Our pledge
We as members, contributors, and leaders of the Ecological Resilience and Forecasting Lab (EcoReF Lab) pledge to uphold the principles of respect and tolerance towards every human being. We will act and interact in ways that contribute to an open, welcoming, diverse, inclusive, and healthy team.
Our values and behaviours
Leaders lead by example; supervisees learn by being curious, independent and willing to try
We demonstrate empathy and kindness towards other people
We are tolerant of ourselves and others and accept that happiness and “good humour” are not constants
We respect differing opinions, viewpoints, and experiences
We listen and assume “best of intentions”
We give honest, but constructive and polite feedback
We are grateful of and seek feedback
We accept responsibility of our acts and are not afraid to apologise to those affected by our mistakes
We accept and learn from failure
When conflicts arise, we seek help from a supervisor when direct resolution seems impossible
Our position with regards to the use of Generative AI
Generative AI (GenAI) can be a genuinely helpful tool — particularly for those who are already well versed in the tasks they are doing and can therefore critically assess its outputs. We understand GenAI is “here to stay”, but believe its capabilities should continue to be put at the service of research and humanity, not override human capabilities, control and staunch learning and development.
We are mindful that this requires keeping humans firmly in control, and that the way we use these tools today shapes both the quality of our science, the skills of the next generation of researchers, and long-term collective human intelligence and knowledge-base.
The following principles guide how we use GenAI in the lab:
Humans remain in the driver’s seat. We are cautious with agent-automated or otherwise unsupervised GenAI workflows, and cautious in using or accepting GenAI output as part of research outputs. A human in the lab is always responsible for, and able to defend, any GenAI-assisted work we produce.
We are mindful of cognitive offloading. Convenience can come at a cost: delegating thinking to a model can erode the deep engagement that underpins learning, training and good research (see, e.g., Georgiou 2025, arXiv:2507.00181). This is especially important for students and trainees, where over-reliance on GenAI can undermine the development of core skills. We encourage GenAI use that supports — rather than replaces — our own thinking and learning, and we are particularly deliberate about its use during training and education. This means, in part, that when GenAI proposes the use of a new technique, method, or approach, we take the time to understand it and its implications before accepting it.
Good use cases. Tasks where GenAI tends to be a useful assistant include:
small writing tasks (e.g., improving the flow of a paragraph or sentence, checking spelling and grammar) and coding tasks (e.g., help with fixing a bug or finding a solution to a particular problem, especially in complex projects with many moving parts);
obtaining inspiration for visual representations of research outputs;
flagging (but not necessarily fixing) spelling, grammar and flow issues in a text.
Configuration and sourcing. Where possible, we use settings in GenAI platforms (ChatGPT, Google AI Studio/Gemini, MS Copilot, etc.) that:
limit the model to data sources provided within prompts;
request outputs are always accompanied with sources of information (e.g., citations, locations within provided sources where information can be found).
Verification. We verify that outputs are factually correct, by checking the cited sources of information.
Limitations and risks. We are aware of limitations and biases of GenAI models, potential breaches of intellectual property these models may entail, and of GenAI “hallucinations” (wrong or made-up information, changes to segments that do not need changes, and other unexpected behaviors).
Confidentiality. We do not upload unpublished work or private documents authored by others, unless we have explicit permission to do so.
Authorship of research outputs. We do not generate whole or partial scientific products or outputs (large code components, images, manuscripts, reports, reviews, responses to reviewers, etc.) using GenAI. GenAI may assist the process, but the substance and authorship remain ours.
Asking for help. If and when we find ourselves struggling to avoid using GenAI or feel pressured to use it in situations where it may not be appropriate – e.g., because there is a feeling of time pressure or of the inability to execute tasks without this tool – we reach out for help to colleagues and/or supervisors.