The Discovery Science conference presents a unique combination of latest advances in the development and analysis of methods for discovering scientific knowledge, coming from machine learning, data mining, and intelligent data analysis, with their application in various scientific domains. The 25th International Conference on Discovery Science (DS 2022) was held in Montpellier, France, during October 10–12, 2022. This was the second time the conference was organized as a stand-alone physical event. For its first 20 editions, DS was co-located with the International Conference on Algorithmic Learning Theory (ALT). In 2018 it was co-located with the 24th Interna- tional Symposium on Methodologies for Intelligent Systems (ISMIS 2018). DS 2019 was a stand-alone event, whereas DS 2020 and DS 2021 were online-only events. DS 2022 received 56 international submissions. Each submission was reviewed by at least two Program Committee (PC) members in a single-blind manner. The PC decided to accept 27 regular papers and 12 short papers. This resulted in an acceptance rate of 48% for regular papers. The conference included three keynote talks. Leman Akoglu (Carnegie Mellon Uni- versity) contributed a talk titled “Unsupervised Model Selection in Outlier Detection: The Elephant in the Room”; Luca Maria Aiello (IT University of Copenhagen) gave a presentation titled “Coloring Social Relationships”; and Stefan Kramer (University of Mainz) contributed a talk titled “35 Years of ‘Scientific Discovery: Computational Explorations of the Creative Processes’ – From the Early Days to the State of the Art”. Abstracts of the invited talks are included in the front matter of these proceedings. Besides the presentation of regular and short papers in the main program, the conference offered a session titled “Late Breaking Contributions” featuring poster and spotlight presentations of very recent research results on topics related to discovery science. We are grateful to Springer for their continued long-term support. Springer publishes the conference proceedings, as well as a regular special issue of the Machine Learning journal on discovery science. The latter offers authors a chance to publish significantly extended and reworked versions of their DS conference papers in this prestigious journal, while being open to all submissions on DS conference topics. This year, Springer (LNCS) supported the best student paper award. For DS 2022, the awardees are Annunziata D’Aversa, Stefano Polimena, Gianvito Pio, and Michelangelo Ceci (for the paper “Leveraging spatio-temporal autocorrelation phenomena to improve the forecasting of the energy consumption in smart grids”). We would like to thank Roberto Interdonato who joined the Program Chairs of the conference for the selection of the best student paper. On the program side, we would like to thank all the authors of submitted papers and the PC members for their efforts in evaluating the submitted papers, as well as the keynote speakers. On the organization side, we would like to thank all the members of the Organizing Committee, in particular Virginie Feche and Elena Demchenko, for the smooth preparation and organization of all conference associated activities. We are also grateful to the people behind EasyChair for developing the conference organization system that proved to be an essential tool in the paper submission and evaluation process, as well as in the preparation of the Springer proceedings. The DS 2022 conference was organized under the auspices of several universities and research institutes in Montpellier: the University of Montpellier, the University of Paul Valery, INRAE, Inria, and CIRAD. Significant support, especially through human resources, was also provided by the University of Montpellier and INRAE. Finally, we are indebted to all conference participants, who contributed to making this exciting event a worthwhile endeavor for all involved.