Adaptive female digital twin can personalize women’s safety in real time

Given the sensitivity of the data used in the system, the study dedicates significant attention to privacy, security and ethical compliance. According to the authors, ANGELA adheres to strict privacy frameworks including GDPR. Users must provide explicit, informed consent before the system collects any information, and they have the right to withdraw consent and request full deletion of their data.

Growing concern over violence and harassment in public spaces is intensifying global pressure to redesign cities with women’s safety at the forefront. Despite widespread deployment of mobile safety apps, city authorities and researchers warn that existing tools lack personalization and fail to account for the emotional, psychological and contextual factors that shape how women perceive risk while moving through urban areas. A new academic study argues that the next major shift in women’s mobility may come from a different direction: the rise of Human Digital Twins.

The paper, “Enhancing Safety in Urban Mobility with Female Digital Twins,” published in AI & Society, introduces AngelaDT, a customized female Human Digital Twin designed to support women during their daily journeys by dynamically modeling their safety perception, emotional state, and physiological responses, while continuously adjusting to changing urban and personal conditions. The authors position AngelaDT as part of a broader technological platform known as ANGELA, which integrates real-time data, behavioral modeling, wearable sensors and AI-driven safety indicators to deliver adaptive safe-route guidance and emergency response capabilities.

The study outlines the system’s architecture, theoretical grounding and life cycle, offering what may be the first comprehensive framework for applying Human Digital Twin technology to women’s safety in urban public spaces.

Human digital twins emerging as a new foundation for urban safety

The study examines why traditional safety tools fall short. Many existing apps provide static maps, crowd-sourced alerts, or panic buttons, yet they do little to assess how a woman personally experiences fear or discomfort in the moment. The authors point out that safety perception varies widely based on age, personality, emotional state, prior experiences, cultural background and the physical environment. For this reason, a one-size-fits-all approach is insufficient for addressing real-world risk.

Human Digital Twins, by contrast, offer a level of personalization and adaptability that conventional apps cannot match. A Human Digital Twin is defined as a digital representation of a person that evolves over time, reflecting physical, behavioral, emotional, social and physiological characteristics. The study explains that HDTs emerged in fields such as fitness tracking, personalized healthcare and precision medicine, where they are used to simulate, predict and monitor an individual’s state with high accuracy.

AngelaDT applies this concept directly to women’s safety during urban mobility. The digital twin collects and processes five major categories of data: profile information such as age and gender; personality traits and emotional tendencies; route-based safety assessments provided by the woman; physiological indicators related to fear, such as heart rate or respiratory rate; and emergency data including real-time geolocation. These inputs allow AngelaDT to learn how each woman perceives safety in various environments and to guide her accordingly.

The authors point up that the system is not merely reactive. AngelaDT’s design allows it to anticipate potential risk scenarios by analyzing physiological changes that may signal stress or fear. If the system detects an anomaly that matches known patterns of distress, it issues a warning. If the user fails to respond or confirms that she needs assistance, AngelaDT can activate audio and video recording, send emergency information to authorized security services, and provide real-time updates from the user’s device.

The system’s intelligence depends heavily on its ability to integrate multiple data streams. The study outlines how the platform’s Hybrid Data Module manages both real-time and stored data using a data lake architecture. This allows the system to combine open data, sensor inputs and long-term behavioral patterns into a unified safety model. The data lake also enables continuous refinement of the platform’s urban safety indicators by analyzing aggregated feedback from users.

The study notes that HDTs are not yet common in mobility applications. Most research and commercial initiatives have focused on manufacturing, human–computer interaction, healthcare or wellness. By applying this technology to urban safety, the authors argue that AngelaDT introduces a new class of digital companion capable of providing personalized guidance with higher precision than traditional apps.

A platform designed to adapt to each woman’s emotional and physical state

The ANGELA platform is a three-module system designed to support AngelaDT and deliver personalized safety services.

• The Metrics Discovery Module identifies the urban, social and economic indicators that influence perceived safety. These include environmental features such as lighting, the presence of surveillance cameras, street cleanliness, visibility obstructions, foot traffic and availability of public transport stops. The authors connect these indicators to established theories from urban planning, social psychology and criminology, such as Defensible Space Theory, Social Disorganization Theory, Broken Windows Theory and Eyes on the Street. By grounding the indicators in these frameworks, the platform aligns its safety calculations with decades of research on environmental risk.
• The Hybrid Data Module is responsible for sourcing, integrating and harmonizing data streams from open databases, social networks, mobile sensors, IoT devices and wearable technologies. The use of a data lake allows the system to store structured, semi-structured and unstructured data in their original formats, supporting complex AI-driven analyses that refine safety indicators over time. As users provide feedback on routes, the system learns which urban characteristics align with different safety perceptions and adjusts its metrics accordingly.
• The Interaction Module houses the mobile app, appSer, and the AngelaDT digital twin. The app generates personalized safe routes, highlights potentially unsafe areas and allows women to rate their perceptions of safety after completing a journey. At the same time, AngelaDT provides advanced support by monitoring physiological patterns and initiating emergency protocols when needed.

The authors describe a scenario in which a woman returning home late at night requests a safe route. AngelaDT considers her personality traits, emotional tendencies, historical safety assessments, and real-time physiological data in conjunction with urban indicators. For a woman who is highly fearful, the system chooses well-lit streets, avoids areas with visual obstacles such as garbage containers, prioritizes zones with visible public services and selects paths that align with her prior experiences.

This personalization ensures that no two users receive the same route recommendations, even if they start at the same point and head to the same destination. The system evolves as the woman’s emotional state and life circumstances change, maintaining accuracy over time.

A life cycle built on privacy, trust and continuous learning

Given the sensitivity of the data used in the system, the study dedicates significant attention to privacy, security and ethical compliance. According to the authors, ANGELA adheres to strict privacy frameworks including GDPR. Users must provide explicit, informed consent before the system collects any information, and they have the right to withdraw consent and request full deletion of their data.

All personal information is encrypted and anonymized when stored in the data lake. In emergency scenarios, however, certain critical data, such as real-time location and physiological signals, may be shared with authorized security services, but only if the user opts in to this feature during setup. This dual approach aims to balance user privacy with the need for rapid emergency response.

The study explains that AngelaDT undergoes a two-phase life cycle: development and evolution. During development, the system’s behavior is modeled using software design tools and tested using synthetic datasets. Once real users join the platform, the evolution phase begins. AngelaDT is configured with the woman’s initial profile and personality traits, then continuously updated through a Human-in-the-Loop strategy, where user feedback helps refine the model. This allows the system to adjust to emotional changes, aging, evolving personal contexts and new urban data.

The authors acknowledge limitations, including variability in wearable device accuracy, subjective differences in safety perception and the need for real-world pilot testing. They also highlight the importance of preventing bias in data models to ensure equitable safety recommendations across different demographic groups.