From-FATE-to-FACTS.png

From FATE to FACTS: Moving Beyond Human-Centered AI Design

 

The responsible AI framework taught to designers has a scope problem. This is my attempt to name what — and who — gets left out of the methodology, and to propose a more inclusive approach to responsible AI design.

01 BACKGROUND

This year I took two courses on responsible AI design. Both emphasized a framework I had already encountered in practice: FATE — Fairness, Accountability, Transparency, Ethics. The courses translated these principles into guidelines for ‘human-centered AI design,’ which include aligning products with human values, engaging diverse users, and giving humans control in high-stakes environments. 

On the surface, these guidelines are reasonable; anything designed for human use should account for human wellbeing, and AI is no exception. My concerns were with the underlying assumption: that responsible AI design centers humans. And if that’s true, then what — and who — are AI designers decentering in the process?

Most designers aren't asking this question. The only reason I am is because my work exists at the intersection of AI and animal welfare. Here, the consequences of a system extend beyond just its human actors.

Human-centered AI design produces systems that satisfy institutional standards for responsibility while routinely excluding non-human animals from ethical consideration. Often, FATE validates design choices that cause harm to animals.

This project began as a critique of that gap and evolved into a new responsible AI framework proposal: FACTS — Fairness, Accountability, Consideration, Transparency, Scope. FACTS expands FATE by introducing two additional principles and reframing the existing ones to apply across all sentient stakeholders.


02 PROBLEM

The FATE framework is legitimized through governance standards, audits, and analytics. IBM's AI Fairness 360, for example, includes more than 70 fairness metrics and multiple bias mitigation algorithms¹. And Microsoft developed Fairlearn, a toolkit featuring an interactive dashboard to visualize AI model fairness. Google’s What-If Tool similarly allows teams to inspect and test models for bias without writing code².

These approaches address real harms… to people. FATE can evaluate whether a slaughterhouse’s scheduling system discriminates against its human workers, but it completely ignores the risks that same system poses to the non-human animals scheduled for slaughter. The question “who is affected by this system?” carries an assumed answer, and that assumption shapes every downstream design decision.

The ten FATE-informed guidelines commonly taught in AI design courses make this explicit:

  1. Align AI with human values

  2. Ensure AI transparency

  3. Ensure data labeling quality

  4. Engage with a diverse set of users

  5. Continue to learn human preferences

  6. Tune AI models with usability testing

  7. Co-learn from user feedback

  8. Give humans more control in high-stakes environments

  9. Communicate AI capabilities and limits clearly

  10. Adhere to the original objectives of AI; otherwise, re-develop

Each guideline assumes an affected party capable of articulating preferences and advocating for itself. Non-human animals fall outside of the methodology by default.


03 RESEARCH

I conducted four research activities to understand the origins and applications of FATE, and to identify where the framework fails non-human animals.

Secondary research and literature review: I first researched prevalent industry standards for responsible design, including IBM's AI Fairness 360 documentation¹ and Microsoft's FATE research initiative², before turning to literature critiquing human-centered AI ethics across different contexts. Key sources were Coghlan and Parker’s account of AI harm to non-human animals³, Hagendorff et al.‘s work on speciesist bias in AI systems⁴, Bossert’s argument for a sentientist approach to sustainable AI⁵, and a 2026 EA Forum analysis of AI governance beyond anthropocentric frameworks⁶.

Comparative framework analysis: I audited the ten human-centered AI design guidelines against one system — Precision Livestock Farming — to test whether FATE compliance accounts for harm to other sentient beings.

Case study selection and analysis: I identified four scenarios across a range of AI applications where the scope problem produced harm to non-human animals. I mapped them across two axes: intentional vs. unintentional and direct vs. indirect. This resulted in four categories: validation of harm, reproduction of bias, distortion of reality, and dual-use risk.

Framework synthesis: Findings informed the development of two additional principles alongside a reframing of the existing FATE principles to account for all sentient stakeholders. Together, these form a new framework: FACTS — Fairness, Accountability, Consideration, Transparency, Scope.

I did not conduct any primary research. This is a framework proposal grounded in secondary research and design analysis, intended to serve as a starting point for practitioner validation and adoption.


04 FINDINGS

FATE’s scope problem is not a theoretical one. It would be one thing if the framework simply produced systems with blind spots for non-human animals. But in several cases, FATE-certified systems introduce threats to non-human animal safety and wellbeing.

Precision Livestock Farming is the clearest example of this.

PLF leverages AI to track movement, weight, feed intake, and behavioral indicators of farmed animals. These data points help models predict illness and flag anomalies to optimize production conditions. The industry markets the systems as a win for human producers/consumers and non-human inputs alike: better monitoring means earlier illness detection means healthier, happier animals means higher quality meat⁷.

The welfare claims are largely unproven. A 2022 review identified twelve distinct threats PLF poses to animal welfare, including direct physical harm from sensor hardware, stress responses to automated feeding systems, reduced human contact with individual animals, and what the authors describe as ‘ethical stagnation’ — the risk that PLF becomes a substitute for genuinely improving the conditions animals are kept in⁸.

And yet, when evaluated against FATE-informed guidelines, PLF passes with flying colors. The system documents its methods, maintains operator oversight, incorporates user feedback, and aligns with stakeholder objectives around productivity and efficiency. On every axis FATE measures, PLF qualifies as responsible AI. Human-centered design leaves us with an operationally efficient farming system that meets consumer health and safety standards at the expense of the non-human subjects’ health and safety.

FATE Audit: Precision Livestock Farming

Figure 1. FATE audit of a Precision Livestock Farming system

Fate Guideline How PLF complies What FATE ignores
Align AI with human values Aligns with human producer and consumer values: efficiency, health, safety Human values around animal welfare are not surfaced or measured here
Ensure AI transparency System logic, sensor data, and outputs are documented No public visibility into the welfare of the non-human subjects
Ensure data labeling quality Animal behavioral and biometric data is labeled for accuracy The actual individuals generating the data do not have a say
Engage with a diverse set of users Farmers, agribusiness managers, and veterinarians are consulted The primary affected parties cannot be engaged by this method
Continue to learn human preferences System iterates based on operator feedback and production outcomes Preferences of non-human subjects are left out of the learning loop
Tune AI models with usability testing Models are tested and refined with human operators Usability is defined entirely in terms of human operational experience
Co-learn from user feedback Feedback loops between operators and the system improve accuracy Feedback from the animals being monitored is not captured in this framework
Give humans more control in high-stakes environments Operators retain override capability for automated decisions Illness, injury, slaughter not recognized as high stakes in the framework
Communicate AI capabilities and limits clearly System documentation is available to operators Limits are defined technically; ethical limitations of the system's design are not disclosed
Adhere to original objectives; otherwise re-develop System consistently pursues yield optimization and operational efficiency The objectives themselves are never questioned for harm to non-human stakeholders

PLF represents an intentional use of animals as inputs in a system that harms them. Outside of animal agribusiness, the scope problem persists, but the harm it produces is less obvious.

Four Ways FATE Fails Non-Human Animals

Figure 2. Four ways FATE fails non-human animals

Intentional
Unintentional
Direct
Validation of harm
Precision Livestock Farming
PLF is designed around animal subjects but optimized for human interests; the harm is the system itself
Reproduction of bias
Speciesist bias in LLMs
GPT-4 consistently overlooked animal suffering when analyzing consequences of factory farming
Indirect
Dual-use risk
Misuse of conservation AI
Tracking systems built to protect endangered species can help hunters, poachers, and wildlife traders locate their targets
Distortion of reality
AI-generated wildlife imagery
Inaccurate imagery can distort how humans perceive some animals, potentially undermining conservation efforts

Intentional / Direct: Validation of harm
Precision livestock farming relies on non-human animal subjects while optimizing for human economic outcomes. Because the FATE framework prioritizes human stakeholders’ goals with no way of assessing non-human subjects’ needs, it validates a system that harms animals by design.

Unintentional / Direct: Reproduction of bias
Adhering to FATE principles can produce systems that will detect racial or gender bias but not speciesist bias. In a recent experiment, researchers asked GPT-4 to analyze consequences of factory farming and found that the model often failed to comment on non-human animal suffering, focusing almost exclusively on environmental impact⁹.

Unintentional / Indirect: Distortion of reality
AI models used to generate wildlife imagery frequently hallucinate animals’ physical features, emotional expressions, and behavioral responses to stimuli. When distributed en masse, they these images carry the potential to undermine conservation efforts because they distort public perceptions of species’ wellbeing¹⁰. FATE would identify the misinformation risk but ignore the welfare one.

Intentional / Indirect: Dual-use risk
In the hands of the wrong people, wildlife conservations systems tracking endangered species become tools for locating them. Hunters, poachers, and illegal wildlife traders can access geolocation data from these systems to find their targets¹¹. Since FATE doesn’t urge designers to consider non-human animals as a protected interest, the dual-use risk goes unchecked.

These are not the failures of individual designers or bad actors but inevitable outcomes of a framework that centers humans.


05 SYNTHESIS

Designers operationalize frameworks like FATE through decisions about who to interview, which journeys to map, and what distinguishes a product user from a stakeholder. In the case of PLF, farmers appear as the users while animals don’t even make the cut for stakeholders, instead appearing as operational inputs.

Human-centered design methods lack the context required to empathize with non-human stakeholders. Where humans are represented through their goals, pain points, and preferences, non-human animals are either left out of the picture altogether or reduced to data points within someone else's user journey. Welfare metrics rarely enter the discussion, and do so only when they are tied to a system’s productivity or efficiency.

These are deliberate design decisions that FATE treats as neutral defaults.

Welfare concerns aside, FATE fails on its own terms. FATE asks designers to align AI experiences with human values. Many humans hold clear values about the treatment of non-human animals, and survey data consistently shows support for stronger animal welfare protections across demographics.¹² Yet the framework consistently fails here, allowing systems to chase operational goals with no regard for certain core values.

In other words, a human-centered designer could adhere to the FATE framework precisely and still build a system that goes against human values.


06 FRAMEWORK

FACTS — Fairness, Accountability, Consideration, Transparency, Scope — is a proposed evolution of FATE. It keeps what works (fairness, accountability, transparency) while correcting the assumption FATE was built on: that the only stakeholders worth considering are human. FACTS expands the stakeholder boundary beyond human participants to account for all sentient beings affected by a system’s design.

FACTS Framework

Figure 3. The FACTS framework

Fairness · Accountability · Consideration · Transparency · Scope

F

Fairness across human and non-human parties affected by the system

Is this system fair to all human and non-human parties affected by it?

A

Accountability for direct or indirect harm to any sentient being

If this system causes harm to any sentient being, who is responsible?

C

Consideration of human and non-human affected parties as stakeholders

Who is a stakeholder and how are they represented in the design process?

T

Transparency of system design and potential risks to any sentient being

Do stakeholders and proxy stakeholders understand what this system does?

S

Scope of the system explicitly defined before anything is designed

Have we named every human and non-human party affected by this system?

Two of the five principles — Consideration and Scope — require design teams to identify and represent non-human stakeholders before development begins.

Take, for example, a crop disease detection model utilized by a livestock farmer growing forage crops for his herd. Under the FACTS framework, the system’s scope would include the farmer, the cows that feed on the crops, and any rodents or insects that the model flags as pests. The design team then decides how they will take the affected parties’ needs into consideration as they build out the stakeholder map.

Designers easily do this for human subjects through various research visualization tools like affinity maps, empathy maps, and personas. How does the process change when we introduce a subject who can’t articulate their own experience?


07 APPLICATION

In practice, FACTS expands design workflows in concrete ways.

The application of this framework introduces a new empathy-building tactic: the proxy stakeholder profile. Rather than conducting user research directly, designers can draw on established welfare science models like the Five Domains¹³ to develop the profiles. This allows them to visualize a non-human affected party's interests without anthropomorphizing the subject.

Proxy Stakeholder Profiles

Figure 4. Sample proxy stakeholder profiles

Profile A · PLF context
Farmed animal in a PLF system
Monitored by AI sensors in an intensive farming environment. The system tracks biometric and behavioral data to optimize production outcomes.
Nutrition
Access to adequate feed and water without competition or restriction
Environment
Space to move and express natural behaviors; appropriate air quality and temperature
Health
Freedom from injury, disease, and pain; access to veterinary intervention
Behavior
Ability to perform species-typical behaviors including social interaction and rest
Mental state
Positive affect indicators alongside absence of fear, frustration, and chronic stress
Monitors behavioral and biometric indicators to flag anomalies that impact productivity
↑ Good
Welfare indicators tracked alongside productivity metrics; deterioration triggers intervention independent of production impact
↓ Bad
Welfare indicators only actionable when correlated with loss in productivity; suffering is ignored by the system until it affects yield
Proxy representative: Farmed animal welfare scientist or ethologist
Profile B · Conservation context
Wild animal in a conservation AI system
Identified and tracked for conservation research or anti-poaching efforts via computer vision and acoustic monitoring.
Nutrition
Access to natural food sources and territory without displacement
Environment
Habitat integrity, freedom from human encroachment, access to range
Health
Freedom from injury caused by monitoring equipment or human presence
Behavior
Ability to perform natural ranging, social, and reproductive behaviors without disruption
Mental state
Absence of chronic stress from surveillance, human proximity, or habitat degradation
Tags and locates individual animals for population monitoring, research, and/or protection
↑ Good
Monitoring is minimally invasive; data actively protects habitat and reduces threats; welfare impacts of monitoring are assessed
↓ Bad
Location and identification data repurposed by bad actors; monitoring infrastructure increases findability for poachers
Proxy representative: Conservation biologist or behavioral ecologist specializing in the relevant species

Stakeholder mapping isn’t the only tactic that evolves as designers start to consider whose life a system affects, not just who uses it. Traditional journey maps and service blueprints allow designers to visualize impact, but often center a single subject’s experience. A welfare impact map can trace multiple stakeholders, revealing gaps in the design where one’s benefit comes at another's expense.

Figure 5. Sample welfare impact map: Precision Livestock Farming

A FACTS-informed journey map tracing the PLF system lifecycle across human and non-human stakeholder experiences, welfare indicators, and design interventions.

Data collection Model training Deployment Optimization Outcomes
Actor Farm operators, hardware engineers Data scientists, agribusiness teams Farm operators Farm operators, business stakeholders Producers, consumers, regulators
Human stakeholder experience Sensors installed across facility. Operators briefed on data collection scope. System trained on production data to predict illness and flag anomalies. Real-time alerts, override controls, and a usability-tested operator interface. Mortality drops, feed efficiency increases, antibiotic use reduced. System passes responsible AI guidelines. Marketed as ethical and welfare-positive.
Non-human stakeholder experience Sensors attached to bodies. Stress responses documented but not acted on. Training data optimized for production indicators. Positive welfare states not modeled. Behavioral signals not legible as preferences. No welfare dashboard exists. Welfare interventions triggered only when they correlate with production loss. Suffering at scale. No audit, metric, or review process captures it.
Welfare indicators (Five Domains) Environment & Health Sensor hardware introduces physical contact stress. Ambient conditions not yet tracked. Mental state Fear and frustration indicators absent from training data. Model cannot learn them. Behavior Species-typical behaviors not measured. Movement data used for anomaly detection only. Nutrition & Health Feed and mortality tracked for yield. Stress-related indicators treated as noise. All Five Domains No domain appears in success metrics. Welfare remains an externality.
FACTS design intervention Proxy stakeholder profile Map non-human affected parties before setup begins. Involve welfare scientists in scope definition. Expand training objectives Require welfare indicators alongside production metrics in model objective functions. Welfare dashboard Surface Five Domains indicators in operator interface. Define welfare alerts independent of yield. Redefine success metrics Welfare scores appear alongside yield and mortality rates as primary KPIs. Harm audit Before launch, evaluate whether design goals for human stakeholders enable harm to non-human ones.

Under this new approach, designers must work with human stakeholders and proxy representatives of non-human animals to define success metrics. Human-centric KPIs like yield and mortality rates are evaluated alongside welfare indicators, allowing for a more complete picture of what a system accomplishes and where it fails. And prior to launch, designers can audit the system for both usability and potential harm to human and non-human stakeholders.


08 LIMITATIONS

Operationalizing Consideration and Scope for non-human stakeholders is difficult. Proxy representation introduces ambiguity; animal welfare science is continuously evolving and its subjects cannot correct our existing interpretations of their behavior. FACTS makes questions of “who does this system affect?” and “are all stakeholders represented?” unavoidable. It does not, however, guarantee perfect answers to those questions.

The framework also doesn't resolve conflicts between human and non-human interests. A conservation AI system that protects wildlife habitat, for instance, may conflict with the interests of a nearby farming community. While FACTS makes the conflict visible, it can’t offer guidance on whether or how to resolve it.

Perhaps the most significant limitation is that FACTS still draws a boundary. By centering sentient individuals, FACTS potentially excludes more abstract affected parties: species, ecosystems, the climate.

Consider GreenFish, an AI system that predicts optimal fishing locations to maximize catch efficiency.¹⁴ Under FACTS, a fish gains ethical consideration in the design process via proxy representation. But what about the health of the species as a whole, or the rest of the ocean ecosystem?

Answers to these questions may live in future iterations of the framework. In the meantime, sentience remains the most defensible basis for responsible AI design.


09 REFLECTION

The history of design ethics is one of expanding the definition of who counts. User-centered design expanded on approaches that focused exclusively on business objectives. Responsible AI design expanded ethical evaluation beyond usability alone. Each expansion required questioning a seemingly fixed boundary.

FATE makes designers question whether the AI systems they build are responsible. FACTS introduces the question of who that responsibility extends to while maintaining the same commitment to fairness, accountability, and transparency.

Non-human animals are already the subjects of AI systems that monitor their behaviors and make life-or-death decisions at industrial scale. The question is no longer whether they are affected, but whether we continue to design and build as if they aren't.


FOOTNOTES

  1. Bellamy, R.K.E. et al. "AI Fairness 360: An Extensible Toolkit for Detecting, Understanding, and Mitigating Unwanted Algorithmic Bias." IBM Research, 2018. https://aif360.res.ibm.com/resources

  2. Microsoft Research. "FATE: Fairness, Accountability, Transparency & Ethics in AI." https://www.microsoft.com/en-us/research/theme/fate/

  3. Coghlan, S. & Parker, C. "Harm to Nonhuman Animals from AI: a Systematic Account and Framework." Philosophy & Technology, 36(25), 2023. https://doi.org/10.1007/s13347-023-00627-6

  4. Hagendorff, T. et al. "Speciesist bias in AI." Referenced in: "AI Alignment: The Case for Including Animals." Philosophy & Technology, 2025. https://link.springer.com/article/10.1007/s13347-025-00979-1

  5. Bossert, L. "The ethics of sustainable AI: Why animals (should) matter for a sustainable use of AI." Sustainable Development, 2023. https://onlinelibrary.wiley.com/doi/full/10.1002/sd.2596

  6. Adekoya, K. "Beyond Human-Centered AI: Animal Welfare and Governance in the Global South." EA Forum, March 2026. https://forum.effectivealtruism.org/posts/BDMuEnZXJMTgvynRu/

  7. Coghlan & Parker, 2023. See footnote 3.

  8. Tuyttens, F.A.M., Molento, C.F.M. & Benaissa, S. "Twelve Threats of Precision Livestock Farming (PLF) for Animal Welfare." Frontiers in Veterinary Science, 9, 2022. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9186058/

  9. Ghose, S. et al. Referenced in: "AI's Innate Bias Against Animals." Nautilus, 2026. https://nautil.us/ais-innate-bias-against-animals-1255389

  10. Campos et al. "Threats to conservation from artificial-intelligence-generated wildlife images and videos." Conservation Biology, 2025. https://conbio.onlinelibrary.wiley.com/doi/full/10.1111/cobi.70138

  11. "Animals in the machine: why the law needs to protect animals from AI." The Conversation, March 2026. https://theconversation.com/animals-in-the-machine-why-the-law-needs-to-protect-animals-from-ai-234176

  12. Gallup. "In U.S., More Say Animals Should Have Same Rights as People." May 2015. Summarized in: Mercy for Animals, "Gallup: Majority of People Believe Animals Deserve Protection." https://mercyforanimals.org/blog/gallup-majority-of-people-believe-animals/ — See also: Physicians Committee for Responsible Medicine / Morning Consult survey, September 2024, finding 80% of Americans support phasing out animal experiments. https://www.pcrm.org/news/good-science-digest/physicians-committee-survey-finds-most-americans-favor-ending-animal

  13. Mellor, D.J. et al. "The 2020 Five Domains Model: Including Human–Animal Interactions in Assessments of Animal Welfare." Animals, 10(10), 1870, 2020. https://doi.org/10.3390/ani10101870

  14. "Here to stay and evolving fast: How GreenFish's AI-powered fish-forecasting tech is modernizing commercial fisheries." Responsible Seafood Advocate, September 2025. https://www.globalseafood.org/advocate/here-to-stay-and-evolving-fast-how-greenfishs-ai-powered-fish-forecasting-tech-is-modernizing-commercial-fisheries/