Early statistical discrimination was visible and contestable because few variables were in play.

Evidence

Second, the forms of discrimination that resulted from data collection were also largely public—and visible as such—resulting, for example, in disputes discussed by historian Dan Bouk about insurance premiums with US commercial insurers. With relatively few variables in play and calculation tasks far less complex than those of today, the reason for someone’s higher insurance premiums (often their ethnicity, gender, or “statistical pasts”) was clear enough, and the correlation crude enough, to be publicly challenged. In the nineteenth-century United States, African Americans succeeded in lobbying for adjustments in insurance-industry practice. — Nick Couldry & Ulises A. Mejias, The Costs of Connection, 379

Categorization

• XREF: This inverse relationship between complexity and public contestability feeds directly into the reader’s attention/platform/surveillance critique: opaque algorithmic variables make modern discrimination unchallengeable. — connects to: Datafied discrimination is opaque to subjects and operators alike, so uncorrected — Nick Couldry & Ulises A. Mejias, The Costs of Connection: How Data Is…, 580; Algorithmic discrimination affects real people despite targeting only probabilistic data doubles. — Nick Couldry & Ulises A. Mejias, The Costs of Connection: How Data Is…, 408

Broader Connections

• Datafied discrimination is opaque to subjects and operators alike, so uncorrected #contradicts (0.82) — Nick Couldry & Ulises A. Mejias, The Costs of Connection, 580
• Algorithmic discrimination affects real people despite targeting only probabilistic data doubles. #contradicts (0.78) — Nick Couldry & Ulises A. Mejias, The Costs of Connection, 408