate outcomes). This distinction is crucial for modeling experiments where measurement choices depend on intermediate results (e.g., sequential measurements in quantum mechanics).

Results

Theorem 39 & 40: There is a one-to-one correspondence between ideal partitions and the outcomes decided by ideal recorders. Any ideal partition can be decided by an ideal recorder, and any ideal recorder decides an ideal partition.
Theorem 49: Ideal partitions are structurally equivalent to branching trees of moment partitions. This provides a geometric interpretation of how experimental outcomes evolve over time.
Theorem 61: If a generalized probability space is convergent (satisfies countable additivity), its extension $(X_\omega, T_\omega, P_\omega)$ is a simple convergent generalized probability space. This ensures that the mathematical machinery for handling infinite sequences of events remains robust even in non-standard probability contexts.
Consistency: The paper demonstrates that collections of experiments can be represented as a single dynamic probability space if and only if they are consistent (agree on overlapping outcomes).

Significance and Claims

The paper claims to provide a formal theory of experimentation that is independent of specific physical laws. Its significance lies in:

Defining the Scope of Science: It posits that a world is scientifically comprehensible if and only if it admits the construction of ideal recorders and the assignment of consistent probabilities. If a universe cannot satisfy these axioms, the scientific method cannot be applied within it.
Bridging Classical and Quantum: The framework is designed to be applicable to both classical and quantum probabilities. It does not assume quantum mechanics but provides the formal structure necessary to describe quantum-like phenomena (such as non-additive probabilities and contextuality) within a rigorous logical system.
Clarifying Experimental Error: The author notes that a universe allowing ideal recorders is one where experimental error can be handled systematically. The formalism separates the "system" from the "apparatus" in a way that preserves the identity of the system being probed while accounting for the recording process.

The paper remains modest in its physical claims, stating that while it formalizes the structure of experiments, it does not dictate the specific dynamics of any physical theory. It serves as a meta-theory, establishing the necessary conditions for any theory to be considered a valid scientific description of a world capable of experimentation.