Something looks fishy! A philosophical exploration of AI for marine conservation

Increased pollution, traffic, overfishing, and climate change threaten fish populations, their habitats, and the sustainability of marine ecosystems. To develop environmental policies that support the conservation of sustainable ocean ecosystems and foster biodiversity, the use of artificial intelligence (AI) is proposed to improve knowledge production on various fish species. These fishial recognition systems (FRS) integrate machine learning, deep learning, convolutional neural networks, computer vision, drones, and robots to identify fish species, monitor endangered populations, track migration patterns, detect illegal fishing, and even track and kill invasive species. While these FRS are employed for biodiversity protection, they may change practices, shift relations, and shape the production of new knowledge and access thereto. The technical community actively adopts FRS, but the ways this technology affects fish, humans, and their shared environments remain underexplored. Therefore, this article conducts philosophically informed, exploratory research into the implications of FRS to address the central research question: how do FRS affect fish in open waters? This paper analyses the technical literature and draws on various bodies of philosophical and social science research to identify the main social and ethical considerations of FRS. It identifies potential (intentional and unintentional) biases found in the data, algorithms, or applications of FRS. It also explores the potential physical and emotional harm to fish caused by underwater FRS. FRS may further exacerbate anthropocentric, quantification, and power asymmetries between humans and fish. In response, this paper develops a fish-centric approach to the development, deployment, and use of FRS.