The mesopelagic zone, a vast and largely unexplored ecosystem, has become a focus of scientific and societal interest due to its potential for resource extraction and climate regulation. Scientific investigation of this zone dates back to the Challenger expedition in the 19th century and has intensified recently, fueled by funding from the European Union and the United States. Scientists are actively working to quantify the mesopelagic zone's role in atmospheric carbon sequestration (estimated at 2–6 billion metric tons per year) and to assess the potential fish resources (estimates vary widely from 1.8 Gt to 19.5 Gt). This scientific work is crucial for decision-making, as it renders the mesopelagic zone 'legible' for governance. Science influences both *de jure* governance (through formal policies and regulations) and *de facto* governance (by shaping how the ecosystem is understood and valued, influencing policy debates and shaping anticipatory knowledge production). This study investigates the interplay between scientific research, public discourse, and the absence of formal governance in shaping the future of the mesopelagic zone.

The authors draw upon existing literature on the mesopelagic zone's biodiversity, its role in carbon cycles, and its potential as a fishing ground. They reference various studies that provide estimates for the mesopelagic zone's carbon sequestration capacity and the potential biomass of fish stocks. The review highlights the wide range of uncertainty in these estimations and the need for further research to better understand the ecosystem's functions and dynamics. The literature also informs the conceptual framework of the study, which draws on theories of network agenda-setting, framing, and the role of science in shaping governance of natural resources.

The study uses a computational methodology based on automated content analysis to investigate the scientific and public discourse on the mesopelagic zone. Data was collected from two sources: 2226 scientific abstracts from Web of Science (published between 1990 and 2020) and 4066 tweets (containing the search terms "mesopelagic" or "ocean twilight zone", or their hashtags, from 2009 to 2020). A Latent Dirichlet Allocation (LDA) topic model was employed to identify 13 key topics within the combined corpus. The LDA model parameters were carefully selected and validated using coherence metrics and manual topic intrusion tests. The frequency of co-occurrence of these topics was analyzed to generate a "network of meaning", visualizing the relationships between dominant ideas. Additionally, the authors conducted a policy analysis by searching for keywords from the topic model in 15 key international and national policy documents. This comparison highlights the disconnect between scientific and public discussions about the mesopelagic zone and the existing governance frameworks.

The analysis revealed two dominant clusters of topics within the network of meaning: one focused on fish (including their diets, reproduction, population estimates, behavior, and bioluminescence) and the other on carbon (emphasizing particulate and dissolved organic carbon). Several other topics addressed specific aspects of mesopelagic biology and ecology (e.g., zooplankton distribution, microbial diversity, stable isotope analysis). Twitter primarily served as a platform for scientists to communicate their research findings to the public and among themselves. The comparison with policy documents showed a stark contrast. The mesopelagic zone was largely absent from the key governing policy documents, while the scientific and social media discussions were rich with information and various perspectives on the potential uses of this ecosystem. The keyword "carbon" was mentioned only once across all policy documents, highlighting that carbon sequestration, a major topic in scientific discourse, is not yet a focus of ocean policy. The most prevalent keywords in the policy documents were "fisheries" and "fish", reflecting the existing focus on managing fish stocks. The authors conclude that scientific knowledge is shaping anticipation and governance of the mesopelagic zone far ahead of the development of *de jure* policies and regulations.

The study's findings demonstrate that, in the absence of formal governance, scientific research significantly shapes how the mesopelagic zone is perceived and potentially managed. The focus on quantifying fish populations and their behavior, essential for fisheries management, illustrates the anticipatory role of science in shaping the future governance of the ecosystem. The development of a "blue carbon" narrative around the mesopelagic zone's carbon sequestration capacity also highlights how scientific work can shape the framing of this ecosystem's functions and potential value. The absence of counter-narratives on Twitter suggests a virtual monopoly on knowledge held by scientists, due to high access barriers. However, the authors note the absence of economic and cultural considerations (like deep-sea mining impacts) in both scientific and policy discussions, pointing to a bias toward fisheries and carbon. The lack of consideration for economic and cultural aspects of deep sea governance highlights a gap in current discourse.

The study emphasizes the de facto governing role of science in the mesopelagic zone due to the lack of comprehensive policy. It suggests that broadening the range of perspectives through transdisciplinary approaches involving diverse stakeholders (including industries, policymakers, NGOs, and the public) is crucial for more equitable and sustainable governance. The authors propose three potential routes for enhancing democratization of knowledge: improving transparency and public engagement in existing institutions (like RFMOs and the ISA), utilizing open digital platforms for data sharing and collaboration, and integrating diverse data sources into a unified global ocean governance infrastructure. Future research should focus on expanding the scope of analysis to include grey literature, popular media, and advocacy-based science.

The study acknowledges limitations in the topic model's inability to account for synonyms, negations, and word order. Future iterations may incorporate advanced techniques that can address these issues and provide more nuanced analyses of text data. Additionally, the policy analysis was limited in scope, focusing on a purposive sample of international and national documents. A more comprehensive analysis including national-level policies in multiple languages could provide a richer understanding of existing *de jure* governance frameworks.