Workshops

Non-human animal (hereafter: animal) behaviour and communication research generates enormous quantities of video and audio data and analysing it manually is labour-intensive and hard to scale. EthoInsights is a non-profit tool that brings together existing open-source AI models into a single, user-friendly interface designed specifically for animal behaviour researchers. Researchers can upload their footage, describe what they are looking for in plain language, and receive structured results ready for further analysis. The tool does not interpret the meaning of animal behaviour or vocalisations; it supports researchers in handling their data and identifying patterns that inform their own analyses.

Using the tool as a case study, this workshop invites participants to discuss the values that should guide tools of this kind, the ways they could be misused, and how the resulting risks might be prevented through design choices and governance.

This workshop is aimed at animal and AI ethicists, and others interested in the ethical discussions surrounding AI in animal behaviour research. No specific background is required.

The workshop starts with a short demonstration of EthoInsights and a brief introduction to AI for animal behaviour research and the ethical questions it raises.

Participants then take part in a world café. Three tables are set up, each with one question:

1. What values should guide the development of the tool;
2. How the technology could be misused, and what unintended harms or side effects it could produce;
3. How can risks be prevented through design and governance?

Each group discusses all three topics in turn, moving between tables at set intervals. A host stays at each table to introduce the topic, document the discussion, and pass on what earlier groups contributed.

Each table host documents the discussion and the recommendations that emerge. The workshop will identify key ethical tensions surrounding AI for animal behaviour research, and the report afterwards will provide concrete feedback for the future development of EthoInsights. Where appropriate, recommendations may be incorporated into the tool’s governance, strategy, or ethical framework.

Dominik Hajosi (Messerli Research Institute, Department of Interdisciplinary Life Sciences, University of Veterinary Medicine Vienna, Medical University of Vienna, University of Vienna, Vienna, Austria and Institute of Comparative Medicine, Columbia University, New York, NY, United States) –

dh2906@cumc.columbia.edu

Herwig Grimm (Messerli Research Institute, Department of Interdisciplinary Life Sciences, University of Veterinary Medicine Vienna, Medical University of Vienna, University of Vienna, Vienna, Austria) –

herwig.grimm@vetmeduni.ac.at

This workshop explores how harm–benefit analysis (HBA) can be operationalized in practice using the Compound Model (CM), a structured framework integrating key elements from existing guidance. Participants will examine conceptual challenges, apply the model to case examples, and discuss its potential to support transparent and consistent ethical review.

The workshop is intended for members of ethics committees, animal welfare bodies, competent authorities, IACUCs, researchers, project applicants, laboratory animal veterinarians, and scholars interested in animal research ethics, governance, ethical review, and decision-making. It is particularly relevant for individuals involved in project evaluation and review under Directive 2010/63/EU. No prior experience with the Compound Model is required.

The workshop will combine a brief introduction to the challenges of HBA with an interactive presentation of the CM. Participants will apply the model to practical case studies, either individually or in small groups, and compare their assessments in a facilitated discussion. The session will focus on identifying key decision points, evaluating which components from existing guidance are essential for HBA, exploring sources of disagreement and differences, and evaluating how a structured approach may improve consistency and transparency in the HBA’s application during the project evaluation process.

Participants will gain familiarity with the CM and its practical application in HBA. They will develop a better understanding of common challenges in the evaluation and decision-making process, learn how to systematically evaluate individual HBA components, and critically reflect on the strengths and limitations of structured HBA approaches, including their normative foundations. The workshop aims to advance understanding of and promote discussion on how HBA can be implemented more consistently and transparently across regulatory and institutional contexts.

Hajosi D, Grimm H (2025) Solving the riddle by puzzling it out: overcoming the challenges of harm–benefit analysis by assembling a compound model and practice-oriented tool. Front. Vet. Sci.,

doi.org/10.3389/fvets.2025.1593954

Hajosi D, Grimm H (2024) Mission impossible accomplished? A European cross-national comparative study on the integration of the harm-benefit analysis into law and policy documents. PLoS ONE,

doi.org/10.1371/journal.pone.0297375

Franck L.B. Meijboom (Utrecht University) –

f.l.b.meijboom@uu.nl

This workshop is designed to help early-career researchers navigate the academic publishing process more strategically and confidently. It will explore key considerations in developing a publication strategy, from identifying appropriate journals and positioning a manuscript for its intended audience to understanding editorial expectations and the peer-review process.

Drawing on practical experience in academic publishing, the workshop will also address common challenges faced by early-career researchers, including responding to reviewers, dealing with revisions and rejection, and making informed decisions throughout the publication process. Participants will have the opportunity to reflect on their own publishing practices, discuss questions and challenges, and gain practical insights for developing a more effective and sustainable approach to academic publishing.

In just a few years, Large Language Models (like ChatGPT and Claude) have become part of our daily lives. The influence of LLMs reaches beyond the individual: they increasingly influence our knowledge, ways of thinking and decision making. For example, AI has the tendency to discriminate against women, people of colour, and other minorities (Hagendorff et al. 2023). Efforts are now being made to tweak the algorithms in order to prevent racist and sexist outcomes. However, discrimination against animals has so far stayed under the radar.

This is problematic, because the datasets on which LLMs are trained show speciesist patterns and ‘AI technologies currently play a significant role in perpetuating and normalizing violence against animals’ (Hagendorff et al. 2023, 717). This speciesist bias can be discerned in language and images produced by algorithms.

For example, objectifying language (such as using the word ‘it’) is often used when referring to animals, and the first images generated by searches for pictures of animals such as chickens are cooked chickens, or recipes, conveying the message that the main purpose of these animals is to be eaten. Because of this, it is important that LLMs generate information that does not violate ethical principles or reproduce harmful ideas about animals.

In this workshop we will explore the practical and conceptual challenges of developing a benchmark test that evaluates speciesist biases that result in animal harm in LLMs. We will discuss a particular LLM that has been designed to produce non-speciesist outcomes. In the first half of this workshop, we will discuss the practical steps of the development and challenges that are involved in the design of the benchmark: how do we approach adequate animal representation, ethical theory in practice and tradeoffs between feasibility, potential impact and ethical principles?

In the second part of this workshop, the participants will create and score scenarios, based on the award winning paper at ACL2024, the Machiavelli framework. By the end of the workshop, the participants have contributed to the finetuning of the benchmark, have discussed ethical and practical considerations for designing such benchmarks and gained insight into the importance and challenges of ethically responsible and fair AI.

Hagendorff, T., Bossert, L.N., Tse, Y.F. and Singer, P. (2023). Speciesist bias in AI: How AI applications perpetuate discrimination and unfair outcomes against animals. AI and Ethics 3, 717-734.

https://doi.org/10.1007/s43681-022-00199-9

This workshop explores gender and broader intersectionality gaps in AI-driven agrifood systems. Participants critically examine AI use cases, identify overlooked impacts on diverse groups, reflect on biases, and formulate research gaps to support more inclusive and responsible AI innovation in agriculture and food systems.

The workshop targets professionals working at the intersection of AI and agrifood, including researchers, policymakers, practitioners, and civil society representatives. Participants may have a background in ELSA of AI communities, digital agriculture initiatives, and ethical and social impacts of agriculture and food.

Using a structured 2 hour co-learning format, participants engage in positionality reflection, small-group case analyses, intersectionality stress testing, and collaborative research-gap identification. Discussions focus on AI agrifood applications such as advisory systems, autonomous machinery, and farm management tools. Facilitated plenary reflections promote awareness of biases and systemic blind spots.

Participants will gain greater awareness of gender and intersectionality dimensions in AI for agrifood, identify critical evidence gaps, and reflect on their own assumptions and research practices. The workshop will generate a set of research-gap statements and foster dialogue on how future AI and agrifood research can become more inclusive, equitable, and socially responsible.