Considerations for Artificial Intelligence and Machine Learning Applications in Electronic Monitoring
This paper is part of a series that summarizes discussions from the 2022 Global Electronic Monitoring Symposium, which convened more than 50 EM experts, both in person and virtually, for a three-day workshop. The symposium focused both on the use of electronic monitoring programs to increase oversight and transparency in international fisheries management and on existing barriers to the uptake of EM. Although this series of papers does not represent an exhaustive discussion of the issues, it includes the key points that symposium participants raised.
Global Electronic Monitoring Symposium (GEMS) participants noted that the EM review process is one of the major challenges to scaling electronic monitoring (EM) worldwide. Extracting data and reviewing video footage, a key element of an EM program, forms the most significant ongoing cost: the more footage reviewed, and the more detailed the data, the more expensive the process. To prevent burdening those responsible with manually reviewing millions of hours of footage annually, many fishery stakeholders are looking to use artificial intelligence (AI) and machine learning (ML) applications to conduct reviews.
Potential uses of AI/ML are numerous across the different phases and uses of EM data; almost any stage of EM in which decisions are made based on data (e.g., video review or annotation) could be automated by ML algorithms. AI/ML also has the potential for EM cost-reducing opportunities. However, stakeholders need to be pragmatic and informed about how AI/ML is developed, including the cost; how AI/ML is integrated into the review process; what AI/ML can be used for currently; and what is needed in the future to integrate AI/ML into new EM programs.
- GEMS Steering Committee members – Andrew Clayton, Claire van der Geest, Esther Wozniak, Eugene Pangelinan, Gerald Leape, Mark Zimring, Papa Kebe, Robert Gillett, Ruth Hoban.
The Pew Charitable Trusts provided funding for this project, but Pew is not responsible for errors in this paper and does not necessarily endorse its findings or conclusions.