KAIST

< (From left)  Integrated M.S.-Ph.D candidate Keonhee Jang, Postdoctoral Researcher Namwoo Kim, Professor Yoonjin Yoon, Researcher Seok-woo Yoon, Postdoctoral Researcher Young-jun Park, (Top) M.S candidate Juneyoung Ro >

KAIST announced on October 29th that its Urban AI Research Institute (Director, Distinguished Professor Yoonjin Yoon of Civil and Environmental Engineering conducted joint research in the field of 'Urban AI' with MIT's Senseable City Lab (Director, Professor Carlo Ratti) and disclosed the results at the 'Smart Life Week 2025' exhibition held at COEX, Seoul, in late September.

KAIST and MIT have been pursuing the 'Urban AI Joint Research Program' to interpret major urban problems using artificial intelligence. At this exhibition, the research results were presented in a form that citizens could directly experience, focusing on three themes: ▲Urban Climate Change, ▲Green Environment, and ▲Data Inclusivity.

Through this collaboration, the two institutions demonstrated that AI technology can expand beyond a tool for calculating urban problems to a new intelligence that promotes social understanding and empathy. They carried out three projects: ▲Urban Heat and Sales, ▲Nature That Heals, Seoul, and ▲Data Sonification.

The first project, 'Urban Heat and Sales,' is a study that analyzes the impact of climate change on urban commercial areas and the small business ecosystem using AI. An AI model was trained on over 300 million data points, including sales and weather for 96 business categories across 426 administrative dong (neighborhoods) in Seoul, to quantify the effect of climatic factors, such as temperature and humidity, on sales by industry type.

The results were visualized into 40,896 'Urban Heat Resilience' indicators, which score how well each region and business category can adapt to and recover from climate change. This allows the level of commercial area resilience to climate risk to be grasped at a glance, showing which areas are strong against temperature risks.

According to the study, for the convenience store sector, 64.7% of the total 426 dong were analyzed as 'climate-neutral areas,' which are relatively stable against climate change, while the remaining 35.3% belong to 'climate-sensitive areas,' which are significantly affected by climate change. This suggests that the operating environment for convenience stores varies significantly by region in terms of climate impact, and the data can be utilized for future location strategy planning from an urban resilience perspective.

< '3D Mesh Structure' that visually represents sales data for 426 regions in Seoul. The height and color of each region indicate the scale of sales. The left shows the distribution of sales in Seoul under actual temperature conditions, and the right shows the sales change predicted by AI when the temperature rises by 5 degrees. >

Visitors to the exhibition could select a region and business type on a real Seoul map and experience a system where the AI predicted sales changes in real-time based on future temperature rise scenarios.

This prediction model is a proprietary technology developed by KAIST, and plans are underway to expand cooperation with other major global cities, such as Boston and London. This research is expected to propose a new direction for establishing opening strategies for small business owners and developing urban climate risk response policies.

< Numerous visitors listening to explanations and experiencing the KAIST-MIT exhibition space >

The second project, 'Nature That Heals, Seoul,' is an extension of MIT's global project 'Feeling Nature' to Seoul. It combines urban environment data (Street View, maps, satellite images, etc.) with citizen survey data to train an AI to estimate the 'psychological green'—the actual psychological experience of green spaces felt by Seoul citizens.

This approach goes beyond simply calculating the area of trees or parks, offering new urban design directions that reflect citizens' emotional resilience and well-being. This research is expected to provide scientific evidence for future Seoul green space policies and locally tailored urban design.

The final project, 'Data Sonification,' is the world's first AI technology that translates over 300 million data points into sounds, like music, to be 'heard.' The AI uses data such as temperature, humidity, and sales to represent information through sound: for example, the pitch rises when the temperature goes up, and the sound lowers when sales decrease. This provides a new sensory experience of 'listening' to urban data through sound instead of sight.

This technology is a prime example of 'Barrier-Free AI' (AI for All), an inclusive AI technology that helps people with visual impairments or children—who may have difficulty accessing visual information—to intuitively understand data.

< A visitor experiencing Data Sonification, the world's first AI technology that converts data into sound >

Man-ki Kim, Chairman of the Seoul AI Hub (Seoul AI Foundation), which sponsored this research, stated, "We have achieved meaningful results by analyzing the urban environment and citizens' lives with artificial intelligence through collaboration with world-class research institutions like KAIST and MIT," adding, "This research has laid the groundwork for understanding urban change from the perspective of citizens and connecting it to policy and daily life."

Director Yoonjin Yoon remarked, "This exhibition demonstrated that artificial intelligence can evolve beyond a technology that merely calculates the city to an intelligence that understands and empathizes with people and the city," and concluded, "We will create data and experiences together with citizens, and collaborate with various cities worldwide to open a more inclusive and sustainable urban future."

This achievement is a global collaborative research project in the AI sector involving the KAIST Urban AI Research Institute and the MIT Senseable City Lab, and was conducted with sponsorship from the Seoul AI Hub.

※Research Results Images/Videos: https://05970c0c.slw-6vy.pages.dev/