As artificial intelligence (AI) and in particular machine learning (ML) become more integrated into marine science, it's vital to move beyond pilot projects and proofs of concept and critically examine the transition to practical applications, especially where the technology falls short. This session therefore invites presentations on the challenges, limitations, and failures of AI/ML in real-world marine contexts. We welcome discussions on:
- Limitations of current AI/ML technology and approaches.
- Technical challenges (data errors, inconsistencies, biases, sparsity, lack of standards).
- Organizational and psychological resistance to change.
- Legal and ethical concerns around human oversight, data ownership, access, and use.
We encourage submissions on negative results, model generalizability issues, mismatches with ecological complexity, AI accidents, or operational/policy failures. We are particularly interested in barriers to adoption in decision-making, including AI mistrust, stakeholder resistance, regulatory constraints, or poor model interpretability. Case studies are welcome.
By highlighting systematic gaps, we promote transparent, honest dialogue towards robust, inclusive, and context-aware AI applications in marine science. Proposed in collaboration with the ICES Working Group on ML in Marine Science (WGMLEARN) and the Data and Information Group (DIG), this session will benefit researchers, and decision-makers working at the intersection of marine science, technology, and policy.
