Spatial-Knowledge-Graph-Grounded LLM Agents for Neighborhood Livability Evaluation
This paper proposes a framework integrating spatial knowledge graphs and large language models to simulate household schedules and evaluate neighborhood livability by revealing how mobility constraints and care responsibilities create travel burdens that static facility indicators fail to capture.
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
We often judge a neighborhood by what is nearby. If a grocery store, a clinic, and a park sit within a short walk of a home, planners and maps often declare the area "livable." This view assumes that if a place exists, it is accessible to everyone. But this static picture misses the reality of daily life. A healthy adult might breeze past a clinic in five minutes, while an older person with slow walking speed might find that same five-minute walk exhausting. A parent might find a park nearby, but if they must also pick up a child from school in the opposite direction, that park becomes a logistical puzzle rather than a convenience. The true measure of a neighborhood is not just what is there, but how different people—with different speeds, schedules, and family duties—can actually use it.
Researchers at The Chinese University of Hong Kong, Shenzhen, have built a new way to test this reality. Instead of relying on simple maps or surveys, they created a digital experiment that simulates the lives of different households over a week. They used a system that combines a detailed map of the neighborhood with a smart computer program capable of reasoning like a human. This program creates "agents," or digital residents, each with a specific personality, age, and set of responsibilities. Some are young professionals rushing to work; others are grandparents living alone; some are parents managing a toddler and a school-aged child. The system asks these digital residents to plan their week, deciding when to go to the doctor, where to buy groceries, and how to get there.
The process begins by feeding the computer a complete picture of the neighborhood, including every road, building, and facility. The system then selects a specific household and asks the computer to generate a schedule for them. Crucially, the computer does not just guess; it checks its work against the actual map. If the schedule says a parent needs to drop a child at school and then run to a pharmacy, the system calculates the real walking time between those two spots. If the math shows the trip is impossible within the available time, the computer rewrites the schedule, perhaps by choosing a closer pharmacy or changing the order of events. This cycle of planning, checking, and revising continues until a realistic, feasible week is created for every household.
Once the schedules are set, the system traces the actual paths the residents would take. It calculates exactly how long each trip takes, what mode of travel is used, and how much effort is required. The result is a detailed log of a week in the life of each digital resident. In a test run within a neighborhood in Shenzhen, the researchers simulated six different households, including a widowed older woman, a young couple with an infant, and a family with a school-age child. The simulation showed that while all the necessary facilities were technically available in the area, the experience of using them varied wildly.
For the young couple without children, the neighborhood felt convenient, with most trips taking less than twenty minutes. For the older woman living alone, however, the same neighborhood felt burdensome. Although she could reach her destinations, her average trip took over twenty-five minutes, and she rarely completed a trip in under fifteen minutes. The parents of young children faced a different kind of challenge: their schedules were filled with coordination. Every trip involving the infant or the school-aged child required an adult to accompany them, creating a heavy load of shared travel. The simulation revealed that the "burden" of living in a place is not just about distance, but about the friction of moving between different tasks while managing the needs of others.
To understand how these digital residents felt about their week, the researchers asked them to rate their experience. The older woman and the parent of the infant gave the lowest scores for healthcare access and the ease of combining errands. They noted that while they could eventually reach a hospital or a pharmacy, the travel time was long and tiring. The young couple, by contrast, rated their experience much higher, finding it easy to manage their daily needs. These ratings were not random guesses; they were directly tied to the specific trips the system had calculated. When the computer showed that a resident had to walk for thirty minutes to get medicine, the resident's digital response reflected that difficulty.
The study suggests that a neighborhood can be full of amenities and still feel difficult to live in for certain groups. The researchers found that simply having a facility nearby does not guarantee it is usable. For a person with limited mobility, a clinic that is two kilometers away might as well be across town. For a parent juggling a stroller and a school run, a park that requires a long detour is effectively out of reach. The simulation proved that the "livability" of a place depends heavily on who is living there and what they are trying to do.
This approach offers a new tool for urban planners and community leaders. Instead of assuming that a map of facilities tells the whole story, they can now simulate how different types of people would actually navigate their daily lives. The system can show where the hidden burdens lie—perhaps a lack of nearby pharmacies for the elderly, or a shortage of family-friendly spaces for parents. By understanding these specific constraints, cities can be designed not just to have services, but to make those services truly accessible to the people who need them most. The work does not claim to have solved the problem of livability, but it provides a clear, evidence-based way to see the gap between what a neighborhood offers and what its residents can actually use.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.