Algorithm Audit
A method by which an independent third party systematically examines the behavior of algorithm models. Its effectiveness depends heavily on data access permissions.
Timeline (last 90 days)
Algorithm auditing is the systematic examination of model behavior by an independent third party, and its effectiveness depends heavily on data access rights, which are held by the platforms.
Algorithm auditing refers to a systematic examination by an independent third party of an AI system's input data, model logic and output results in order to identify and correct bias; common methods include statistical analysis by dimensions such as race, gender and income, as well as the use of counterfactual testing.
For Seattle's legislation to be enforced in practice, it may need to be accompanied by an algorithmic audit mechanism requiring merchants to retain logs of pricing model input features for traceability.
Reasonable pathways to curbing the risks of predictive enforcement include clear legislative authorization, independent algorithmic audit mechanisms, and guarantees of the right to be informed and to appeal for individuals who have been flagged.
All Facts (4)
Algorithm auditing is the systematic examination of model behavior by an independent third party, and its effectiveness depends heavily on data access rights, which are held by the platforms.
50%UnverifiedAlgorithm auditing refers to a systematic examination by an independent third party of an AI system's input data, model logic and output results in order to identify and correct bias; common methods include statistical analysis by dimensions such as race, gender and income, as well as the use of counterfactual testing.
50%UnverifiedReasonable pathways to curbing the risks of predictive enforcement include clear legislative authorization, independent algorithmic audit mechanisms, and guarantees of the right to be informed and to appeal for individuals who have been flagged.
50%UnverifiedFor Seattle's legislation to be enforced in practice, it may need to be accompanied by an algorithmic audit mechanism requiring merchants to retain logs of pricing model input features for traceability.
50%