Classification systems constitute infrastructures with political consequences in AI training.

Evidence

Star and Bowker’s foundational STS study analyzes how classification systems (medical, scientific, racial) shape practice and possibility. Their method is infrastructure studies focused on classification. Sorting Things Out is essential reading for COMM/AI/equity doctoral work. For doctoral candidates, Star and Bowker provide the foundational analytic for understanding how AI training labels (including racial and gender categories) constitute classification infrastructures with political consequences. — , The Braided Reading EPUB, 1526

Categorization

• XREF: Extends to contemporary ML/LLM label design debates, connecting to Noble, Benjamin, and Eubanks in this chapter. — connects to: Machine learning increasingly intertwines technical problems with human and social questions; AI training data is invisible yet permanently shapes deployed system behavior — Brian Christian, The Alignment Problem, 150; Machine learning systems are not inherently fair or just by default (+3 more)

Broader Connections

• Machine learning increasingly intertwines technical problems with human and social questions #expands (0.62)
• AI training data is invisible yet permanently shapes deployed system behavior #expands (0.55) — Brian Christian, The Alignment Problem, 150
• Machine learning systems are not inherently fair or just by default #complicates (0.50)