Profile picture
, 5 tweets, 17 min read
A widely used #AI algorithm that predicts health risk exemplifies automated racial bias
science.sciencemag.org/content/366/64…
@oziadias @UCBerkeleySPH @UCBerkeley @m_sendhil @UChicago and collaborators @sciencemagazine
@oziadias @UCBerkeleySPH @UCBerkeley @m_sendhil @UChicago @sciencemagazine Extensive coverage, insight of this important paper science.sciencemag.org/content/366/64… by @ruha9 @PrincetonAAS (and the New Jim Code)
washingtonpost.com/health/2019/10… by @Carolynyjohnson w/ @suchisaria (at least it can be corrected)
sciencemag.org/news/2019/10/h…
not just @Optum
@oziadias @UCBerkeleySPH @UCBerkeley @m_sendhil @UChicago @sciencemagazine @ruha9 @PrincetonAAS @Carolynyjohnson @suchisaria @Optum 2 other very good explainers:
1. Thread by co-author @brianwpowers
2. @WIRED: wired.com/story/how-algo… by @tsimonite w/ @blkwomenshealth's Linda Goler Blount
@oziadias @UCBerkeleySPH @UCBerkeley @m_sendhil @UChicago @sciencemagazine @ruha9 @PrincetonAAS @Carolynyjohnson @suchisaria @Optum @brianwpowers @WIRED @tsimonite @blkwomenshealth It's all about the inputs.
As in the case, #AI algorithms are neutral and can't be blamed for bias.
The humans (@Optum) who unwittingly accepted costs of #healthcare as a metric of health blew it.
@oziadias @UCBerkeleySPH @UCBerkeley @m_sendhil @UChicago @sciencemagazine @ruha9 @PrincetonAAS @Carolynyjohnson @suchisaria @Optum @brianwpowers @WIRED @tsimonite @blkwomenshealth They may want to rethink their We Know How, We Are the How ads. Especially the ones How "healthcare works better for all of us"
Missing some Tweet in this thread? You can try to force a refresh.

Enjoying this thread?

Keep Current with Eric Topol

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Related hashtags

#AI #healthcare

More from @EricTopol see all

Profile picture
Eric Topol
@EricTopol
No, It won't @TIME
@TIME Let me tell you why it won't @TIME @TIMEHealth
1. Robots don't actually see; they ingest pixels
2. I don't want to see a robot for my medical care, nor do most people
3. We have enough depersonalization & lack of humaneness in #healthcare now; we don't need to make it⬆️robotic
“Human connection is the only way we will win in our eventual war with the robots.”— Amy Poehler :) @smrtgrls
Read 3 tweets
Profile picture
Eric Topol
@EricTopol
Instead of 10,000 years, it took ~200 seconds.
A computing milestone "comparable in importance to the Wright brothers 1st flight"
nature.com/articles/s4158… @NatureNews
nature.com/articles/d4158… @NatureNV
blog.google/perspectives/s… @Google by @sundarpichai
technologyreview.com/s/614608/googl…
@NatureNews @NatureNV @Google @sundarpichai @GoogleAI @techreview "It’s early days, but where quantum mechanics shines is the ability to simulate molecules... Drug discovery is a great example"
"when I look at medicines, this is why I am confident one day QC will drive progress"@sundarpichai
A simple molecule: caffeine has 10 to the 48 states
@NatureNews @NatureNV @Google @sundarpichai @GoogleAI @techreview A wise, balanced editorial @nature on this milestone achievement in science and engineering vis-à-vis the quantum gold rush
nature.com/articles/d4158…
and more perspective
nature.com/articles/d4158…
by @LizzieGibney w/ video
Read 3 tweets
Profile picture
Eric Topol
@EricTopol
"Prime" #CRISPR genome editing: a new chapter integrating reverse transcriptase for search and replace that is yet more versatile and precise; might be suitable for approaching 89% of disease-causing mutations
nature.com/articles/s4158…
@nature by David Lui and team @broadinstitute
@nature @broadinstitute 2 different ways to represent this significant advance 😉
1. for the genomics community
by @davidrliu @liugroup @broadinstitute
2. for the uninitiated
technologyreview.com/s/614599/the-n… by @antonioregalado @techreview
@nature @broadinstitute @davidrliu @liugroup @antonioregalado @techreview And a 3rd representation, explainer by @MeganMolteni and @WIREDScience wired.com/story/a-new-cr…
and a very good critique thread by @circuit_logic via @antonioregalado (+1, get rid of the 89%)
Read 3 tweets

Related threads

Profile picture
Tage Rai
@tage_rai
1/10 Some thoughts on @oziadias @m_sendhil et al new paper in @sciencemagazine on racial bias in health algorithm. science.sciencemag.org/content/366/64…
2/ I've seen many papers documenting racial bias in algorithms. But authors usually only have access to output and have to do cool detective work to infer how bias arises. This paper is different b/c authors had access to algorithm's inputs too.
3/ The algorithm is a screening tool designed to identify people who would benefit from preventative health care program. Authors find racial bias in screening that may affect millions. The question is how does it happen?
Read 11 tweets
Profile picture
Brian Powers
@brianwpowers
New paper w/ @oziadias, @m_sendhil, and Christine Vogeli out today in @sciencemagazine.

Does a predictive algorithm widely used for population health management exhibit racial bias?

Short answer: YES.

Full Article: science.sciencemag.org/content/366/64…

THREAD (1/8)
cc: @sacjai @ashishkjha
BACKGROUND

Care management programs—which provide high-need, complex patients with additional services and support—have proliferated. An entire industry now exists to supply delivery organizations with predictive algorithms to help identify patients for these programs. (2/8)
WHAT WE DID

We looked for evidence of racial bias in a predictive algorithm used to identify patients for care management programs. This algorithm generates a “risk score” which prioritizes patients for increased services. It is one of the most widely used.(3/8)
Read 8 tweets
Profile picture
Ashley Webster
@crashwebster
Thread: Blatant #ClimateDenial has lost. The new conflict is whether #ClimateChange is a crisis requiring urgent social and #economic transformation, or something we can adapt to incrementally while carrying on with business as usual. 1/40
2/ A recent opinion piece by Bjorn Lomborg in the @globeandmail is representative of the 'business as usual argument'. Let's deconstruct it to see how this gets passed off as the pragmatic point of view, and why it's false. theglobeandmail.com/opinion/articl…
3/ The basic tactic is this: provide a few common-sense truths to gain the air of pragmatism and then feed us the story we're desperate to hear: that everything will be fine, that growth will outpace #ClimateDamages and life can carry on as usual.
Read 41 tweets
Profile picture
Stéphane Debove
@stdebove
Comment réparer la science ??? 🤔🤔

Cette vidéo mérite bien plus de 1730 vues (même si le son est un peu pourri).

Quelques nouvelles encourageantes du front que je vous résume 🔽 [1/n]

#psychologie #science

Le pré-enregistrement : consiste pour les chercheurs à dire à un journal AVANT de commencer leur étude le but de leur étude et les méthodes / statistiques qu’ils vont utiliser.

Si le journal l’accepte, il s’engage à publier l’étude *quel que soit son résultat*, positif ou nul
Ça permet que les chercheurs ne soient plus incités à rendre leur étude aussi sexy que possible pour qu’elle soit publiée, et ça permet que les journaux arrêtent de ne publier que les résultats positifs (qui trouvent un effet, confirmant l’hypothèse du chercheur)
Read 13 tweets
Profile picture
Dagmar Monett
@dmonett
A "worrying analysis":

"18 [#deeplearning] algorithms ... presented at top-level research conferences ... Only 7 of them could be reproduced w/ reasonable effort ... 6 of them can often be outperformed w/ comparably simple heuristic methods."

Paper:
lnkd.in/dTaGCTv

#AI
[Updates worth tweeting]

2/
 There is much concern about #reproducibility issues and flawed scientific practices in the #ML community in particular & #academia in general.

Both the issues and the concerns are not new.

Isn't it time to put an end to them?
3/
 There are several works that have exposed these and similar problems along the years.

👏👏 again to @Maurizio_fd et al. for sharing their paper and addressing #DL algorithms for recommended systems (1st tweet from this thread).

But there is more, unfortunately:
Read 18 tweets

Trending hashtags

#HBWH #HousingAffordability #WG3 #PDoH #GlobalWarming #SDGs #openaccess #economic #ClimateDenial #AI #bonnchallenge #SR15 #carnassials #tweetorial #WG2 #deeplearning #science #metaheuristic #UniteBehindTheScience #cholesterol #HealthyPublicPolicy #LDL #Foreshadowing #bayes #CNN
Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!