On *reviewability* of automated decision making (ADM), rather than *explainability*
Reviewability does not necessarily involve explanations. It is about exposing the decision-making process, including human processes, structures, & systems around a model
papers.ssrn.com/sol3/papers.cf…
Reviewability does not necessarily involve explanations. It is about exposing the decision-making process, including human processes, structures, & systems around a model
papers.ssrn.com/sol3/papers.cf…
Explanations focused on how a model has arrived at an output may miss much of what is important. A more holistic view could include information on testing & auditing procedures, training data, effects of decisions on protected characteristics, & more. 2/ @jennifercobbe
Judicial review of public sector decision-making does not simply assess the decision itself, but the decision-making process as a whole
An understanding of human decision-making as a process that begins before the decision and that has consequences that resonate afterwards 3/
An understanding of human decision-making as a process that begins before the decision and that has consequences that resonate afterwards 3/
Above quotes are from "Reviewable Automated Decision Making", by @jennifercobbe and Jatinder Singh 4/
papers.ssrn.com/sol3/papers.cf…
papers.ssrn.com/sol3/papers.cf…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
