DangerMaps

This project develops a system for providing personalized safety or accessibility advice to travelers.

Planning a trip can be a daunting task, especially for trips to potential unsafe locations. For a given destination, safety-related questions might be: “Where is it safe for me?”, “Which areas should I avoid at night?”, or “Which direction should I head toward to get to safety?”

When planning a trip—or when seeking information just-in-time in a location—numerous online information sources would need to be searched and reviewed to determine the answer to these questions. Compounding this problem is that each traveler’s perception of danger and safety is different. What is perceived as a risk (or potential danger) for one person in one location may not be perceived as a risk by another person in the same location. The personal demographic attributes of a person (e.g., age and gender) determine the person’s safety-related need for information.

Search engines fail to satisfy these personal information needs, because they do not personalize their results to contextual information. Further complicating the search for safety-related travel information is that this information is dispersed on the Web and may be hard to find. Therefore, researching the safety of a location requires cognitive effort that can be regarded as a sensemaking activity. This type of cognitive activity involves foraging for information: tedious online research on neighborhoods, hotels, modes of travel, locations, and many other safety-related factors. This highlights the need for solutions that provide personalized safety advice to tourists and travelers.

Large language models (LLMs) could provide an opportunity to support tourists and travelers in researching safety-related information for their travel destinations. But while LLMs are potentially capable to adapt and personalize their responses to users’ personal information needs, their intrinsic knowledge is not granular enough for detailed travel advice. However, the LLMs’ ability to understand instructions and learn in-context could be employed to provide the LLM with the missing information needed to provide personalized safety advice to travelers. Given personal details about a traveler, such as basic demographics and whether the person is traveling alone or in a group, an LLM could provide a personal safety assessment for a given geographic location. For instance, the LLM could be provided with the following information: “I am a woman, 21 years of age, I travel on foot and alone, this is my current location, these are points of interest (POIs) in the vicinity.” Given this information, the LLM could answer the safety-critical question: “How safe am I in this location?”

This project will implement a prototype that uses a large language model for the knowledge-intensive task of providing personalized safety advice to travelers.

References

1. Jonas Oppenlaender. 2025. DangerMaps: Personalized Safety Advice for Travel in Urban Environments using a Retrieval-Augmented Language Model. https://arxiv.org/abs/2503.14103

2. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval-augmented generation for knowledge-intensive NLP tasks. In Proceedings of the 34th International Conference on Neural Information Processing Systems (NIPS ’20). Curran Associates Inc., Red Hook, NY, USA, Article 793, 16 pages. https://proceedings.neurips.cc/paper_files/paper/2020/file/6b493230205f780e1bc26945df7481e5-Paper.pdf

3. Denny L. Starks, Tawanna Dillahunt, and Oliver L. Haimson. 2019. Designing Technology to Support Safety for Transgender Women & Non-Binary People of Color. In Companion Publication of the 2019 on Designing Interactive Systems Conference 2019 Companion (DIS ’19 Companion). Association for Computing Machinery, New York, NY, USA, 289–294. https://doi.org/10.1145/3301019.3323898

4. Christine Satchell and Marcus Foth. 2010. Fear and Danger in Nocturnal Urban Environments. In Proceedings of the 22nd Conference of the Computer-Human Interaction Special Interest Group of Australia on Computer-Human Interaction (OZCHI ’10). Association for Computing Machinery, New York, NY, USA, 380–383. https://doi.org/10.1145/1952222.1952308

5. Michela Arnaboldi, Marco Brambilla, Beatrice Cassottana, Paolo Ciuccarelli, and Simone Vantini. 2016. How Twitter Reveals Cities within Cities. In Proceedings of the 7th 2016 International Conference on Social Media & Society (SMSociety ’16). Association for Computing Machinery, New York, NY, USA, Article 13, 11 pages. https://doi.org/10.1145/2930971.2930985

6. Marianne Aubin Le Quéré, Ting-Wei Chiang, Karen Levy, and Mor Naaman. 2022. Information Needs of Essential Workers During the COVID-19 Pandemic. Proc. ACM Hum.-Comput. Interact. 6, CSCW2, Article 306 (nov 2022), 19 pages. https://doi.org/10.1145/3555197

7. Mark Bilandzic, Marcus Foth, and Alexander De Luca. 2008. CityFlocks: Designing Social Navigation for Urban Mobile Information Systems. In Proceedings of the 7th ACM Conference on Designing Interactive Systems (DIS ’08). Association for Computing Machinery, New York, NY, USA, 174–183. https://doi.org/10.1145/1394445.1394464

8. Jan Blom, Divya Viswanathan, Mirjana Spasojevic, Janet Go, Karthik Acharya, and Robert Ahonius. 2010. Fear and the City: Role of Mobile Services in Harnessing Safety and Security in Urban Use Contexts. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’10). Association for Computing Machinery, New York, NY, USA, 1841–1850. https://doi.org/10.1145/1753326.1753602

9. Wes Gurnee and Max Tegmark. 2023. Language Models Represent Space and Time. arXiv:2310.02207

10. Margaret Hagan, Nan Zhang, and Joseph ’Jofish’ Kaye. 2012. Safe Mathare: A Mobile System for Women’s Safe Commutes in the Slums. In Proceedings of the 14th International Conference on Human-Computer Interaction with Mobile Devices and Services Companion (MobileHCI ’12). Association for Computing Machinery, New York, NY, USA, 47–52. https://doi.org/10.1145/2371664.2371675

11. M. Hamilton, F. Salim, E. Cheng, and S. L. Choy. 2011. Transafe: A Crowdsourced Mobile Platform for Crime and Safety Perception Management. SIGCAS Comput. Soc. 41, 2 (dec 2011), 32–37. https://doi.org/10.1145/2095272.2095275

12. MD Romael Haque, Katherine Weathington, Joseph Chudzik, and Shion Guha. 2020. Understanding Law Enforcement and Common Peoples’ Perspectives on Designing Explainable Crime Mapping Algorithms. In Conference Companion Publication of the 2020 on Computer Supported Cooperative Work and Social Computing (CSCW ’20 Companion). Association for Computing Machinery, New York, NY, USA, 269–273. https://doi.org/10.1145/3406865.3418330

13. Cristina Kadar and Irena Pletikosa Cvijikj. 2014. CityWatch: The Personalized Crime Prevention Assistant. In Proceedings of the 13th International Conference on Mobile and Ubiquitous Multimedia (MUM ’14). Association for Computing Machinery, New York, NY, USA, 260–261. https://doi.org/10.1145/2677972.2678008

14. Takuhiro Kagawa, Sachio Saiki, and Masahide Nakamura. 2017. Visualizing and Analyzing Street Crimes Using Personalized Security Information Service PRISM. In Proceedings of the 19th International Conference on Information Integration and Web-Based Applications & Services (iiWAS ’17). Association for Computing Machinery, New York, NY, USA, 208–214. https://doi.org/10.1145/3151759.3151785