This training school was organised within the Working Group on Reference Frameworks (WG3) of the COST Action on Artistic Intelligence, which investigates the collective capacity of artistic and practice-based research to generate impact and value beyond the singular outcomes of individual projects.The training school took the form of a collaborative workshop exploring how different dimensions of artistic research may intertwine, reference and influence one another. In this context, artistic practice operates transversally - as method, process, and result. These dimensions continuously inform and reshape each other, challenging the expectation that research must culminate in fixed or rigid formats. Instead, the workshop foregrounds modes of inquiry that are dynamic, reflexive, and responsive to the specific conditions, social contexts and materialities of practice. Two platforms formed a starting point for the workshop: the Research Catalogue (RC), a digital platform for documenting artistic research (provided by SAR) by means of creating online graphical presentations of text, image, sound and video. The other was KOBI: an AI-powered knowledge system that uses large language models and semantic knowledge graphs to connect, visualise, and help users navigate diverse sources of knowledge, fostering collective intelligence and creativity. Image
Despite this technological starting point, it was certainly also a social and collaborative occasion. The training school gave the opportunity to use and discuss these tools, in a group environment. The aim was to investigate how do the artistic research projects of the participants relate to each other, are there concepts, methodologies and findings that can be transferred between projects and researchers in ways that are unique to the field of artistic research? How can tools like RC and KOBI help with this? To make exploring such ideas concrete, the format of the workshop was designed as follows: after a brief introduction by the participants, we started to collect a number of areas of interest by means of keywords, which can be found here: https://www.researchcatalogue.net/view/4667244/4667245/1338/55. After a short introduction to the Research Catalogue editor, we then split the group into two: one half of the group would work in pairs, using the keywords, and contribute to the main space within the RC exposition, while the other half would work on an individual page (switching roles after a break). Especially the pairs were very important, we wanted to experiment with collaborative writing in a shared context. A key aspect was that all these pairs worked within the same RC page shared by all participants (the "blackboard"), so that any input had to be embedded into the context of materials what the other researchers had contributed before. You can see a visual overview of that collective page and the individual pages here: https://map.rcdata.org/#/exposition?id=4667244. Image
The resulting exposition content was fed into the KOBI environment to create an interactive semantic space and lexicon. We experimented and discussed the space formed by KOBI. Despite the brief time and the usual technical challenges, overal the response to the workshop was very positive! I was pleased to see that we were able at several moments to move beyond keyword concepts and reference ideas in a more embedded and specific manner. For me, this is also what makes KOBI different from other LLM based systems: there is always a clear and explicit reference back to the context from which something is generated. Also participants enjoyed a lot being able to navigate and even stretch the semantic space, which would create new dynamic perspectives on the materials. One conclusion that I would make is that it was somehow the combination of the physical in-person interaction and the digital spaces of both RC and KOBI that made the exchange different and more rich than any of the tools would be taken in isolation.
|
Image
