Jake Fawkes

📣 Thrilled that our paper "Returning the Favour: When Regression Benefits from Probabilistic Causal Knowledge" has been accepted for an oral at #ICML2023. Joint work with my wonderful co-authors @FawkesJake @sejDino arxiv.org/abs/2301.11214

@jeanfrancois287 @FaaizTaufiq @BrieucLehmann Thanks @jeanfrancois287 !

@FaaizTaufiq @BrieucLehmann @FawkesJake

I'm trying to get an idea of who's doing research (at least loosely) related to fairness in stats / ML ⚖️ e.g. on topics including (but not limited to) - distribution shift - ascertainment bias - transfer learning Would v much appreciate any suggestions! 👇 #StatsTwitter #ML

This creates difficulties for causal fairness as it means correct causal models will not have independent noise . This makes fitting causal models harder and also means the application of many results from the causality literature will not be valid without further justification.

Further in many cases we can explicitly reject ignorability as samples are often selected. We show that unless this selection satisfies the following scaled version of counterfactual fairness, ignorability will not hold:

Recent paper “Selection, Ignorability and Challenges With Causal Fairness” with @ItsAStatLife and @sejDino to appear at @CLeaR_2022: arxiv.org/abs/2202.13774.