Paper short abstract:

This paper explores how a series of art works made with participants, plants, and machine learning co-produced knowledge, and whether this paper contributes to how that question can be answered.

Paper long abstract:

Building a machine learning image model from scratch requires taking hundreds of photographs to inform a model. This paper explores how involving participants in producing that data set can contribute to STS studies of vegetal-socio-technical interactions. This case study of two art works Seeding Things No.3 2020, and Gathering Downstream 2022, explores what was learned when knowledge was created in interactions between plants, machines and human participants. The first worked with participants to plant and care for a grass and clay mountain, and to contribute photographs to a collective model. Emerging characteristics / biases were specific to that group, and enabled new understandings of machine learning. The second used museum archives and environmental images to ask questions about the impact of industrial histories on environmental futures. Influenced by STS and Anthropocene issues of environmental impact, artists are increasingly concerned with the agency of the more-than-human. Ideas of authorship and agency have been challenged through collaborative, participatory, and site-based practices in which ‘context is half the work’ (APG 1965). AI and Machine Learning have led to further investigations of where agency lies in creative production (Zeilinger 2021, Zylinska 2020), and ownership of training images. The case studies I use assembled new visual knowledge within the Machine Learning models and engaged publics and museums in discussion of more-than-human learning, and the impact of creating ML models. Finally, a discussion of writing will problematise exhibition texts and academic papers which ironically, can cement whether the work is recorded as research and/or engagement.