Examining the Historical Roots of IQ as AI Redefines Intelligence

Historian Angus McLaren has demonstrated that these efforts were rooted in eugenics, framing intelligence as an inherited biological trait subject to a racialized genetic hierarchy. Similar frameworks were applied to Indigenous populations. Sociologist Alexandra Giancarlo documented that Kenneth E. Kidd, who later established the Indian-Eskimo Studies program at Trent University, conducted IQ testing at the Mohawk Institute residential school in Ontario during the 1930s. The resulting publications reinforced racist theories of intelligence that informed Department of Indian Affairs policies across the Canadian residential school system. These testing mechanisms helped establish universities as arbiters of human capability, creating grading, testing, and ranking structures designed to sort and exclude individuals.

Modern Algorithmic Metrics and Education

Today, modern AI systems risk inheriting these classification habits. Researchers at the Rutgers AI Lab have highlighted how AI can condition individuals to align with narrow definitions of health, capability, or social value. These systems often operate through algorithms, data, and evaluations that embed historical biases concerning human potential.

Similarly, discourse surrounding cognitive decline from AI usage—sometimes termed "de-skilling" or "enfeeblement"—mirrors the pseudoscientific eugenic concept of "feeble-mindedness." Disability studies scholar Rua Williams notes that failing to recognize this historical context risks allowing modern AI initiatives to replicate historical projects that diminish human worth through arbitrary or constructed ideas of intelligence. Meanwhile, concerns over unregulated AI have aligned with broader anxieties regarding authoritarian governance, leading to demonstrations such as the protest outside the OpenAI DevDay 2026 conference in San Francisco where demonstrators demanded a pro-democracy commitment.

In higher education, fear of AI-driven cheating has led some educators to revert to handwritten or oral examinations, creating new barriers and concerns around ableism. At the same time, critics argue that an overreliance on standardized assessment and grading made educational work vulnerable to automated replication in the first place.

What it means for developers

For engineers and machine learning practitioners, this history provides crucial context for how automated models are evaluated and deployed. When benchmark suites attempt to evaluate models or humans using rigid, standardized scoring systems, they risk codifying century-old assumptions about intelligence that were flawed from their inception.

Developers building evaluation frameworks, automated hiring tools, or educational applications must consider whether their evaluation benchmarks reduce complex human abilities to simplistic, one-dimensional rankings. System prompts and algorithmic sorting mechanisms can inadvertently re-encode discriminatory biases under the guise of objective performance metrics.

When exploring model behaviors and building applications across diverse foundation models, developers can try top AI models cheaply through one API at https://apixoai.online to compare how systems handle complex tasks and diverse perspectives.

Ultimately, building responsible tools requires recognizing that intelligence cannot be neatly captured by a single number or a narrow set of automated tests. By shifting focus away from rigid rankings and toward human connection, critical reasoning, and adaptable problem-solving, developers and educators can avoid repeating the exclusionary mistakes of the past.

Source: Why the ugly history of IQ testing matters for AI and education — The Conversation AI. Written by the Apixo team from that report.