How we use AI

AI does not currently interact directly with visitors

Generative AI is currently used behind the scenes during development and maintenance.

Blue Biome does not currently provide visitors with an AI chatbot or other generative-AI system that independently produces answers, recommendations or decisions while they use the website.

Interactive games, mathematical simulations and other computational tools should therefore not automatically be interpreted as generative AI.

Where we use AI

AI-assisted workflows may be used for activities including:

• website development, programming and maintenance
• drafting, editing and structuring written material
• scientific information synthesis and quality checking
• translation and multilingual publishing
• development of educational resources
• creation and editing of illustrations
• animation and multimedia production
• organization and development of the Knowledge Graph
• simplification of scientific concepts for Scenario Lab activities
• game development and supporting materials
• communication and outreach.

The exact tools and models used may change as technology develops and as we improve our workflows.

Human oversight and review

AI-generated material is not published directly to Blue Biome without human involvement.

Changes to the website and its content receive at least a human editorial check and approval before deployment.

However, human approval should not be interpreted as meaning that every resource has undergone the same type or depth of specialist review.

Different resources may undergo different forms of assessment, including:

Editorial review Checking relevance, clarity, consistency and obvious problems before publication.

Scientific review Checking scientific statements, relationships, values or supporting evidence.

Educational review and testing Assessing materials with educators, schools, learners or other intended users.

Ongoing review Some resources continue to be refined as new feedback, evidence or specialist input becomes available.

The way a resource was produced and the extent to which it has been reviewed are therefore treated as separate questions.

Scientific information

AI outputs are not treated as scientific evidence in themselves.

Where Blue Biome presents scientific claims, we aim to base them on identifiable scientific literature, authoritative datasets, established scientific knowledge or documented expert interpretation.

AI may assist with finding, organizing, summarizing or checking information, but it does not replace the underlying evidence.

Species and ecoregion information is reviewed by the Blue Biome team before publication. This review is part of our internal quality process and should not be interpreted as formal external scientific certification.

Scientific content may continue to be corrected, expanded or improved as Blue Biome develops.

Knowledge Graph

AI has been used to support the technical organization and implementation of parts of the Blue Biome Knowledge Graph.

The graph is built primarily from ecological and trophic relationships already documented and reviewed within Blue Biome's species content. AI and automated processing help structure and connect this existing information within the graph. Some relationships may also be inferred or simplified for educational purposes and are identified accordingly where relevant.

Automated processing can nevertheless introduce errors, omissions or unintended relationships. We therefore continue to inspect and improve the graph, its relationships and their supporting evidence.

Scenario Lab

AI is used during development of the Scenario Lab, particularly to help explore ways of translating complex scientific processes and mathematical concepts into lighter educational simulations.

The primary objective of Scenario Lab is learning, exploration and systems thinking, rather than quantitative prediction.

Its activities are designed to help users explore questions such as:

• Which processes influence a marine ecosystem?
• How can pressures interact?
• What may happen when one component changes?
• What types of ecological consequences can follow?

Scientific processes may therefore be deliberately simplified to make them understandable, interactive and suitable for educational use.

Unless explicitly stated otherwise, Scenario Lab outputs should not be interpreted as scientific forecasts, operational models or quantitative predictions for real-world decision-making.

Translation

English is the primary working language for Blue Biome content.

To make Blue Biome accessible in multiple languages with the resources currently available to the initiative, we use automated translation workflows combining different language models and machine-translation systems.

Models and workflows may differ between languages and tasks. We favor locally operated and open-weight models where appropriate, while also using specialized translation tools when they provide useful results.

Current and recent workflows have included models from the TranslateGemma family, specialized machine-translation models such as Helsinki-NLP OPUS-MT, and services such as DeepL.

Translation is not treated as a one-off automated process.

We progressively refine our translation workflows, compare different approaches and carry out targeted revision passes, particularly for scientific terminology, ecological concepts and technical language.

Given the volume of Blue Biome content and the resources available to a small nonprofit initiative, complete human linguistic review of every translated page is not currently possible.

Translated versions therefore display a notice explaining that automated translation has been used and that inaccuracies may remain.

Our approach is one of continuous improvement through better models, language-specific workflows, targeted checks and feedback from speakers and users of the different languages.

Illustrations and multimedia

AI is also used as part of Blue Biome's visual-production workflow.

Some educational species illustrations are created using biological reference material to help reproduce recognizable characteristics while producing new artistic representations.

AI may also assist with:

• image generation
• editing and enhancement
• background removal and upscaling
• animation
• transformation of existing Blue Biome visual material.

These materials are primarily created for education, communication and engagement.

Unless specifically identified otherwise, artistic illustrations should not be interpreted as diagnostic scientific imagery or as substitutes for photographs, taxonomic identification material or other specialist scientific resources.

Where appropriate, Blue Biome aims to make the use of AI in visual material transparent through metadata, labels or resource information.

Transparency at resource level

Blue Biome is progressively developing simple metadata and status labels to help users understand how a resource was produced and the type and extent of review it has received.

The examples below illustrate the labels; they do not certify the review status of this page or of other resources. Labels are added where supporting information is available. A missing label does not establish how a resource was made or whether it has been reviewed.

Educational simplification describes how scientific ideas are represented. It is separate from review status and does not imply scientific validation.

The purpose of these labels is not to classify AI-assisted content as inherently more or less trustworthy. They help educators, learners and other users make informed decisions about how individual resources should be interpreted and used.

Data used with AI tools

AI-assisted development primarily uses publicly available information, public scientific data and Blue Biome working material.

Blue Biome does not intentionally use sensitive personal information as input to its AI-assisted development workflows.

Because generative AI does not currently operate directly during ordinary visitor interaction with the website, browsing Blue Biome does not in itself require visitors to provide content to a generative-AI system.

Information concerning website analytics, forms, cookies and personal-data processing is addressed separately in our privacy documentation.

Read our privacy policy

Errors, corrections and feedback

AI-assisted workflows can make mistakes, just as human-created workflows can.

Blue Biome is an evolving educational initiative, and we welcome feedback that helps us improve its resources.

Users can contact us regarding:

• scientific inaccuracies
• incorrect ecological relationships
• translation problems
• educational content
• website bugs
• attribution or rights concerns
• other errors or questions.

Contact: info@bluebiome.no

Where an issue is identified, we can review the relevant material and make corrections where appropriate.

Our approach

Blue Biome uses AI extensively because these technologies allow a small nonprofit initiative to develop educational, scientific, multilingual, technical and creative resources that would otherwise require substantially greater resources.

We do not consider transparency to mean avoiding AI.

For us, responsible use means being clear about where AI contributes, maintaining human oversight, distinguishing generated material from supporting evidence, making important limitations visible, improving our processes over time and correcting mistakes when they are identified.

AI can support the work. Responsibility for Blue Biome remains human.

This page describes Blue Biome's current approach to AI-assisted development. Our tools and workflows will continue to evolve, and this information may be updated accordingly.