Facial recognition for sheep. Image from Telling (2018).
Precision livestock farming (PLF) is a reality for many farmed animals today, and its use is likely to accelerate in coming years. Systems that are already being rolled out in factory farms track many aspects of farmed animals’ health, such as feed consumption, weight, lameness, and disease prevalence. This data is then processed through machine learning algorithms to identify patterns, enabling producers to more efficiently manage livestock.
Will PLF usher in a new era of welfare transparency across animal agriculture (“precision welfare”), or are we witnessing the takeoff of never-before-seen levels of fully automated animal exploitation?
🧩 Central questions
How PLF works: How do sensor technology and machine learning come together in PLF?
Defining and measuring welfare: How is the welfare of farmed animals defined in the context of PLF, and how does this relate to the data that is collected?
Risks/Benefits: How might PLF lead to improvements in farmed animal welfare, and how could it lead to greater suffering and exploitation?
Welfare vs. profit: Is there a tension between farmed animal welfare and profit? Could PLF be leveraged to optimise for both?
Steering: What are the most effective intervention points – from the initial data infrastructure and algorithm design to industry regulation – for ensuring PLF systems genuinely prioritize the interests of farmed animals?
🧭 Learning objectives
Understand: Explain how PLF works, and identify major benefits (e.g. improved transparency and accountability) and risks (proxy drift, welfare washing).
Assess: Anticipate potential tensions between animal welfare and economic efficiency, drawing connections to outcomes for animals. Compare and contrast different conceptions of farmed animal welfare, highlighting assumptions and risks.
Reason: Formulate evidence-based arguments about whether PLF is likely to be net positive or negative for farmed animals.
Next Steps: Identify strategic recommendations and key intervention points for steering PLF toward animal-positive outcomes.
💡
Use the table of contents on the right to quickly navigate this page.
Resources
‣
Required readings
Please review all of these resources prior to your session.
Estimated time: 45m
These first readings provide a basic introduction to PLF technologies with a view to the stakes for farmed animals and how animal advocates can get involved to advance welfare-positive outcomes.
Walter Veit, Amber Elise Sheldon, and Virginie Simoneau-Gilbert | 31 min panel discussion (0:00-31:09 only)
This expert panel juxtaposes optimistic and pessimistic perspectives on PLF's overall impact on farmed animals, concluding with principles for welfare-centric development and deployment.
These ideas are explored further in Virginie Simoneau-Gilbert’s 2024 blog post coauthored with Jonathan Birch) and Walter Veit’s 2023 Princeton lecture.
This short post makes the case that advocates shouldn’t wait for the industry to build welfare-positive tech, but should proactively design and build the "precision welfare" systems they want to see.
This talk compares two hypothetical PLF systems – one profit-driven, the other welfare-driven – to reveal surprising ways welfare and economic incentives can align. It builds on Boddy’s call to action by offering a design philosophy for precision welfare tech and a strategic path forward from this session’s central debate.
Benjamin Hilton (2024) | 10 min read; 16min audio available (52:11-1:08:14)
This excerpt from the 80,000 Hours problem profile on factory farming contextualises PLF among other key possible future developments, including advances in alternative proteins and the advent of transformative AI.
‣
Additional readings (please read at least 1)
These resources provide a much closer examination of what PLF technologies actually look like on factory farms and what they enable producers to do.
An in-depth case study of AI4Animals, a PLF system used in 20 European slaughterhouses. Morales explains this collaboration between Deloitte and welfare orgs, detailing the incentives for adoption, the metrics dashboard, and how AI insights have led to protocol changes.
This podcast episode surveys the potential of data-driven welfare enhancements, with a special focus on AI insights into farmed animal behavior and a discussion on data ownership.
When assessing the stakes involved in PLF, it’s important to consider how industrial animal agriculture might scale, at least in the near-term future. This report published by Rethink Priorities projects forecasts trends in animal farming over the next 10 years. Access the full report here.
Aaron Boddy (2025) | 2 min read; 4 min audio available
This short memo argues that advocates often miss key opportunities by dismissing industry initiatives (like McDonald's broiler commitments) as mere "humanewashing". Having a "seat at the table" allows advocates to influence corporate metrics from within – potentially shaping industry standards more effectively than external pressure alone.
Daniel Foy & Dr. Jim Reynolds (2025) | 53 min podcast
This podcast episode provides crucial historical context, tracing the origins of US farmed animal welfare audits from early, less scientific assessments to today's multi-metric systems. It highlights the pivotal role of early welfare audit programs (like Validus) and advocates in establishing standards, while looking ahead to how real-time AI monitoring could transform future auditing beyond annual checks.
Walter Veit contends that livestock producers have a moral obligation to adopt PLF, arguing it uniquely improves welfare (through individualized, real-time monitoring) often without sacrificing productivity. He directly tackles skepticism regarding efficacy, cost, and the risk of exploitation, positioning PLF as a powerful tool aligned with both ethical and economic interests.
This blog post argues PLF's benefits for individual animals are likely outweighed by the possibility that lower costs could enable more animals to be farmed, resulting in greater total suffering. The author therefore proposes advocating for restrictions on industrial PLF as a key intervention, suggesting political feasibility due to potential coalitions and public skepticism.
Frank Tuyttens, Carla Molento, & Said Benaissa (2023)
Building on Coghlan & Parker (2023), this academic article identifies a range of both direct and indirect threats to animal welfare from PLF (e.g. injuries from wearable sensors; farmers losing husbandry skills).
This talk highlights AI's potential for early disease detection and discerning animals’ preferences while also cautioning against hype, emphasizing the immense practical challenges (quality training data, setup costs, farmer skepticism, the gap between lab and farm).
Virginie Simoneau-Gilbert & Jonathan Birch (2024) | 11 min read
This article proposes four concrete principles for advocates and regulators to help steer PLF technologies for the benefit of farmed animals: (1) preventing AI from being used to justify higher stocking densities, (2) mandating public transparency for welfare data, (3) ensuring accountability for unaddressed issues, and (4) protecting farmer autonomy.
This short series of blog posts explores the current state and near-future trajectory of AI in animal agriculture, providing further context to Simoneau-Gilbert and Birch’s four “Goldilocks principles”.
Lewis Bollard and Dwarkesh (2025) | 6 min video (9:55-16:00 only) ; transcript available
In this interview segment, Lewis Bollard probes the apparent tension between economic incentives and welfare in animal agriculture, highlighting the need for coordinated efforts across policy and technological innovation.
This paper reviews how PLF uses sensors for individual animal monitoring, primarily targeting production efficiency but already yielding welfare co-benefits (e.g. early disease detection). While acknowledging PLF's potential, the authors caution that current systems lack the integration needed for a truly comprehensive welfare assessment – ending with a call for future systems specifically designed for holistic, multi-dimensional welfare evaluation.
This meta-analysis reviews 50 years of PLF research, identifying research hotspots like precision dairy, AI/ML applications, and behavior monitoring (lameness, estrus detection). The report notes animal welfare as a key research theme alongside social, environmental, and IT aspects.
A major EU-funded project aiming to evaluate and improve broiler and pig welfare across Europe through the use of novel sensor technologies and AI to track animal-based welfare indicators at slaughterhouses, providing real-time feedback to farmers and other stakeholders. This serves as a large-scale, real-world example of developing PLF systems with an explicit focus on welfare monitoring and accountability.
Pre-session exercises
Please spend 20-30 minutes completing the following three exercises.
You can write your responses in bullet point format if that’s easier.
Submit your responses in the weekly Slack thread created by your facilitator in your channel at least 24 hours before your regularly scheduled meeting.
Leave at least one comment on somebody else’s response.
What is PLF?
“PLF is a bundle of many different technologies… Some PLF systems are plausibly good for animals and some are likely bad.” (Simoneau-Gilbert & Birch 2024)
[125 words] In your own words, briefly define precision livestock farming, making sure to connect these three key elements in your response:
What is the role of sensors and data collection?
What kinds of data are being collected?
Machine learning is used to turn this data into useful insights. How does this help producers?
Defining and measuring welfare
PLF: a pessimist perspective
[125 words] How exactly welfare is defined & measured by PLF systems makes all the difference. How might PLF worsen farmed animal suffering if these definitions don’t actually correspond with farmed animals’ interests?
PLF: an optimist perspective
[125 words] How can we transform PLF from a tool for producers into a tool for public accountability and welfare? What would advocates need to demand in terms of technology design and industry regulation?
