しましまのFAT*2019まとめ

しましまのFAT*2019の参加・聴講記録 https://fatconference.org/2019/
0
前へ 1 ・・ 14 15 次へ
ACM FAccT @FAccTConference

Up next: A Taxonomy of Ethical Tensions in Inferring Mental Health States from Social Media (@snchancellor M Birnbaum. E Caine, V Silenzio & @munmun10) doi.org/gftmqb #fat2019

2019-02-01 05:42:00
Yuval Pinter @yuvalpi

Obermeyer and Mullainathan: an important part of the problem space is that some illnesses affect black populations at much higher rates than white populations and vice versa; in addition, in general white Americans have more access to health care #FAT2019

2019-02-01 05:42:19
Yuval Pinter @yuvalpi

Obermeyer and Mullainathan: so, even though costs are predicted correctly by the algorithm, the problem stems from a systemic discrepancy between these costs and actual health #FAT2019

2019-02-01 05:43:36
Yuval Pinter @yuvalpi

Obermeyer and Mullainathan: epilogue - we contacted the vendor who was very responsive and following a much larger-scale study a new version of the algo was deployed with less bias. #FAT2019

2019-02-01 05:44:50
Yuval Pinter @yuvalpi

Stevie Chancellor @snchancellor , Michael Birnbaum, Eric Caine, Vincent Silenzio, Munmun De Choudhury @munmun10 : A Taxonomy of Ethical Tensions in Inferring Mental Health States from Social Media dl.acm.org/authorize?N675… #FAT2019 [@mlatgt @ICatGT @gtcomputing ]

2019-02-01 05:46:09
Yuval Pinter @yuvalpi

Chancellor et al.: social media companies are well-able to detect mental health issues with users from the data they have access to. Unfortunately, this field is very multidisciplinary and there's no clear-cut methodology #FAT2019

2019-02-01 05:48:18
Yuval Pinter @yuvalpi

Chancellor et al.: in addition, there are many risks from bad predictions and many possibly dangerous implications. We develop a taxonomy of these tensions in order to make them explicit #FAT2019

2019-02-01 05:49:28
María Paz Hermosilla @mphermosilla

Excelente paper de @m_sendhil y @oziadias sobre la evaluación de sesgo racial en un algoritmo para detectar pacientes crónicos en el sistema de salud #FAT2019 dl.acm.org/authorize?N675… pic.twitter.com/ECJBAC9X0T

2019-02-01 05:49:50
拡大
Yuval Pinter @yuvalpi

Chancellor et al.: deep-dive into the second of three top-level nodes of the taxonomy - Validity, Interpretability, and Methods: Models range from linear regression (most interp.) to deep neural nets, which perform much better #FAT2019

2019-02-01 05:51:04
ACM FAccT @FAccTConference

Last in this session, Clear Sanctions, Vague Rewards: How China's Social Credit System Currently Defines Good and Bad Behavior (@severinengelma1 @silence_chen21 @fischerfel, C-Y Kau & J Grossklags) doi.org/gftmqd #fat2019

2019-02-01 05:52:00
Yuval Pinter @yuvalpi

Chancellor et al.: how interpretable does an algo need to be for us to be sure that someone is in a dangerous situation requiring intervention? How much can we risk false negatives - missing a person in need? #FAT2019

2019-02-01 05:52:09
Yuval Pinter @yuvalpi

Chancellor et al.: we try to start and solve these problems. First step - bringing key stakeholders (companies; clinicians) into the process from the very start #FAT2019

2019-02-01 05:54:03
Ian Stewart @alethioguy

Chancellor (@snchancellor) et al.: making predictions on mental health based on social media data can give clinicians new insight into diagnoses, as long as the clinicians are included in the entire prediction/interpretation process. #FAT2019 pic.twitter.com/TJ3PrrOHqR

2019-02-01 05:54:43
拡大
Yuval Pinter @yuvalpi

@snchancellor et al.: standardization of methods and evaluations for these systems across research communities is also an important next step; finally, designing mechanisms beyond ethic boards for these matters. #FAT2019

2019-02-01 05:55:46
Yuval Pinter @yuvalpi

Severin Engelmann, Mo Chen, Felix Fischer, Ching-yu Kao, Jens Grossklags: Clear Sanctions, Vague Rewards: How China’s Social Credit System Defines “Good” and “Bad” ‌Behavior dl.acm.org/authorize?N675… #FAT2019

2019-02-01 05:56:33
María Paz Hermosilla @mphermosilla

El paper que estaba esperando... este es un caso que vemos en el diplomado de #BigData para políticas públicas en @GobiernoUAI #FAT2019 dl.acm.org/authorize?N675… pic.twitter.com/r4Ze8asT0Z

2019-02-01 05:58:15
拡大
Yuval Pinter @yuvalpi

Engelmann et al.: we look into the Chinese Social Credit System (SCS), where "good" and "bad" behavior is recorded and used to reward or withhold services from individuals and their families #FAT2019

2019-02-01 05:58:21
Yuval Pinter @yuvalpi

Engelmann et al.: from the philosophical literature, we differentiate between the freedom condition and the epistemic condition (the individual is aware of the system - the case at hand) #FAT2019

2019-02-01 05:59:28
Yuval Pinter @yuvalpi

Engelmann et al.: our first study looked into the content of news stories that reflect moral standards (in China), found that there's a "virtue cascade" - explicit mention of a person's good values - and a negative stereotype used for suggesting the sanction of shaming #FAT2019

2019-02-01 06:01:52
Yuval Pinter @yuvalpi

Engelmann et al.: 2nd study was analysis of blacklists and redlists from the Beijing SCS database. Blacklists include an option for appealing inclusion decision, redlists include commendations (honorary titles / 5-stars) #FAT2019

2019-02-01 06:03:38
Yuval Pinter @yuvalpi

Engelmann et al.: the three most common reasons for inclusion in blacklists correspond to the "lao lai" stereotype condemned by the news stories #FAT2019 pic.twitter.com/86MufPMTCu

2019-02-01 06:04:53
拡大
Yuval Pinter @yuvalpi

Engelmann et al.: currently, there is still no uniform SCS deployed. #FAT2019

2019-02-01 06:06:00
Yuval Pinter @yuvalpi

Thus endeth my #FAT2019 tweetathon. Thank you @fatconference for a terrific conference, both in organization and content! It was a pleasure.

2019-02-01 06:07:17
madeleine clare elish @m_c_elish

#FAT2019 @jennaburrell same exact question, thank you for asking so eloquently

2019-02-01 06:07:52
Ben Packer @packer_ben

Talk by @benzevgreen at #FAT2019 on "Disparate Interactions: An Algorithm-in-the-Loop Analysis of Fairness in Risk Assessments," with Yiling Chen. Winner of the @fatconference's Best Technical + Interdisciplinary Paper award. pic.twitter.com/gjZVq8hgcp

2019-02-01 06:08:44
拡大
前へ 1 ・・ 14 15 次へ