The Causal Uncertainty Principle: Manifold Tearing and the Topological Limits of Counterfactual Interventions
Authors: Rui Wu, Hong Xie, Yongjun Li
Organizations: School of Management, University of Science and Technology of China 96 Jinzhai Road, Hefei, 230026, Anhui, China · School of Computer Science and Engineering, University of Science and Technology of China 96 Jinzhai Road, Hefei, 230026, Anhui, China
Judea Pearl's do-calculus provides a foundation for causal inference, but its translation to continuous generative models remains fraught with geometric challenges. We establish the fundamental limits of such interventions. We define the Counterfactual Event Horizon and prove the Manifold Tearing Theorem: deterministic flows inevitably develop finite-time singularities under extreme interventions. We establish the Causal Uncertainty Principle for the trade-off between intervention extremity and identity preservation. Finally, we introduce Geometry-Aware Causal Flow (GACF), a scalable algorithm that utilizes a topological radar to bypass manifold tearing, validated on high-dimensional scRNA-seq data.
Figures & tables
Figure 1: Conceptual Overview of the Topological Limits of Counterfactual Interventions. (Left: The Deterministic Failure): Attempting to transport the factual measure to an extreme out-of-distribution target across a geometric void. To minimize transport cost (preserving identity), characteristic curves inherently intersect, inducing a finite-time singularity (Manifold Tearing). (Right: The Entropic Necessity): The injection of geometric entropy (via Schrödinger Bridges / SDEs) allows the probability mass to fluidly bypass the void. However, this enforces the Causal Uncertainty Principle: topological validity requires irreversible identity smearing.
Figure 2: Comprehensive Verification of Topological Tearing and Scaling Laws. (Top Left): Deterministic ODE flows force characteristic curves to violently cross within the geometric void. (Top Center): The precise, smooth collapse of the cumulative Jacobian determinant det(Jt)→0 , providing rigorous mathematical proof of the loss of diffeomorphism. (Top Right): Calibrated GACF effectively bypasses the singularity via adaptive entropic tunneling. (Bottom): Empirical validation of the tc∝1/D scaling law. The observed singularity times (blue dots) strictly track the theoretical O(1/D) Riccati bound (gray dashed line), confirming the determinable boundary of the Counterfactual Event Horizon.
Figure 3: Curvature Effects and the Causal Uncertainty Pareto Front. (Left): The evolution of the Jacobian determinant under different Riemannian geometries via strict Riccati integration. Positive curvature accelerates manifold tearing, while negative curvature extends the survival window tc . (Right): The mathematically honest Pareto front evaluated on a non-linear topological canyon. The ODE falls short of full survival. The standard SDE survives but severely smears individual identity (variance 0.893 ). GACF optimally bounds the Causal Uncertainty Principle, achieving a strictly superior balance by reducing identity loss by 58.4% (variance 0.372 ) while guaranteeing topological validity ( tc=1.0 ).
Figure 4: Support Sensitivity and Geometric Singularity. Empirical validation of the critical topological entropy ε∗ required to prevent manifold tearing across varying factual support diameters Δ . As the support broadens towards the critical geometric threshold ( Δ≈2.67 ), the required entropy diverges, validating the singularity in the Causal Uncertainty Principle. Beyond this threshold (Region I), finite entropy cannot stabilize the deterministic flow.
Figure 5: High-Dimensional Scalability and Universal Singularity in Neural Flows. (Left/Exp A): The collapse time tc exhibits a catastrophic non-linear decay as the latent dimension n scales to 100, confirming that the curse of dimensionality violently accelerates topological tearing. (Right/Exp B): Dual-axis tracking of a neural flow ( n=100 ). The theoretical Jacobian determinant (red) collapses smoothly, yielding a true singularity at t=0.345 . The Hutchinson-estimated scalar divergence (blue) amplifies the topological risk via a sharp Riccati blow-up. Using the dynamically scaled threshold ( λthresh=−10.0 ), the radar successfully triggers at t=0.010 , providing a massive Δt=0.335 lead time to safely inject entropy before catastrophic failure.
Figure 6: Real-World Validation on PBMC 3k scRNA-seq Data. The factual state (black triangle) is intervened upon to reach the target (black star). (Red Dashed Line): The deterministic ODE minimizes cost by crossing the void, tearing the manifold and producing an invalid ”Biological Chimera.” (Green Solid Line): The GACF triggers entropic recovery before the singularity, adaptively utilizing the valid single-cell manifold (gray dots) to safely transport the cell to a topologically valid counterfactual state.
Current continuous generative models (e.g., Diffusion Models, Flow Matching) implicitly assume that locally consistent causal mechanisms naturally yield globally coherent counterfactuals. In this paper, we prove that this assumption fails fundamentally when the causal graph exhibits non-trivial homology (e.g., structural conflicts or hidden confounders). We formalize structural causal models as cellular sheaves over Wasserstein spaces, providing a strict algebraic topological definition of cohomological obstructions in measure spaces. To ensure computational tractability and avoid deterministic singularities (which we define as manifold tearing), we introduce entropic regularization and derive the Entropic Wasserstein Causal Sheaf Laplacian, a novel system of coupled non-linear Fokker-Planck equations. Crucially, we prove an entropic pullback lemma for the first variation of pushforward measures. By integrating this with the Implicit Function Theorem (IFT) on Sinkhorn optimality conditions, we establish a direct algorithmic bridge to automatic differentiation (VJP), achieving O(1)-memory reverse-mode gradients strictly independent of the iteration horizon. Empirically, our framework successfully leverages thermodynamic noise to navigate topological barriers ("entropic tunneling") in high-dimensional scRNA-seq counterfactuals. Finally, we invert this theoretical framework to introduce the Topological Causal Score, demonstrating that our Sheaf Laplacian acts as a highly sensitive algebraic detector for topology-aware causal discovery.
Rui Wu, Hong Xie, Yongjun Li
School of Management, University of Science and Technology of China · School of Computer Science and Engineering, University of Science and Technology of China
Counterfactual prediction estimates an individual's outcome under an alternative intervention given their factual observations. Such outcomes are generally not identifiable from observational data without additional assumptions. Even within the class of fully observed additive noise models (ANMs), different causal graphs can generate the same observational distribution yet imply different individual counterfactual outcomes. Predictions based on a single estimated graph ignore this structural uncertainty. We therefore target a Bayesian counterfactual posterior predictive distribution that combines predictions from plausible SCMs. We introduce CAFE (CounterfActual Prediction via Fast Posterior Estimation), an amortized inference framework that directly approximates the Bayesian counterfactual posterior predictive distribution. We pretrain a transformer-based model on synthetic counterfactual tasks generated from a diverse prior over ANMs. Given an observational dataset, an individual's factual observations, and an intervention, CAFE approximates the corresponding posterior predictive distribution in a single forward pass. Experiments show that CAFE accurately predicts individual counterfactual outcomes in identifiable settings and approximates the posterior predictive distribution when structural uncertainty induced by observationally indistinguishable causal graphs exists. Strong performance in realistic manufacturing and viticulture settings further demonstrates its empirical robustness beyond the assumptions of the training prior.
This paper introduces a categorical account of infinitesimal causality in Frobenius Markov categories equipped with tangent-bundle semantics. IDC captures the infinitesimal layer in which interventions act as tangent deformations of copy/discard structure. Two distinct Frobenius structures interact: (1) the categorical Frobenius algebra on classical variables encoding copying, comparing, and discarding; and (2) the geometric Frobenius integrability condition, namely involutive closure of the intervention distribution, distinct from the algebraic Frobenius structure. Categorical causal sufficiency is defined as the compatibility of these two notions. A key observation is that, for structural causal models, infinitesimal causality is most naturally formulated in the slice of deterministic mechanisms over exogenous variables, with visible stochastic kernels obtained only after pushforward. Interventions are tangent vectors that deform the Frobenius copy/discard operations; their Lie brackets measure whether this deformation preserves classical information-flow structure. Pearl's do-calculus is used as a guiding example of intervention identities: ignoring irrelevant interventions corresponds to counit invariance, action/observation exchange to coproduct compatibility with pushforward, and independence to involutive bracket closure of the visible intervention distribution.
Sridhar Mahadevan
Adobe Research and University of Massachusetts, Amherst