-
"All models are wrong, some are dangerous." -- @Abebab (at the complexity lounge) @BarbaraFrontera/1505917218105790469
-
amazing critique of ML from Abeba here... People focus on ML algorithm accuracy, when they need to recognize that modeling complex systems is not actually possible. The focus should be on awareness of harms and real life impacts, GIVEN that the models will be flawed.
-
Big takeaway for me. Abeba says the consequences are already bad. We already know how bad they are, and for whom (marginalized groups). So, stop assuming the public data sets used for ML are good. Assume they are bad unless mitigations have been employed.
-
"nothing about us without us"