University of Zürich
University of Zürich
01/27/2026 | News release | Distributed by Public on 01/27/2026 03:34
When Spirituality Meets AI
27.01.2026 Postdoc Team Award
When Spirituality Meets AI
A team of four researchers from the fields of spiritual care, computer science, computational linguistics and educational science has been awarded the UZH Postdoc Team Award for developing SpiritRAG. The AI tool searches thousands of United Nations documents to retrieve grounded answers to questions about religion and spirituality.
Carole Scheidegger
Categories
The team behind SpiritRAG: Patrick Montjouridès, Anastassia Shaitarova, Fabian Winiger and Yingqiang Gao. (Picture: UZH)
The world has access to more information than ever before, and yet we still often struggle to find the facts that matter. Given the sheer volume of data, locating relevant facts can feel like searching for a needle in a haystack. This is especially true of large or poorly structured collections of documents, such as the archives of various UN bodies.
The problem is even more pronounced for topics where meaning varies depending on cultural or historical context. Anthropologist Fabian Winiger, for example, studies UN documents on different aspects of religion and spirituality. "If you search for the word 'spirit', you get a huge number of hits," he explains, "but most of them are useless because they simply refer to phrases like 'in the spirit of …' in the preamble to a resolution." For researchers in his field, this is a frustrating limitation.
Interdisciplinary expertise
Rather than resigning themselves to this problem, Winiger joined forces with three colleagues from other fields to develop an AI tool designed to make these searches more effective. The team includes computer scientist Yingqiang Gao, computational linguist Anastassia Shaitarova and education scholar Patrick Montjouridès. Each brought their specific expertise to the table: Gao led the technical implementation, Winiger provided subject-matter expertise on religion and spirituality, Shaitarova provided general textual analysis of the data as well as insights into the temporal distributions of religious themes, while Montjouridès contributed insights into political contexts and international relations and, drawing on his experience as a former UNESCO staff member, explained the United Nations administrative architecture and document structures.
This collaboration led to the creation of SpiritRAG, an interactive question-answering system that connects state-of-the-art large language models with more than 7,500 United Nations documents related to religion, spirituality, health and education. The tool is designed to support rigorous, document-grounded knowledge acquisition in settings where concepts are context-dependent, contested and historically situated.
Science in hardcore mode
The four researchers behind SpiritRAG have now received the University of Zurich's 2025 Postdoc Team Award. Winiger is delighted that their efforts are being recognized: "Interdisciplinary work is basically science in 'hardcore mode'," he says, referring to the ultimate difficulty level in many video games. "It requires not only excellence in one's own field, but also a good working understanding of how others do science. Otherwise, you end up talking past each other. It takes time and effort - often much more so than working in one's own disciplinary habitus."
The four researchers not only come from different disciplines but also have diverse cultural and linguistic backgrounds. This, however, was no obstacle: "We worked together incredibly well and efficiently; communication was always flowing," Gao says. "We're like the Beatles, each with our own quirks," Shaitarova jokes. "Fabian and Yingqiang are our John and Paul. Then Patrick, just like George, brought in an exotic tune with his UNESCO experience. I joined last. I guess I'm Ringo with my steady beat of linguistic data analysis."
UZH Postdoc Team Award
The UZH Postdoc Team Award was introduced in 2022. The award recognizes interdisciplinary postdoctoral teams for their outstanding and independent scientific achievements, with a focus on research projects that stand out for their societal relevance and/or impact. The award is worth CHF 15,000, funded through a generous donation to the President's Fund of the UZH Foundation.
The work did not proceed entirely without challenges, however. After the first tests, the team realized that many users interacted with the tool as if it were a normal chatbot, which led to odd responses. SpiritRAG isn't a general-purpose chatbot, Montjouridès notes, but operates as a research infrastructure. "It retrieves relevant resolutions, reports and policy texts from a carefully curated corpus of United Nations documents and then uses generative AI to formulate answers." To help users get the most out of the tool, SpiritRAG's homepage features a sample query: "How does the UN address religion and/or spirituality in conflict-affected regions?" The answers always include the underlying sources, allowing users to consult the original documents and pursue their research further if they wish. The tool is aimed primarily at researchers and policymakers.
What's RAG?
RAG refers to Retrieval Augmented Generation, a technique that enables large language models (LLMs) to retrieve information from external data sources and incorporate it into their responses.
SpiritRAG was presented at the prestigious EMNLP 2025 conference in Suzhou, China, as well as at the AI+X Summit in Zurich. The tool, which is also regularly used by people outside of UZH, has prompted interest from several organizations looking to apply it for their own work.
Thanks to its lightweight, modular open-source architecture, SpiritRAG can be deployed with different datasets while using minimal resources. This also gives other disciplines that deal with context-sensitive topics and complex datasets a better chance of finding the proverbial needle in the haystack.
Carole Scheidegger, UZH News
University of Zürich published this content on January 27, 2026, and is solely responsible for the information contained herein. Distributed via Public Technologies (PUBT), unedited and unaltered, on January 27, 2026 at 09:34 UTC. If you believe the information included in the content is inaccurate or outdated and requires editing or removal, please contact us at [email protected]