Ideological stagnation & propagation via algorithmic affordances (snuggle paper)
"subjective" algorithms and the hidden social problems they can cause (Tufecki & Sandvig)
standpoints and successors, "successor system" and "successor infrastructure" (Harding & Haraway "standpoint & successor", Geiger "successor system", Ribes "infrastructure")
The system: Infrastructure for successor technologies: The People's Classifier Service
GOAL: AI as a service. Easy access from central platform or bespoke code --> New intelligent UIs
Open reporting of process & fitness metrics & transparency of signal enables critique --> identification of problems & development of improved feature sets
Open coder systems --> new and better training sets --> new and better classifiers
The evaluation:
Increased salience of breakdowns? Evidence of reflection? (talking about what the alg's miss)
Propagation of new ideological views via algorithmic affordances? (tools that use signal in new ways)
Effects on objective outcomes (e.g. productivity)?