Galilean Frame Uncertainty Has No First-Order Pudding-Theoretic Residual

Authors: Sterling Geisel, QBist Lab

Abstract

Kelly and Giamou formulate uncertainty on the special Galilean group, SGal(3), rather than on a loose product of time, pose, and velocity coordinates. Their paper is a useful boundary case for Pudding Theory. The source domain is classical estimation. Its observer is an inertial reference frame, not a conscious field. Its uncertainty is covariance on a Lie group, not selection among stochastic physical outcomes. This Working Paper applies Chaos Susceptibility only as an exclusion criterion. A nonchaotic coordinate model should not admit a first-order Pudding-theoretic correction. The prediction is therefore null. Once SGal(3) covariance is used, observer-indexed augmentation should not improve calibrated estimation. A reproducible improvement would not confirm ordinary Galilean geometry. It would falsify the present boundary application and require a new physical coupling term outside the source paper.

Source Synopsis

Kelly and Giamou study uncertainty in classical Galilean spacetime. The paper is not about quantum measurement, consciousness, or anomalous probability. It is about representing uncertain time, position, orientation, and inertial velocity on the manifold where these variables already interact.

The mathematical object is the special Galilean group, SGal(3). This ten-dimensional Lie group combines spatial rotation, spatial translation, time translation, and Galilean boosts. In Galilean relativity, time intervals are absolute across inertial frames, while spatial positions and velocities transform between frames. The paper describes Galilean spacetime as a fibre-bundle structure in which each instant of absolute time has an associated Euclidean spatial fibre. That description fixes the physical setting. The observer here is a coordinate frame.

The technical claim is that uncertainty should be carried by the group. If time is uncertain, position uncertainty changes under velocity. If velocity is uncertain, temporal and spatial measurements are coupled. Treating time as an external scalar breaks this structure and can give inconsistent covariance propagation. The authors provide the group representation, Lie algebra, exponential and logarithmic maps, adjoint action, and a compact Jacobian for uncertainty propagation.

The demonstration is an estimation problem for a time-varying transformation between inertial frames using noisy observations collected at distinct times. Because noise appears in all measured variables, the paper uses a total least squares formulation. The reported result is statistical consistency. Estimators built on SGal(3) respect the geometry and perform better than formulations that separate time from the motion group.

The conclusion is modest and technical. Classical space, time, and inertial motion should be estimated on the coupled Galilean manifold. The paper corrects a covariance model. It does not introduce a new physical source of outcome bias.

Postulate Lens

This Working Paper applies Chaos Susceptibility as a boundary criterion.

The fit is negative but concrete. The postulate states that systems with positive Lyapunov exponents amplify small coherent inputs into macroscopic statistical shifts. The source paper does not study such a system. SGal(3) is a symmetry group for classical inertial frames. Its estimation residuals arise from sensor noise, finite samples, and model geometry. They are not produced by a chaotic substrate waiting to amplify an informational bias.

This matters because the word “observer” can mislead. In Kelly and Giamou, an observer is a reference frame. In Pudding Theory, an observer relevant to probability bias must be a physical information-bearing system coupled to a susceptible target. Those are not the same object. A frame transformation has no internal instability, no positive Lyapunov response, and no stochastic barrier-crossing channel.

The postulate therefore licenses silence at first order. It says that a Pudding-theoretic correction should not be added merely because a paper contains uncertainty. The correction is admissible only when the target dynamics can amplify a small coherent input. SGal(3) covariance is not such a target. It is the ordinary geometry that should be used before any anomaly is claimed.

Pudding Theory Prediction

The prediction is a preregistered null boundary. For the class of estimation problems studied by Kelly and Giamou, statistically consistent uncertainty propagation should be exhausted by SGal(3) geometry, measurement noise, and estimator design. No observer-indexed parameter should survive out-of-sample testing once those ingredients are modeled correctly.

Let \(M_0\) be a baseline estimator that treats time outside the motion group. Let \(M_G\) be the SGal(3) estimator. Let \(M_P\) be an augmented estimator that adds observer identity or observer-state terms to \(M_G\). Pudding Theory predicts that \(M_G\) will outperform \(M_0\), because \(M_0\) has the wrong geometry. It also predicts that \(M_P\) will not outperform \(M_G\) after model complexity is penalized.

The appropriate comparison is not raw error alone. It is calibration. The normalized estimation error squared, NEES, should lie within the expected confidence bounds for the correct covariance model. If \(M_P\) lowers raw error by becoming overconfident, it has not found a physical residual. It has damaged uncertainty calibration. The relevant test must therefore require both improved out-of-sample NEES and preserved calibration.

A clean protocol follows from the source paper. Collect noisy observations of frame transformations at distinct times. Randomize and blind observer labels so that no estimator can infer labels from sensor geometry, sampling time, or trial order. Fit \(M_0\), \(M_G\), and \(M_P\). Use a fixed complexity penalty, such as WAIC or held-out log predictive density with nested cross-validation. Require replication in at least three independent laboratories.

The expected result is simple. \(M_0\) should be less consistent. \(M_G\) should be calibrated. \(M_P\) should collapse to \(M_G\), with observer coefficients indistinguishable from zero. Any apparent observer effect should vanish under blinded relabeling. If it does not, the finding belongs to a new experiment, not to ordinary Galilean covariance.

Falsifiable Observable

The distinguishing observable is the out-of-sample NEES for SGal(3)-based frame estimation under blinded observer assignment, with calibration assessed by the appropriate chi-square confidence interval and model complexity corrected by nested cross-validation. If the observer-indexed reduction in out-of-sample NEES were measured to be at least 10 percent relative to the unaugmented SGal(3) estimator, with calibration preserved, \(p<0.001\), and replication in at least three independent laboratories under unchanged sensor noise, this Postulate would be falsified.

Editorial Dialogue

Tanaka: You are using Pudding Theory to predict no Pudding-theoretic effect. That is thin. Would ordinary model comparison reach the same conclusion?

Sterling: Mostly yes. That is the point. The source paper is a control case. It shows a place where ordinary geometry should do the work.

Tanaka: Then why write it as a Lab paper?

Sterling: Because a theory needs exclusion rules. Chaos Susceptibility says a target must have instability that can amplify a coherent input. SGal(3) covariance is not such a target.

Tanaka: You are making the postulate act as a gatekeeper, not a generator.

Sterling: Correct. Gatekeeping is a scientific function. Without it, every covariance residual becomes a temptation.

Tanaka: Suppose the observer-indexed model improves NEES under blinding.

Sterling: Then this boundary application fails. The result would imply a missing coupling between observer state and classical estimation residuals. That would be new physics for this domain.

Tanaka: And until then?

Sterling: Until then, the source paper should be read as a geometry paper. Its observer is a frame. Its uncertainty is a Lie-group covariance. No extra term is justified.

Discussion

The limitation is source dependence. This Working Paper does not derive a new SGal(3) result. It uses Kelly and Giamou as a boundary test for when Pudding Theory should remain absent from a model. That is a narrower claim than a positive application.

The open empirical question is residual structure. If SGal(3)-based estimators leave reproducible residuals correlated with blinded observer identity, then the null boundary is wrong. The evidence would need to survive sensor controls, time-order randomization, laboratory replication, and complexity penalties. A one-laboratory gain or an uncalibrated error reduction would not be enough.

The present conclusion would change only under that kind of residual. Analogy will not change it. The words observer, time, and uncertainty are insufficient. The relevant test is whether a susceptible physical target exists and whether an observer-indexed term improves calibrated prediction. In the source domain, the answer should be no.

References

1. Jonathan Kelly and Matthew Giamou. “Uncertainty in Space, Time, and Motion on the Special Galilean Group.” arXiv:2602.13327, 2026.

2. Sterling Geisel. “Pudding Theory: A Topological Theory of Information Fields.” QBist Lab foundational paper, September 10, 2025.

3. Vladimir I. Arnold. Mathematical Methods of Classical Mechanics. Springer-Verlag, 2nd edition, 1989. doi:doi:10.1007/978-1-4757-1693-1.

4. Ray E. Artz. “Classical Mechanics in Galilean Space-Time.” Foundations of Physics 11, 679-697, 1981. doi:doi:10.1007/BF00726944.

5. Timothy D. Barfoot and Paul T. Furgale. “Associating Uncertainty with Three-Dimensional Poses for Use in Estimation Problems.” IEEE Transactions on Robotics 30, 679-693, 2014. doi:doi:10.1109/TRO.2014.2298059.

6. Ajit Bhand and Andrew D. Lewis. “Rigid Body Mechanics in Galilean Spacetimes.” Journal of Mathematical Physics 46, 102902, 2005. doi:doi:10.1063/1.2060547.

7. Jean-Marc Lévy-Leblond. “Galilei Group and Galilean Invariance.” In Group Theory and its Applications, Volume II, Academic Press, 1971. doi:doi:10.1016/B978-0-12-455152-7.50011-2.

8. Joan Solà, Jeremie Deray, and Dinesh Atchuthan. “A Micro Lie Theory for State Estimation in Robotics.” arXiv:1812.01537, 2021.