Community-led AI systems for scholarly communication

For years, the academic discovery market has been dominated by a few proprietary systems. In the shadows of these giants, however, an alternative infrastructure has emerged, built on thousands of archives, repositories, and aggregators, and championed by libraries, non-profits, and open-source developers. Unlike commercial players, these systems make their (meta-)data openly available, driving innovation and fostering the development of diverse discovery tools. As the scientific corpus continues its rapid growth, a new generation of discovery systems is emerging, based on new technologies such as AI, visualisation, and semantic networks. However, the increasing shift toward subscription-based AI tools introduces new paywalls and thus barriers to knowledge access. This raises a critical question: How can we develop free, open alternatives aligned with Open Science principles? In this presentation, we discuss the participatory approach of Open Knowledge Maps (OKMaps) as a response to this challenge. OKMaps (https://openknowledgemaps.org) is a charitable non-profit organisation dedicated to dramatically increasing the visibility of scientific knowledge. To this end, we operate the world's largest AI-based search engine for research, enabling users to create visual overviews of research topics across all disciplines. These so-called knowledge maps provide an immediate overview of a topic, highlighting key sub-areas and linking them to relevant resources and concepts. OKMaps draws from over 400 million outputs from its primary data providers, BASE and PubMed. To date, more than 3 million knowledge maps have been created by users from over 200 countries and territories. OKMaps is a diamond open infrastructure: all of our services are free and open. We share our source code, content, and data under open licences, ensuring community ownership. Furthermore, we develop our services together with our community, who play a crucial role in governance and decision-making. Supporting members and the wider community determine two-thirds of our technical roadmap, with the goal of creating an inclusive and fair infrastructure that balances the needs of its stakeholders equitably. A key outcome of this participatory approach is the development of institutional services, known as Custom Services. These enable institutions to enhance their discovery offerings with AI-based search components from OKMaps. Examples of integrations include a wide variety of library catalogues such as at ETH Zurich and the University of Eastern Finland, the EOSC platform GoTriple, and the Austrian Social Science Data Archive (AUSSDA). In conclusion, this approach can create a win-win-win scenario for researchers, institutions, and the open infrastructure. However, a significant challenge remains: attracting enough supporting members to secure basic funding for maintenance and further development. We will conclude our presentation with strategies to address this challenge. See this presentation in this video recording.