🔬 physics

Toward a Hybrid Digital Twin of Society: Quantifying Cognitive-Spatial Linkages Through Online-Offline Feedback Networks

This paper introduces a "Feedback Network" framework that integrates Google Search and Location History data to quantify the co-evolution of online cognitive exploration and offline physical mobility, revealing that while digital interests are more repetitive than physical movements, stable cognitive-spatial loops exist and were differentially disrupted by the COVID-19 pandemic, thereby laying the groundwork for Hybrid Digital Twins of Society.

Original authors: Rafiazka Hilman, Julia Koltai

Published 2026-06-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Rafiazka Hilman, Julia Koltai

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 your daily life in a city as a dance with two partners: Your Body (moving through physical streets) and Your Mind (scrolling through digital screens). Usually, researchers study these two dancers separately. They look at where you walk, or they look at what you search for, but rarely do they watch how the two steps influence each other in real-time.

This paper proposes a new way to watch the dance together. The authors call this the "Feedback Network," and they suggest it's the key to building a "Hybrid Digital Twin of Society."

Here is a simple breakdown of what they did and what they found, using everyday analogies.

1. The Two Worlds: The Map and the Menu

Think of your city life as having two layers:

The Offline World (The Map): This is where you physically go. You visit the grocery store, the office, the park, or the gym.
The Online World (The Menu): This is what you search for on Google. You look up "best pizza near me," "how to fix a leaky faucet," or "history of Budapest."

The paper argues that these aren't separate lists. They are a loop. You might look up a restaurant (Online) before you go there (Offline). Or, you might visit a museum (Offline) and then immediately search for "exhibition hours" or "reviews" (Online).

2. The Experiment: Tracking the Loop

The researchers acted like digital detectives. They asked 90 people in Budapest, Hungary, to share their private data from 2018 to 2022. This included:

Google Location History: Where they physically went.
Google Search History: What they typed into the search bar.

They looked for a specific rule: If you search for something, and then visit a place within one hour, that's a "link." If you visit a place and then search for something within an hour, that's also a "link." By connecting these dots, they built a giant web showing how thoughts turn into steps, and steps turn into new thoughts.

3. The "Radius of Gyration": How Far Do You Roam?

To measure how "adventurous" people were, the authors used a concept called the Radius of Gyration. Imagine a rubber band stretched around all the places you visited or topics you searched for.

A small rubber band: You only go to the same few places or search for the same few things (Routine).
A large rubber band: You travel far and wide, or search for many different, unrelated topics (Exploration).

The Big Discovery:
The researchers found a fascinating imbalance.

Your Mind (Online) is a "Routine Robot": People's search habits were very concentrated. They stuck to a narrow set of topics (like work, shopping, or news) and didn't wander far into new semantic territory.
Your Body (Offline) is a "Wandering Explorer": People physically moved around much more. They visited a wider variety of neighborhoods and types of places than the topics they searched for online.

The Analogy: Imagine your mind is a dog on a very short, tight leash, always sniffing the same three bushes. But your body is a dog off-leash, running all over the park.

4. The "Feedback Network": The City's Nervous System

The authors built a map of the connections between these two worlds. They found that the city has "Highways of Habit."

The Strongest Highway: There is a very strong loop between Business/Professional Services and General Information. People often visit offices or business districts and then immediately search for general info, or search for general info and then go to a business district.
The Shopping Loop: There is also a tight loop between Shopping searches and Retail visits.

However, the network isn't just a few straight lines. It's a mix of strong, predictable routes and many smaller, exploratory paths. This means our city life is a mix of doing the same things on autopilot and occasionally trying something new.

5. The Pandemic Effect: When the Dance Stopped

The data covered the years before, during, and after the COVID-19 pandemic. This acted like a sudden pause button on the dance floor.

The Body Stopped: Physical movement dropped sharply. People couldn't go to offices, restaurants, or parks.
The Mind Kept Dancing: Surprisingly, people's search habits didn't change as much. They kept searching for the same types of things.

This created a temporary gap: The "Rubber Band" of physical movement shrank, but the "Rubber Band" of mental exploration stayed the same size. As the world opened up again, physical movement slowly returned to its previous, more diverse patterns.

6. What is a "Hybrid Digital Twin"?

Finally, the paper suggests we need a new kind of "Digital Twin" of a city.

Old Digital Twins: These are like 3D models of buildings and roads. They show you where the traffic jams are.
The New Hybrid Twin: This model includes human thought. It doesn't just show where people are; it shows why they are there based on what they were thinking about just moments before.

In Summary:
This paper tells us that to truly understand how a city works, we can't just watch people walk. We have to watch what they are thinking about while they walk. Our digital searches and physical movements are locked in a continuous loop, but our minds tend to stay in a smaller, more repetitive circle than our feet do. By mapping this loop, we can build smarter, more human-centered models of our cities.

Technical Summary: Toward a Hybrid Digital Twin of Society

Problem Statement
Contemporary urban behavior exists within two interconnected yet analytically distinct domains: an offline visit space defined by physical mobility and destination encounters, and an online search space defined by semantic information seeking. Despite the growing recognition that digital platforms mediate how individuals perceive and navigate cities, urban mobility research typically analyzes these domains separately. This analytical distance obscures the "cognitive–spatial dualism" of modern urban systems, where online search often precedes offline mobility (revealing latent intentions) and physical encounters subsequently generate cycles of digital interaction (reviews, navigation). The paper argues that understanding urban mobility requires moving beyond physical displacement to model the recursive feedback loop between digital attention and spatial encounter.

Methodology
The study proposes a computational framework called the Feedback Network to capture continuous interactions between cognitive activity in digital environments and behavior in physical space. The research utilizes a unique dataset collected via a data donation framework in Budapest, Hungary, comprising Google Search and Location History data from 90 representative residents spanning 2018 to 2022.

The methodology proceeds through the following technical steps:

Data Preprocessing and Clustering:
- Online Space: Approximately 10 million search queries were cleaned and processed using a multilingual E5 transformer model to generate semantic embeddings. These were reduced in dimensionality via UMAP and clustered using HDBSCAN, resulting in 30 stable semantic clusters mapped to Google Trends categories (e.g., "Reference," "Shopping," "Jobs & Education").
- Offline Space: Google Location History data (300,000 visit records) was filtered for validity (minimum 5-minute duration) and reverse-geocoded. Place types were mapped to Foursquare taxonomy and clustered into 10 distinct spatial categories (e.g., "Business & Professional Services," "Dining & Drinking," "Landmarks & Outdoors").
Dual Radius of Gyration ( $R_g$ ):
To quantify exploration patterns, the authors adapted the radius of gyration metric for both domains:
- Spatial $R_g$ : Measures the geographic dispersion of physical visits around a centroid.
- Semantic $R_g$ : Measures the dispersion of search queries in the embedding vector space.
  This dual formulation allows for the comparison of physical movement breadth against cognitive exploration breadth.
Feedback Network Construction:
The core of the framework is a directed weighted graph where nodes represent semantic or spatial clusters. Edges are constructed exclusively between heterogeneous domains (online $\to$ offline or offline $\to$ online) if events occur within a one-hour time window.
- Forward Linkage Probabilities: A weighted adjacency matrix is constructed to estimate the probability of transitioning from a specific search cluster to a visit cluster (and vice versa).
- Aggregation: Individual networks are aggregated into a Collective Feedback Network to identify population-level behavioral dynamics.
Concentration Entropy ( $C_i$ ):
To evaluate the structure of these linkages, the authors introduce Concentration Entropy, defined as the complement of normalized Shannon entropy. This metric quantifies whether behavioral flows are concentrated around routine pathways (high $C_i$ ) or distributed across exploratory transitions (low $C_i$ ).

Key Results

Asymmetry in Exploration: Analysis of the Radius of Gyration reveals that online search behavior is systematically more concentrated than offline mobility. Individuals tend to reuse a narrower set of query themes compared to their physical destinations, which remain relatively more diverse. This suggests digital behavior is shaped by repetitive semantic interests, while physical movement retains broader exploratory potential.
Stable Cognitive-Spatial Loops: The Feedback Network identifies persistent linkages, particularly between "Business & Professional Services" (offline) and "Reference" (online), as well as between "Retail" and "Shopping." These stable loops indicate that specific urban routines are increasingly mediated by digital information ecosystems.
Pandemic Impact: The COVID-19 pandemic (2019–2020) disrupted spatial routines more severely than cognitive exploration. Spatial concentration dropped significantly while cognitive concentration remained stable, temporarily widening the gap between digital intent and realized movement. Post-pandemic data (2022) shows a rebound in spatial routines and a network structure that is more distributed than in 2018, with stronger co-occurrences between online and offline events.
Network Topology: The Collective Feedback Network exhibits a hybrid structure with concentrated hubs (e.g., commercial and professional service clusters) acting as bridges, alongside peripheral exploratory links. The average Concentration Entropy (0.60) indicates a moderate level of behavioral overlap, suggesting that while routine reinforcement is dominant, exploratory pathways remain embedded in the system.

Significance and Claims
The paper claims to establish the conceptual and computational foundations for a Hybrid Digital Twin of Society. Unlike traditional Digital Twins that focus on physical infrastructure (traffic, buildings, environmental sensors), this framework integrates the cognitive and behavioral dimensions of urban systems.

Theoretical Contribution: It challenges the view of mobility as solely spatial displacement, reframing it as a cognitive process embedded in informational ecosystems. It demonstrates that urban mobility relies on the interaction between informational exposure and spatial encounter.
Methodological Contribution: The Feedback Network provides a unified entropic framework to quantify the coupling between digital and physical behaviors, moving beyond isolated analyses of online or offline data.
Practical Implication: By modeling the "why" (digital intent) alongside the "where" (physical movement), the framework offers a foundation for future adaptive behavioral twins capable of predicting urban demand linkages and understanding how digital attention shapes urban flows.

The authors acknowledge limitations regarding the sample size ( $N=90$ ) but assert that the interpretable structures found at the individual and collective levels provide a robust proof-of-concept for analyzing cognitive-spatial linkages in digitally mediated urban environments.