Conservative AI and social inequality: Conceptualizing alternatives to bias through social theory

"Bias" alone is insufficient terminology: The paper argues that sociologists' existing frameworks for understanding inequality are more precise than the catch-all term "bias." AI discussions should engage with specific theories of how inequality is systematically reproduced in society, rather than treating all problems as data bias issues.

AI systems can either reinforce or disrupt inequality: The author distinguishes between "conservative" approaches that reproduce the status quo and "radical" approaches designed to counter systemic inequality. This framing suggests that neutrality in AI design is impossible—systems actively either strengthen or challenge existing inequalities.

Conservative approaches to bias have structural limitations: The paper identifies specific shortcomings in how current AI addresses class, gender, and racial bias, suggesting that surface-level bias mitigation strategies fail to address the deeper social structures that perpetuate these inequalities.

Social context must be central to AI development: The author contends that "societal issues can no longer be out of scope for AI and machine learning, given the impact of these systems on human lives," calling for greater integration of social science and humanities perspectives in how AI systems are designed and governed.

Critical AI scholarship must move beyond data-focused solutions: The paper advocates for engagement with scholarship that analyzes "structured ways of perpetuating inequality" rather than limiting focus to biased training data, opening possibilities for fundamentally different approaches to responsible AI.