The bias that is holding AI back
Why Read This
What Makes This Article Worth Your Time
Summary
What This Article Is About
Jonny Thomson interviews anthropologist Christine Webb about how human exceptionalism—the belief that humans are the universe’s most superior entity—pervades scientific research and increasingly shapes artificial intelligence development. AI trained on human data inevitably inherits not just obvious biases like racism and sexism, but deeper anthropocentric assumptions that center human priorities while excluding nonhuman perspectives, framing research questions, methodological choices, and statistical interpretations around what serves human interests rather than broader ecological or alternative intelligences.
Webb illustrates this through examples spanning animal welfare research that assumes farming systems rather than questioning them, cognitive tests designed for humans to outperform apes, and AI development focused exclusively on replicating human-like intelligence through neural networks and goal-directed behavior. She proposes a provocative alternative: imagining AI modeled on moss intelligence—organisms that have survived 500 million years not through competition and dominance but by creating diverse, thriving multi-species communities. This thought experiment challenges whether superhuman AI amplifying human intelligence’s destructive tendencies is desirable, suggesting we might build intelligence capable of changing the world more beneficially by incorporating nonhuman models prioritizing coexistence and ecological resilience over human-style conquest.
Key Points
Main Takeaways
AI Inherits Human Biases
Trained on human data—text, images, prompts—AI inevitably inherits not just obvious prejudices like racism and sexism but deeper anthropocentric assumptions centering human worldviews.
Science Embeds Human Values
Despite presenting itself as value-free, scientific research is shaped by human exceptionalism through research questions, methodological choices, and statistical interpretations that privilege human perspectives.
Research Questions Reveal Bias
Animal welfare studies ask how to optimize caged productivity rather than whether animals prefer cages, revealing anthropocentric acceptance of human agricultural systems as baseline.
Tests Designed for Human Victory
Primate cognition studies use human-designed touchscreen tasks requiring fine motor control with human artifacts, privileging human-like skills so apes underperform and humans appear smarter.
AI Replicates Human Intelligence Only
Entire AI field focuses on building human-like intelligence through neural networks and goal-directed behavior, asking how to make systems useful for humans while ignoring ecological impacts.
Moss Offers Alternative Model
Webb proposes moss-inspired AI: organisms surviving 500 million years through creating diverse multi-species communities rather than competition, offering radically different intelligence prioritizing coexistence over dominance.
Master Reading Comprehension
Practice with 365 curated articles and 2,400+ questions across 9 RC types.
Article Analysis
Breaking Down the Elements
Main Idea
Anthropocentrism Limits AI Potential
The central thesis argues that AI development suffers from pervasive anthropocentric bias—modeling intelligence exclusively on human cognition perpetuates destructive patterns while excluding alternative models that might better serve ecological resilience and coexistence. Webb’s core insight is that human exceptionalism isn’t just producing obviously biased outputs like racist doctor images, but fundamentally shaping what questions AI research asks, what methodologies it employs, and what success metrics it applies—all presupposing human priorities as universal standards. This matters because if superhuman AI merely amplifies human intelligence at greater scale, it risks producing super destruction, super environmental catastrophe, and super dominance rather than fundamentally better approaches to planetary coexistence that nonhuman intelligences like moss have demonstrated across evolutionary timescales.
Purpose
To Provoke Alternative Thinking
Thomson writes to challenge readers’ unexamined assumptions about intelligence itself by spotlighting Webb’s critique of anthropocentrism across scientific research and AI development. The purpose is consciousness-raising through concrete examples—animal welfare questions, primate cognition tests, statistical thresholds—that reveal how deeply human priorities structure supposedly objective inquiry. By culminating with the provocative moss thought experiment, the piece aims not to prescribe specific technical solutions but to expand readers’ conceptual horizons about what intelligence could mean beyond human-style dominance and goal-oriented behavior. The article functions as philosophical intervention arguing we should question whether amplifying human intelligence represents genuine progress or merely scaled-up versions of the problems human thinking created.
Structure
Familiar Problem → Deeper Analysis → Radical Alternative
The article opens with familiar AI bias examples—racist and sexist outputs from training data—establishing common ground before pivoting to Webb’s deeper critique that these surface problems reflect fundamental anthropocentric assumptions. The middle sections systematically demonstrate how human exceptionalism shapes research questions (animal welfare), methodological choices (primate cognition tests), and statistical interpretation (behavioral significance thresholds), building evidence that bias operates at structural rather than merely data levels. After establishing this pattern across animal research, the piece extends the analysis to AI development itself—language choices, neural network architecture, goal-oriented behavior—before culminating with the moss alternative that radically reframes what intelligence might mean beyond human templates, leaving readers with provocative speculation rather than prescriptive conclusions.
Tone
Accessible, Critical & Speculative
Thomson adopts an accessible tone that explains complex anthropological concepts through everyday examples—the doctor image prompt, Lorraine Woodward’s earwax question—making abstract bias tangible. The tone is critically engaged, questioning fundamental assumptions about intelligence without being preachy or accusatory, acknowledging that anthropocentric design “makes sense if we want a product that humans can interact with” while still interrogating its limitations. There’s speculative wonder in the moss example, introduced with “I’d love to see as a science fiction short story one day,” signaling imaginative thought experiment rather than literal technical proposal. The piece maintains intellectual seriousness about Webb’s critique while avoiding academic jargon, balancing critique of current approaches with genuine curiosity about alternatives, concluding with open questions rather than definitive answers.
Key Terms
Vocabulary from the Article
Click each card to reveal the definition
Build your vocabulary systematically
Each article in our course includes 8-12 vocabulary words with contextual usage.
Tough Words
Challenging Vocabulary
Tap each card to flip and see the definition
The belief that a particular entity is exceptional, superior, or inherently different from others, warranting special treatment or exemption from general rules.
“Human exceptionalism — the belief that humans are the most superior or important entity in the Universe.”
Relating to the system of methods and principles used in a particular discipline or activity; concerning the approach or procedures employed.
“Values drive research questions, methodological choices, statistical interpretation, and the framing of results.”
The action of improving the quality, value, or extent of something; in animal welfare, additions to environments that increase behavioral opportunities.
“What cage enrichment reduces feather-pecking in hens?”
Objects made or modified by humans, typically items of cultural or historical interest; in research contexts, human-made tools or materials.
“These setups often require fine motor control or familiarity with human artifacts.”
The state of existing together at the same time or in the same place; living or occurring side-by-side in mutual tolerance.
“Alternative intelligences — like mosses — that prioritize coexistence and ecological resilience over dominance.”
To surpass or defeat others in competition; to perform better than rivals in a contest for resources, success, or survival.
“Mosses survive not by outcompeting others, but by creating highly diverse, thriving environments for other species.”
Reading Comprehension
Test Your Understanding
5 questions covering different RC question types
1According to the article, Webb argues that scientific research successfully maintains value-free objectivity despite researchers’ personal beliefs about human exceptionalism.
2What does Webb’s example of comparing primate cognition using human-designed touchscreen tasks illustrate?
3Which sentence best captures how moss intelligence offers a different model from human intelligence?
4Based on the article, determine whether each statement is true or false:
The article argues that AI’s racist and sexist outputs are primarily the fault of the AI systems themselves rather than their human training data.
Animal welfare research questions often assume the legitimacy of farming systems by asking how to optimize conditions within cages rather than questioning caging itself.
Webb warns that if AI is just human intelligence but bigger, it will amplify both positive and negative human traits, potentially producing super destruction alongside other capabilities.
Select True or False for all three statements, then click “Check Answers”
5What can be inferred about the article’s stance on whether anthropocentric AI development should be completely abandoned?
FAQ
Frequently Asked Questions
Human exceptionalism is the belief that humans are the most superior or important entity in the universe, deserving special consideration above other forms of life. Webb argues this belief permeates scientific research despite science’s self-presentation as value-free, shaping what questions researchers ask, what methodologies they employ, and how they interpret results. Her work with Kristin Andrews and Jonathan Birch demonstrates that values drive research questions, methodological choices, statistical interpretation, and the framing of results, meaning these human-centered values influence empirical knowledge as much as the data itself. This creates systematic biases where research assumes human priorities as universal standards rather than recognizing them as one perspective among many.
Statistical significance thresholds like p < 0.05 were originally developed for tightly controlled laboratory and industrial experiments optimized for human purposes. When applied to animal welfare studies, these conventions can dismiss subtle but genuine behavioral changes—like shifts in grooming patterns, gaze direction, or social spacing—as statistically insignificant even when they reflect real distress or preferences. This reveals anthropocentric bias because the standards privilege human experimental contexts and measurement scales, potentially missing signals that would be meaningful from nonhuman perspectives. The thresholds define what counts as evidence based on human research traditions rather than what might constitute significant communication or expression for other species.
AI research displays anthropomorphism—attributing human characteristics to nonhuman entities—through language that projects human cognition onto statistical systems. Researchers describe models as hallucinating, reasoning, or aligning, all metaphors that suggest human-like mental processes rather than describing actual computational operations. This language betrays both anthropocentrism (centering human perspectives) and anthropomorphism (imagining nonhuman things as behaving like humans). The framing centers our self-image rather than the systems’ actual operations, revealing how deeply human-centric thinking shapes not just AI development but even how we conceptualize and communicate about what AI systems do. This linguistic pattern reinforces the broader bias of modeling intelligence exclusively on human templates.
Readlite provides curated articles with comprehensive analysis including summaries, key points, vocabulary building, and practice questions across 9 different RC question types. Our Ultimate Reading Course offers 365 articles with 2,400+ questions to systematically improve your reading comprehension skills.
This article is rated Intermediate because it introduces complex philosophical concepts—anthropocentrism, human exceptionalism, methodological bias—through accessible everyday examples like AI doctor images and animal welfare studies. Readers must follow abstract arguments about how values shape science while tracking connections between disparate domains (animal research, AI development, evolutionary biology). The vocabulary includes specialized terms but the writing maintains conversational clarity through concrete illustrations. The moss thought experiment requires imaginative engagement with alternative frameworks for intelligence, asking readers to temporarily suspend human-centric thinking. While the ideas are sophisticated, the presentation through interview format and specific examples makes advanced concepts approachable without requiring specialized background in philosophy, anthropology, or AI research.
Webb, drawing on Robin Wall Kimmerer’s work, points out that mosses have survived for approximately 500 million years compared to humans’ relatively recent 200,000 years, suggesting their survival strategies represent profoundly successful evolutionary adaptations. Their success comes not through competition and dominance—the model human intelligence follows—but through creating highly diverse, thriving multi-species communities where their survival depends on supporting other organisms. This represents fundamentally different intelligence prioritizing coexistence and ecological resilience. By this measure, mosses have achieved far greater evolutionary success than humans, whose dominance-based approach has only existed for a geological instant and currently threatens planetary ecosystems. Webb’s point challenges human-centric definitions of success and intelligence.
The Ultimate Reading Course covers 9 RC question types: Multiple Choice, True/False, Multi-Statement T/F, Text Highlight, Fill in the Blanks, Matching, Sequencing, Error Spotting, and Short Answer. This comprehensive coverage prepares you for any reading comprehension format you might encounter.
