AIS-based profiling of fishing vessels falls short as a “proof of concept” for identifying forced labour at sea


Forced labour at sea is a serious human rights violation that must be addressed with policies that consider complex social, economic and political dynamics. So when a recent research paper in PNAS (McDonald et al.) concluded that a machine learning algorithm could identify vessels that are at “high risk” of  engaging in labour abuse, researchers from the Nippon Foundation Ocean Nexus Center took notice. 

Upon a closer review of the methodology, the researchers were left with strong reservations towards this study due to the limited dataset, unsubstantiated assumptions made by the authors, and the failure to robustly validate the model’s findings. Led by deputy director Wilf Swartz, the researchers were quick to submit a letter to the PNAS Editorial Board to voice their concerns.

“In addition to the reservations listed above, the weakly validated AI-driven profiling risks being overly reductive and potentially discriminatory. Given the real and severe implications of labour abuse, we argue that this type of approach is inappropriate,” said Swartz, Marine Affairs Program, Dalhousie University.

Other authors of the commentary include Andrés M. Cisneros-Montemayor, Institute for the Oceans and Fisheries, University of British Columbia; Gerald Singh, Department of Geography, Memorial University of Newfoundland; Patrick Boutet, Department of Computer Science, University of British Columbia; and Yoshi Ota, School of Marine and Environmental Affairs, University of Washington.

The full details of the Ocean Nexus letter to the editor can be viewed here.

For more information about the commentary, please contact Wilf Swartz at wilf.swartz@dal.ca.