Speciesism in AI: Built-in biases hinder animal welfare reform and plant-based diets

05 Feb 2024 --- The proliferation of AI systems is exacerbating the issue of speciesism and undermining animal welfare reform and the transition to plant-based diets, warn animal protection groups. Incoming regulation, notably the highly anticipated EU AI Act, fails to address the potential impacts of AI systems on animal welfare.

Speciesism is an increasingly influential idea in animal ethics that moral favoritism toward members of one’s own species is a prejudice, similar to sexism and racism.

Currently, Serbia is the only country in the world that mentions the effects of AI on animals in its legislation. The government’s ethical guidelines state: “...artificial intelligence systems that are being developed must be in line with the well-being of humans, animals and the environment.”

However, AI systems have also helped advance animal welfare standards, for example, with animal testing alternatives, alternative protein production and vegan-friendly chatbots.

Inherent biases
According to Mimi Bekhechi, campaigns advisor at PETA Netherlands, research suggests the language we use to describe other animals can affect how we view and treat them.

“AI language systems absorb our human-centric biases, so if we use phrases that make light of cruelty to animals or contain archaic, negative perceptions of certain species, we perpetuate speciesism and make it more likely that AI tools will as well,” she tells Food Ingredients First.

“Users and developers alike can — and should — seek to use more inclusive and respectful language (as PETA entities worldwide do in their use of ChatGPT and other platforms) when describing the fellow living, feeling beings with whom we share this planet.”

Dr. Joanna Swabe, senior director of public affairs for Humane Society International–Europe, highlights a recent study published by Springer Nature, which explores how ChatGPT reinforces biases about the intrinsic value of farm animals.

“There is an inherent risk of bias in AI systems because they are dependent on the initial data and algorithms used to train them,” she tells us.

“I saw a pertinent example on Instagram: an AI system had been asked to produce an image depicting a vegan Christmas dinner, yet — despite being given clear prompts about the meal being plant-based — it still managed to churn out a scene featuring a roast turkey. The only way this can happen is for the machine to have learned that festive fare inevitably includes meat.”

Regulatory recognition
Governments worldwide have so far failed to deliver regulatory frameworks to limit the potential risks of AI systems on wider society. Julia Buech, a senior consumer foods analyst at Rabo Research Food & Agribusiness, recently described the AI space as a “lawless jungle.”

Although the EU AI Act, which is expected to come into full force in 2026, will aim to limit AI systems based on their potential risks to society, the framework offers no specific protection for animals.

“I haven’t seen any reason to think there’s going to be any mention of animals in the EU AI Act,” Sam Tucker, executive director at Open Paws and CEO at VEG3, tells Food Ingredients First.

“More broadly, the general framework of regulating AI systems differently depending on their level of risk is a very good approach. High-risk systems should have more regulation, but there needs to be a consideration of the risks to animals as well as humans.”

“There’s huge impact potential through new regulation — but, obviously, animals don’t have the ability to lobby for their own interests. So it’s up to us as humans to consider them in legislation.”

A recent paper published by Springer Nature provides a systematic account of how AI technologies could harm non-human animals and how their welfare should be better recognized.If dogs are considered pets, why are cows considered food? Animal rights activists are concerned that AI systems perpetuate speciesism.

AI for positive change
Despite the issues of inherent bias and lacking regulation, AI systems have also produced positive outcomes for animal welfare standards and more sustainable food systems.

“The era of tormenting sentient rabbits, rats, dogs and other animals in laboratories is coming to an end, thanks in part to mind-blowing innovations that AI is helping to advance, such as organ-on-a-chip technology, which mimics the structure and function of human organs and can more accurately predict human drug responses than tests on animals can,” says Bekhechi.

For Tucker, the possibilities for AI systems as a force for good in animal welfare standards are “only limited by our imagination.” Open Paws and VEG3 are two platforms helping combat speciesism through AI-powered services.

“AI is a rapidly evolving technology, and we’re finding more use cases all the time, like the production of high-quality alternative proteins — there’s a lot of great work being done to make creating plant-based foods more efficient, tastier and cheaper,” he says.

“There’s massive potential for AI to help as a tool for public advocacy and marketing and provide resources to people who are interested in making lifestyle changes for animals, for example, chatbots that assist people in finding new plant-based recipes.”

However, the bias against certain species of animals is still very widespread across most AI systems.

“AI systems usually adopt the bias of whatever is in their training data…[and] there’s a lot of bias and toxic information on the internet. A lot of this was later trained out of AI systems, because we acknowledged it as a problem,” adds Tucker.

“But, none of the major AI companies currently have any mention of animals in their ethical guidelines, and they’re not instructing data workers to consider how responses affect animals.”

By Joshua Poole