madeleine clare elish: Hi #FAT2020! Here is the table + our paper link. TL;DR Achieving accountable and trustworthy AI means designing interventions as *socio*-technical systems with constant input from stakeholders and local expertise (not necessarily an explainable model). https://arxiv.org/abs/1911.08089 https://t.co/sCQScPbpEG
7 replies, 150 likes
madeleine clare elish: Hot off the press: Our #FATML case study on AI in healthcare. Tl;dr: developing ML systems is a ⚡️sociotechnical⚡️problem, people and institutions shape use, explainability is not the only means to accountability + much more! https://arxiv.org/pdf/1911.08089.pdf @MarkSendak @JFutoma #ML4HC
2 replies, 123 likes
Jacob Metcalf is at #fatconf2020 for the calçots: My favorite talk thus far at #fatconf2020: is algo explainability necessary if the underlying phenomenon itself is not explainable and there are other conflicting values that may be preferred, e.g., autonomy, trust, representation and sustained feedback? #whiteboxisthenewblackbox
1 replies, 35 likes
Eric Sears: If you want to read recent, rigorous research on the use of an AI tool in a clinical healthcare setting check out the paper "The Human Body is a Blackbox..." by @m_c_elish @MarkSendak and others that was presented at the 2020 @FAccTConference https://arxiv.org/abs/1911.08089
0 replies, 22 likes
Mark Sendak: We're thrilled to share a pre-print of a project we'll be presenting at the upcoming @fatconference: "The Human Body is a Black Box": Supporting Clinical Decision-Making with Deep Learning https://arxiv.org/abs/1911.08089
1 replies, 22 likes
Kevin Bankston: What fascinates me most about this #FAT2020 paper is the mention of the “emotional labor” that had to be done by (presumably mostly female) nurses mediating between this diagnostic ML system and the egos of (presumably mostly male) doctors. https://t.co/k3DLA3JfkD
2 replies, 15 likes
Joe Futoma @#FAT2020 (NYC job-hunting!): At #FAT2020? Want to learn abt a *real-world* ML system integrated into *actual* clinical practice, running live in @DukeHealth ED 24/7, & what it took to build trust & accountability w stakeholders to get there? Come to Sess. 2 tmrw 9am on explainability! @MarkSendak @m_c_elish
0 replies, 12 likes
Trooper Sanders: Great paper taking a comprehensive approach to fairness, accountability, and transparency in machine learning.
1 replies, 5 likes
Thomas Arnold: @katecrawford @HMRoff I bristle at the title, but this piece from @m_c_elish et al. is a much more serious and grounded medicine/black box reflection than Hinton's edgelord binarism. https://arxiv.org/pdf/1911.08089.pdf
0 replies, 5 likes
Jamie Clark: What a great start.
Building accountable and trustworthy #AI •requires• building accountable and trustworthy human processes behind it.
HT @m_c_elish @lizjosullivan
0 replies, 4 likes
rebecca bartel: AI in Healthcare: Achieving accountable and trustworthy AI means designing interventions as *socio*-technical systems with constant input from stakeholders and local expertise #FATML #FAT2020 https://arxiv.org/abs/1911.08089 @MarkSendak @m_c_elish #machinelearning https://t.co/TSnw7Dtifs
0 replies, 3 likes
Neil Carleton: No med classes means catching up on writing and reading... A particular joy to read has been this #arXiv paper on critical assessment of #MachineLearning systems into clinical care and the many considerations that may go into it. https://arxiv.org/abs/1911.08089
2 replies, 2 likes
Likelihood T. Prior: Hot off the (pre-print) press!! "The Human body is a black box". Fairness, accountability, and transparency requires much more than 'interpretability' https://arxiv.org/abs/1911.08089 @MarkSendak @JFutoma
0 replies, 2 likes
Wilneida Negrón: Incredibly practical framework for anyone looking to apply machine learning applications to complex social problems.
0 replies, 1 likes
Emanuel Moss: Great talk and paper! How do we think about algorithmic explainability for sepsis, something doctors don’t think of as “explainable”?
0 replies, 1 likes
Internet Ethics: #ethics #AI #health #tech #data
0 replies, 1 likes
Found on Jan 28 2020 at https://arxiv.org/pdf/1911.08089.pdf