Date Range
Sort by Relevant

GenAI Guidelines for Environmental Journalism

These guiding principles and editorial standards, outlined for Internews’ Earth Journalism Network (EJN) staff, fellows and partner organizations, are intended to support responsible, ethical, and transparent use of generative AI technologies. EJN is dedicated to supporting the production and dissemination of high-quality environmental journalism—journalism that is well-informed, accurate, impartial, and maintains integrity while addressing the complexities of climate and environmental issues. The thoughtful application of AI tools may aid in achieving these objectives, enhancing the capacity to report with depth and precision. It also presents numerous risks to the craft of journalism and to journalists, not to mention to climate and the environment due to the energy and water requirements implied in its use.

Understanding Generative AI

Generative AI (genAI) is a type of artificial intelligence that uses neural networks, deep learning models, complex algorithms, and large and varied training datasets to produce synthetic content based on user input. The content that generative AI creates includes written text, images, video, audio, computer code and more. In the realm of environmental journalism, generative AI holds significant promise by providing analytical insights, assisting with complex data interpretation, creating compelling visual representations, enriching narrative storytelling, and enhancing accuracy through fact-checking, as exemplified by EJN’s soon-to-be-launched genAI tool based on the ChatGPT large language model (LLM), EarthCheckr.

GenAI synthesizes new content such as text, images, video or audio, while non-generative AI, such as machine learning techniques, analyzes existing data to identify patterns and insights. This document pertains to the former.

Despite its potential, the use of genAI comes with inherent risks that must be vigilantly managed. Key amongst these are the potential for plagiarism, the perpetuation of existing biases present in the training data, hallucinations and false information and the production or amplification of mis/disinformation, all of which can seriously compromise the trust and reliability of journalistic output. Recognizing and addressing these risks is crucial to maintaining the integrity of our reporting. At EJN, we are dedicated to upholding transparency and the development of strategies to mitigate these risks, ensuring that our grantees’ and partners’ use of genAI adheres strictly to our ethical standards and responsibilities in journalism.

In summary: When genAI technology is used to generate, synthesize, summarize, transcribe, translate, or paraphrase content, full disclosure of such is strongly encouraged. As a matter of principle, EJN urges its partners not to publish any content that has not been robustly edited and fact-checked by an EJN mentor and available editorial support, as well as the journalist’s newsroom editor and affiliated support staff.

Primary Principles for Responsible Use

1. Accuracy and verification

In environmental journalism, where accurate data is crucial, any content generated or assisted by genAI must undergo the same stringent verification procedures and checks. This includes detailed fact-checking and proper sourcing to prevent plagiarism and ensure authenticity, especially when handling scientific data. By adhering to these rigorous standards, EJN and its partners can ensure that AI-supported reporting is both innovative and trustworthy.

2. Transparency and disclosure

It’s important to clearly disclose the use of genAI in content creation, from grant applications to published stories (provided this does not conflict with the publishing outlet’s editorial policies). Audiences should be informed when AI has been used to generate content. An example of this is a note (see below) that explains how genAI was used to assist with the development of one of EJN’s online courses: 

Editors note on AI

EJN seeks to abide by the best practices prescribed by the Poynter Institute and Trusting News, both of which highlight the need to maximize transparency surrounding synthetically produced content, which includes genAI.

3. Bias and fairness

Recognizing and addressing the inherent biases in AI algorithms is essential for maintaining the fairness and accuracy of AI-assisted content. These biases can arise from skewed datasets or the preferences encoded by their human developers, potentially leading to content that inadvertently perpetuates stereotypes or propagates mis- and disinformation. We aim to foster fairness by actively detecting and correcting biased content. This can be achieved through several strategic steps, from pre-processing data to ensure diversity and representativeness, to post-processing decisions to correct biased outputs. This paper, titled Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies, provides a deeper look into incorporating a holistic approach for mitigating bias in generative AI, involving diverse data collection, transparency, and continuous monitoring.

4. Further ethical considerations

A journalist’s use of genAI in journalism must address issues such as lack of attribution and potential copyright violations. AI-generated content should be transparently disclosed and rigorously verified to avoid misleading information and to maintain trust. Journalists must ensure that AI tools do not oversimplify complex issues or present speculative data as facts, especially in critical areas of their environmental reporting. AI-generated content should be used judiciously, with a strong editorial justification for its inclusion.
 

Editorial Standards for genAI Integration

1. Sourcing and attribution

When employing genAI for data collection or analysis, it is imperative that primary sources are clearly attributed to maintain the integrity of the information. Additionally, any potential limitations or uncertainties inherent in AI-generated analysis must be transparently communicated to the audience. This transparency helps maintain the credibility of the reporting and allows audiences to better understand the context and reliability of the information presented.

2. Editorial oversight

Journalists who engineer genAI outputs must make their mentors and editors aware of all uses of this content throughout their reporting, such as employing LLMs for initial research or a summarization tool for interviews, while subsequently verifying all data and quotes. It is essential that final fact checks and editorial decisions involve a human editor to ensure genAI tools complement but do not replace human judgment. This “human in the loop" approach safeguards against errors that genAI might not detect and ensures the content aligns with journalistic values.

3. Legal and ethical compliance

All EJN staff and partners are expected to adhere to legal standards and ethical norms as defined by national and international law, as well as professional journalism guidelines. This adherence includes respecting intellectual property rights and privacy laws, which are dictated by both local legal frameworks and global agreements, depending on the context. Additionally, the use of genAI should align with the ethical standards set by prominent journalism and media ethics bodies, such as those outlined by the Society of Professional Journalists (SPJ) and the International Federation of Journalists (IFJ). These bodies provide frameworks that help ensure AI applications respect the dignity and rights of individuals and communities, maintaining the public's trust and upholding the standards of responsible journalism.

4. Transparency

EJN adheres to the principles outlined in the Paris Charter on AI and Journalism, emphasizing transparency in AI use and clear distinctions between authentic and synthetic content. Media outlets must disclose significant genAI impacts on journalistic content and maintain public records of AI systems used, while ensuring AI-generated content does not mislead the public or mimic physical capture of the real world (e.g., photos, audio, videos, etc.).

In summary: EJN is bound by our parent organization Internews’ guidelines on AI, “Internews and AI”. We also abide by the Associated Press’ Standards around generative AI and the Poynter Institute’s, and Trusting News’ guidelines on the use of AI in our grantees’ and partners’ newsrooms. This includes:

  • Treating AI-generated content as unvetted material, applying editorial judgment and sourcing standards.
  • Not using generative AI to add or subtract elements or other­­wise significantly alter original text, photos, videos, audio or any other forms of content.
  • Clearly labelling AI-generated illustrations or artworks in news stories.
  • Avoiding putting confidential or sensitive information into AI tools.
  • Exercising caution to ensure material from external sources is free of unverified AI-generated content.
  • Employing standard journalistic practices like reverse image searches to verify the authenticity of content.

EJN has also developed specific policies for the use of genAI in applications to its grants, fellowships, workshops and trainings and other competitive opportunities: “Applicants are required to be transparent about the use of generative AI tools, if any, to revise their proposals. EJN reserves the right to disqualify applicants from consideration if they have been found to have engaged in unethical or improper professional conduct, including, but not limited to, submitting AI-generated content as their own.”

EJN maintains that journalists and media organizations seeking to use genAI should do so responsibly, prioritizing transparency, human oversight and adherence to journalistic ethics. It is important to note that genAI tools are developing rapidly and that recommended guidelines are continuously evolving as journalists and newsrooms adapt to these changes, which require ongoing learning to ensure the continued ethical usage of these tools.

GenAI technologies, if used ethically, have the potential to significantly enhance environmental journalism. By adhering to the guiding principles and editorial standards outlined in this document, EJN and its partners can leverage these technologies to inform, educate, and equip the public with the information it needs to demand action on environmental issues, while upholding high standards of journalistic integrity and audience trust in the media.