💻 computer science

Multi-Scale Equilibrium under Variable Indicator Dimensionality: Faithful Reduction of Dynamic Attractors in Urban Mobility Systems

This paper establishes a formal framework for analyzing disrupted urban mobility systems through a dynamic multi-layer equilibrium attractor, deriving conditions and error bounds under which lower-dimensional indicator projections faithfully preserve the system's structural stability, recovery dynamics, and decision-making power to guide authorities in selecting essential monitoring metrics.

Original authors: Ali Ghoroghi, Yacine Rezgui, Afrouz Ghaemi, Cristina De Nardi, Andrei Hodorog

Published 2026-07-24

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Ali Ghoroghi, Yacine Rezgui, Afrouz Ghaemi, Cristina De Nardi, Andrei Hodorog

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

Imagine a city as a giant, living machine. It's not just roads and cars; it's a complex dance of energy, communication, people, and rules all moving at once. When something goes wrong—a storm, a strike, or a pandemic—this machine gets jolted. Scientists who study these systems use math to predict how the city will bounce back. They look for a "sweet spot" called an equilibrium, a state where everything settles down and runs smoothly again. To find this spot, they usually track dozens of different numbers, like traffic speed, bus frequency, or power usage. These numbers are called indicators. Think of them as the dashboard lights on a car. If the "check engine" light is on, you know something is wrong. But what if you could only see three lights instead of thirty? Could you still tell if the engine is fixed, or if it's just hiding a bigger problem? This is the puzzle this paper tackles: how much of the dashboard can we turn off before we stop understanding the car?

The paper, titled "Multi-Scale Equilibrium under Variable Indicator Dimensionality," is a deep dive into the math of urban recovery. It asks a very practical question: If a city can't afford to track every single piece of data during a crisis, which specific pieces of data are absolutely essential to know if the city is truly recovering? The authors don't just guess; they build a rigorous mathematical framework to prove exactly when a "reduced" view of the city is trustworthy and when it's a dangerous illusion.

The City as a Two-Layer Cake

To understand the solution, imagine the city's recovery as a two-layer cake. The top layer is the Performance Layer. This is the "fast" part of the city. It's the traffic jams clearing up, the buses running on time, and the immediate stress levels dropping. This layer reacts quickly to changes, like a rubber band snapping back.

The bottom layer is the Strategic Layer. This is the "slow" part. It's the city planners and the long-term learning. It involves big decisions like changing bus routes, building new bike lanes, or teaching people to drive differently. This layer moves slowly, like a glacier, but it sets the rules for how the fast layer behaves.

The paper introduces a concept called an Attractor. Imagine a marble rolling down a bumpy hill. No matter where you drop the marble, it eventually rolls into the same deep valley at the bottom. That valley is the "attractor." For a city, the attractor is the state of "normalcy" it tries to return to after a disaster. The paper's main job is to figure out: if we stop watching half the dashboard lights (the indicators), does the marble still roll into the same valley? Or does it get stuck in a fake valley that looks like normalcy but isn't?

The "Faithful" Reduction

The authors propose a new way to decide which data to keep. They call it a Faithful Reduction. It's not enough to just have some data; the data you keep must tell the true story of the city's recovery.

They prove that you can throw away some indicators, but only if three strict rules are followed:

The Fast Layer Must Stay Stable: The quick reactions (traffic, stress) must still settle down correctly. If you remove a key indicator, the math might say the city is fine when it's actually spiraling out of control.
The Slow Layer Must Stay Connected: The long-term plans must still match the short-term reality. If you remove data about how people are learning to adapt, the city might plan for a future that never happens.
The "Antifragility" Test Must Work: This is a fancy word for "getting stronger after a shock." The city wants to know if it bounced back better than before. The paper shows that if you remove the wrong indicators, you might miss the fact that the city actually improved, or worse, you might think it improved when it didn't.

The Three Levels of Data

To test their theory, the authors created three "city profiles" in a computer simulation:

The Minimal City: This city only tracks three things: how much traffic is moving (throughput), how stressed the network is, and how big the disturbance was. It's like checking the car's speedometer, the temperature gauge, and the oil light.
The Standard City: This city tracks six things, adding in network redundancy (backup routes) and equity (fairness for different groups). This is like adding a fuel gauge and a tire pressure monitor.
The Full City: This city tracks all twelve indicators, including energy use, communication systems, and behavioral changes. This is the full dashboard.

What the Simulations Revealed

The paper ran thousands of simulations to see what happens when these cities try to recover from a shock. Here is what they found:

1. The "Hidden" Slow Motion
The most surprising finding is about Restoration Time. If a city only watches the Minimal indicators, it will always think the city recovered faster than it actually did. It's like watching a movie in fast-forward and thinking the scene is over, while the slow-motion details are still playing out in the background. The paper proves mathematically that reduced monitoring can only understate the time it takes to recover; it can never make a slow recovery look slower than it is. This is a one-sided bias: you might think you're safe, but you're actually still in danger.

2. The "Antifragility" Trap
When the city tries to decide if it has become "antifragile" (stronger than before), the Minimal City fails the test. It lacks the data to see the subtle improvements in fairness or adaptation. The paper shows that to pass this test, you need specific indicators related to how people are learning and how different systems (like energy and transport) are working together. If you don't have these, you can't claim the city is antifragile, even if it is.

3. The "Greedy" Selection
The authors also figured out how to pick the best few indicators if you can't have all of them. They used a "greedy" strategy: pick the one indicator that gives you the most new information, then pick the next one that adds the most new info, and so on. They proved that if your data sources are independent (not repeating the same info), this greedy method is the perfect way to build a small, powerful dashboard.

The Bottom Line

This paper doesn't just say "more data is better." It gives city leaders a mathematical rulebook for knowing exactly how much data they need.

If you just want to know if traffic is moving, a few indicators might be enough.
But if you want to know if the city is truly healing, becoming fairer, or stronger than before, you need a specific set of "slow" indicators that track learning and equity.

The paper warns that cutting corners on data collection is risky. It might make a city look like it's recovering quickly, but that recovery could be an illusion, hiding deep cracks in the system that only show up when you look at the full picture. The authors conclude that while we can't always track everything, we must track the right things to ensure the city's "attractor"—its return to normal—is real, not just a mathematical mirage.

Technical Summary: Multi-Scale Equilibrium under Variable Indicator Dimensionality

Problem Statement
Urban mobility systems are increasingly modeled as high-dimensional dynamic attractors defined by a vector of indicators (e.g., throughput, redundancy, equity, stress). However, data availability varies significantly across cities and during disruptions, meaning the full indicator vector is often unobservable. The core mathematical problem addressed is the well-posedness of a high-dimensional dynamic attractor when only a subset, projection, or aggregation of its indicators is available.

Existing literature offers mature results on local dimension reduction (e.g., center manifold theory), multi-layer network aggregation, and composite indicator sensitivity. However, none of these frameworks directly address the specific challenge where an equilibrium attractor must simultaneously satisfy three concurrent requirements:

Asymptotic Stability: The projected system must preserve the stability class of the performance layer.
Fixed-Point Interpretation: The reduced performance and strategic layers must retain the same hierarchical fixed-point structure (coupling traffic assignment, modal split, and capacity learning).
Statistical Decision Power: The reduced indicator set must retain sufficient information to distinguish between recovery, resilience, and antifragility via a statistical decision rule.

Standard projection methods often focus on trajectory reconstruction or spectral preservation but fail to guarantee the preservation of external decision rules or the specific hierarchical structure of coupled fixed points.

Methodology
The paper proposes a formal framework for multi-scale equilibrium reducibility under variable indicator dimensionality. The methodology proceeds in four stages:

Model Formulation: The authors define a dynamic multi-layer equilibrium attractor for disrupted urban mobility.
- Performance Layer: A fast-relaxing state variable $E(t)$ evolves toward a target function $\bar{E}(x)$ defined over a 12-dimensional normalized indicator vector $x$ .
- Strategic Layer: A slow-timescale hierarchical fixed point combining user-equilibrium traffic assignment (Beckmann program), modal split (multinomial logit), and capacity learning.
- Antifragility Rule: A statistical decision rule comparing the post-event performance ratio to a baseline-dependent threshold to classify system behavior.
Reduction Theory: The paper defines faithfulness relative to the attractor rather than just trajectory reconstruction. It introduces four theoretical results:
- Stability Preservation: Necessary and sufficient conditions for preserving local asymptotic stability under linear indicator projection, formulated via the projectability of the vector field and the Hurwitz property of the reduced Jacobian. It extends this to approximate projectability, bounding trajectory errors using the projectability defect ( $\varepsilon_\Pi$ ) and the contraction rate ( $\alpha$ ), including a correction for non-normal transient amplification.
- Fixed-Point Preservation: A theorem proving that a coupled two-layer fixed point is preserved if the reduction maps commute with the fixed-point operator and the reduced coupling matrix remains a contraction ( $\rho(K_\Pi) < 1$ ). It bounds the displacement of the fixed point when commutation is only approximate.
- Information Preservation: A measurement model linking state variables to observable Key Performance Indicators (KPIs) via a linear-Gaussian observation equation. The authors derive the retained Fisher information in closed form and prove that detection power is monotone in nested indicator supports. They show that greedy indicator selection is exactly optimal for independent channels.
- Restoration-Time Bias: A proof that restoration time estimated from a reduced indicator set is one-sidedly biased: it never exceeds, and generically understates, the restoration time of the full system.
Simulation Protocol: The theoretical results are illustrated via a simulation study on three stylized city configurations (Bratislava, Larissa, Thessaloniki) representing full, standard, and minimal indicator supports. Five experiments verify the theoretical bounds on projection error, spectral structure, fixed-point uniqueness, decision power, and minimum admissible support.

Key Contributions
The paper makes five distinct theoretical contributions:

A Self-Contained Attractor Formalism: A unified model for disrupted urban mobility integrating a relax-to-target performance law, a hierarchical strategic fixed point, and a statistically testable antifragility rule.
Stability and Faithfulness Conditions: A necessary and sufficient condition for preserving asymptotic stability under linear projection, coupled with a "practical faithfulness" classification that quantifies error bounds when projectability is imperfect.
Coupled Fixed-Point Theorem: A proof establishing the preservation of the joint performance-strategic fixed point, identifying the contraction boundary where uniqueness is lost, and bounding fixed-point displacement under approximate commutation.
Measurement and Information Theory: A closed-form derivation of retained Fisher information for any indicator support under Gaussian noise, proving the monotonicity of detection power and the exact optimality of greedy selection for independent channels.
One-Sided Restoration Bias: A proof that reduced monitoring inevitably leads to an optimistic bias in reported restoration times, as slow modes in the kernel of the projection are hidden.

Results
The simulation study verifies the theoretical claims:

Practical Faithfulness: Observed projection errors remain within the analytic bounds derived from the projectability defect and contraction rate. The "minimal" tier (3 indicators) crosses the tolerance threshold at lower coupling strengths than the "standard" tier (6 indicators).
Restoration Time: Reduced systems consistently report restoration times earlier than the full system, with the gap widening as the initial displacement excites modes in the kernel of the projection.
Fixed-Point Convergence: As the reduced coupling matrix approaches the contraction boundary ( $\rho(K_\Pi) \to 1$ ), the number of iterations required for convergence grows hyperbolically, and the dispersion of limit points increases, signaling the loss of uniqueness.
Decision Power: The minimal indicator set fails to achieve the target statistical power (0.80) for antifragility classification, whereas the standard and full sets succeed. This confirms that information loss disproportionately affects the classification of post-disruption behavior.
Greedy Selection: The greedy addition of indicators based on innovation gain yields the minimum admissible support for the decision rule, confirming the theoretical optimality for independent channels.

Significance and Claims
The paper claims to provide a principled basis for city authorities to decide which indicators must be maintained to support specific equilibrium claims. It argues that "faithful reduction" is a stronger condition than simple approximation; a reduced model can be numerically convenient yet mathematically unfaithful if it fails to preserve stability, fixed-point structure, or decision power.

The framework shifts the definition of minimum data infrastructure from administrative convenience to mathematical necessity. It asserts that:

A city does not need full indicator coverage for all screening questions but needs enough coverage to preserve the specific property being asserted (e.g., stability vs. antifragility classification).
Reduced monitoring can only understate recovery duration, never overstate it, implying that reported improvements in restoration time under minimal monitoring should be treated as optimistic upper bounds.
The loss of an indicator has a dual role: it may shift the target value of the performance layer and simultaneously reduce the statistical power of the decision rule.

The authors explicitly state that the paper is theoretical and does not claim empirical validation of the framework on real-world data, nor does it rely on generative agent modeling. The significance lies in establishing the formal conditions under which partial indicator coverage is admissible for equilibrium analysis, distinguishing between reductions that are merely approximate and those that are structurally faithful to the attractor.