We used fictitious resumes to find employer bias.
We used statistics to find bias in a healthcare algorithm.
Side by side these studies show why discrimination perpetrated by algorithms is so very different from discrimination perpetrated by people
nyti.ms/2qqPUPW
12
349
750
Any discussion of re-opening the economy must recognize not all businesses are the same. Some contribute more to spread than others.
Cellphone data can help identify them - see for yourself in our OpEd nyti.ms/3c8gm3q ๐งต
7
203
497
We want algorithms that are simple and interpretable.
We want algorithms that are fair
A paper with Jon Kleinberg that shows why you probably can't have both
arxiv.org/abs/1809.04578
8
226
487
๐งต1/ New paper: We argue machine learning should be used to generate truly novel hypotheses.
The application: judge decision-making. We find a single factor is heavily related to who gets jailed: the defendantโs mugshot.
What are judges looking at?
Our method unpacks this...
7
100
396
258,023
Our paper on how machine learning fits in econometric toolbox w/@jannspiess J.Econ.Perspectives bit.ly/2pF5OkN 1pic summary:
4
170
366
1/2 Exciting to announce this summer school in machine learning, economics and policy!
Whether you want to work in this are or already do, it's a great way to learn new tools and get research ideas.
bit.ly/3dz0YQH
Amazing list of faculty already lined up +more to come
2
110
351
Excessive force. Unnecessary arrests. The policing problem seems intractable.
A new large-scale RCT with Chicago police offers a glimmer of hope. And insight into *one* overlooked reason (obviously not the only reason) for the problems of policing
bit.ly/3Ex8Rmm๐งต
2
105
335
90,951
Self-replication:
We re-ran the studies from our 2012 scarcity paper with larger N: learned a lot - not everything replicates (though a lot does).
osf.io/7z38r/
Would encourage others try self replication
Any lab interested in replicating our work? (DM me)
4
110
294
I rarely feel comfortable sharing personal stories in my columns, but this one was so apt I pushed past my reticence. Full column here: nyti.ms/2qqPUPW
5
61
306
1/3: Our recent @sciencemagazine paper finds large racial bias in an important health care algorithm.
We want to help fix the problem. If youโre at a health system or a manufacturer that wants to know more or do better โ contact us at caai@uchicago.edu
bit.ly/2BUSNKU
6
109
247
Since science is entirely a team effort, I'm always a little skeptical of individual prizes. But the @infosysPrize has a wonderful mission: "inspire young Indians to choose a vocation in research." And what an honor to contribute to that goal
16
36
240
Heartbroken about police killings of African Americans, I dug into the data. It changed my thinking
nyti.ms/1MIq8Jf
21
166
236
We had to act quickly to contain the virus
We have to move equally quickly to contain the economic fallout
My attempt to think through some policies that should be on the DO NOW list
nytimes.com/2020/03/19/businโฆ
9
88
212
Are you doing a ecoomics-related PhD?
Interested in AI/ML?
Then apply to our Summer Institute in Machine Learning!
Learn from awesome faculty
Meet other like-minded students
Get exposed to cool, new ideas
bit.ly/3ZRvTxr (march 24 deadline)
3
77
191
86,014
To companies expressing support for racial equity ๐๐พ๐๐พ
Now show you mean it. Give workers (& contractors) paid time off to vote #time4voting
Hourly & gig workers are already stretched - don't force them to make a financial sacrifice to vote.
nyti.ms/2XWFwNG
49
139
To fight discrimination we must catch it but the inscrutability of human choice processes makes detection and proof hard. We argue algorithms make this crucial challenge easier but only with new policies in place. First draft bit.ly/2WZ57Ds Kleinberg Ludwig @CassSunstein
4
49
125
Terrific job market paper on cognitive taxes of poverty by Claire Duquennois: exam questions that involve money reduce accuracy on subsequent questions. Once money is on the mind even poor children apparently find it hard to think about other things. h/t @FrankSchilbach
Claire Duquennois
JMP: "Fictional Money, Real Costs: Impacts of Financial Salience on Disadvantaged Students"
Website: sites.google.com/view/claireโฆ
47
125
Great rundown of discrimination research in economics by my friend and long-time collaborator Marianne Bertrand: review.chicagobooth.edu/behaโฆ
1
53
124
Long slog but finally our paper on machine learning applied to bail decisions
nber.org/papers/w23180
@jure @hima_bindu
3
59
124
The applied AI Center @UChicago is hiring predocs.
Apply a range of empirical tools (ML, causal inference, generative models, ontologies) to interesting problems (e.g. in health, education, discrimination..even baking)
Join us before econ/stats/CS PhD! bit.ly/2HF7Z5t
5
41
127
Voting shouldn't be a financial burden.
Employers should give paid time off to your employees & contractors, especially hourly or gig workers.
If you agree
1) please retweet and
2) contact your employer #time4voting
nyti.ms/2XWFwNG
8
65
111
1/2: Fighting the coronavirus is hard enough. Regulations shouldn't make it harder
nytimes.com/2020/03/24/businโฆ
(column from @R_Thaler and me)
On the front lines and see a regulation that needs suspending? Tell us - see next tweet for how to get in touch
3
63
117
Very impressive resource for behavioral interventions
Find proven behavioral strategies across #health, #financialinclusion, #education, & more (& check out #bhub's new look!) bhub.org
1
53
118
Excited to be part of this paper.
We study a question about language models that I find fascinating: have they come to โunderstandโ the world behind the sequences they are trained on?
New paper: How can you tell if a transformer has the right world model?
We trained a transformer to predict directions for NYC taxi rides. The model was good. It could find shortest paths between new points
But had it built a map of NYC? We reconstructed its map and found this:
3
18
116
30,797
A sampling of 19 great Black economists working today - I'm lucky to have the opportunity to learn from their research and even luckier to know a few personally. #Juneteenth
bit.ly/3ehLL4F
1
22
104
Winning best actress award increases divorce rates (not so for best actor): example of hidden tax on successful women
nyti.ms/2tgVyVu
11
53
97
What many people get (very) wrong about the marshmallow test study n.pr/2rhswCu
4
57
95
Bob Frank's fantastic course inspired me to become an economist. Whether you've had the privilege of being his student or (especially if you have) not, you'll want to listen to his retirement lecture
I was at first disappointed to learn that my retirement lecture would have to be delivered remotely. But since everyone's welcome to tune in, I'm delighted that several thousand people have already signed up. Register here to join us at noon tomorrow:
ecornell.com/keynotes/overviโฆ
10
98
Really excited about this paper. Part of our continued efforts to make sense of what ML models are and arenโt doing.
Can an AI model predict perfectly and still have a terrible world model?
What would that even mean?
Our new ICML paper formalizes these questions
One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws ๐งต
3
17
102
11,135
Come work in our lab as a Predoc!
We use diverse tools: machine learning, experiments, observational causal inference
Cross disciplines - economics, psychology, computer science
To answer questions about people
& hopefully make the world a bit better
bit.ly/3o20ZRh
1
26
88
Well, this was a great way to greet the new decade! Obviously Cass is being overly generous but I'll take it! :)
Best nonfiction book of the decade? Scarcity, by @m_sendhil and Eldar Shafir.
amazon.com/Scarcity-Science-โฆ
3
9
82
Great prizes week for behavioral science:
Richard Thaler @R_Thaler - Nobel
Betsy Levy Paluck @betsylevyp - Macarthur Fellow @macfound
24
83
Easy to get randomized evaluations wrong. Great #EvaluationToolkit from @JPAL_NA on how to do them well: j-p.al/kr6
1
19
82
What will AI do to economy? Who knows. Let's prepare for the unpredictable: nyti.ms/2gAw9QO (happy to get star trek ref in)
2
65
77
Agreed! But worth changing the "If you fail" framing? Going the non-academic route (or leaving ABD) is not a failure but using one of the options a PhD program offers to your advantage. I fear PhD stress is in small part due to such negative framings.
4
4
74
Great job market paper on topic close to my heart: algorithmic hypothesis generation.
(As a bonus, I get be proud to have had the chance to advise a student as creative as Raf)
7
76
12,796
Possibly biased - this symposium on econometrics in J.Economic.Perspective is great
bit.ly/2oY8pJQ
1
47
71
We talk about algorithms and data. We should spend more time talking about metrics. @math_rachel lays out teh case persuasively: fast.ai/2019/09/24/metrics/
2
23
71
Behavioral economists have a lot to offer in the building and evaluation of LLMs. Exhibit 1:
How can the same LLM pass AP math and also claim 9.10 > 9.9?
Our #ICML2024 paper: LLMs are wrong in ways that people can't predict. Benchmarks can be misleading because people decide how LLMs are used.
Solution: measure LLM alignment with the "human generalization function" ๐งต
5
11
70
10,664
New article on how to increase diversity in the economics profession by Amanda Bayer, @GaryAHoover Ebonya Washington
A must read. Great data and essential suggestions.
pubs.aeaweb.org/doi/pdfplus/โฆ
1
18
63
Interesting paper on sleep...
"Why Don't We Sleep Enough?" In a field experiment using wearable devices, multiple surveys, and randomized monetary incentives we explore the role of behavioral biases in sleep choice. Here the @iza_bonn DP iza.org/publications/dp/1277โฆ Joint with @MalloryAvery16 and Peiran Jiao.
2
13
58
And now not behind a paywall. scholar.harvard.edu/files/seโฆ
Long slog but finally our paper on machine learning applied to bail decisions
nber.org/papers/w23180
@jure @hima_bindu
2
24
55
Making papers accessible in this way is such a service to the profession!
Does alleviating a worker's financial strain allow them to better focus at work and therefore be more productive?
From @supKaur (@berkeleyecon), @m_sendhil (@ChicagoBooth), @suannaoh (@PSEinfo), @FrankSchilbach (@MITEcon):
piped.video/-ZjijA4GVLk
3
6
56
A great opportunity for anyone interested in social applications of Machine Learning.
Last year was a real joy, despite it being virtual. This year might end up being even more fun.
(1/2) Our Center is hosting its Machine Learning in Economics Summer Institute this summer!
- Open to 2nd year+ Ph.D. students and Jr faculty
- Leading faculty in ML, Econ, and Policy
- New methods, new tools, and big questions!
Apply by May 18 here: tinyurl.com/fa8uhxvr
1
12
53
My secret wish for our new paper (w/ Jon Kleinberg & @manish_raghavan): that we had titled it
โMy relationship with Doritos (Bonus: may explain your love/hate of social media)โ
๐งตto explain what connects Doritos, machine learning, and user data.
bit.ly/39ZEZn3 1/N
2
10
55
Hiring a few awesome people to build something delightful and impactful that combines behavioral science and technology.
Come learn, work and laugh with us.
Or pass this on to someone who will. nitter.app/bettertheapp/status/10โฆ
3
25
52
Why are people suddenly googling "iphone slow"? My latest @NYtimes @UpshotNYT column. nyti.ms/1nHuFyB
12
52
43
2/2: If you see a regulation that is keeping us from fighting the coronavirus effectively please tell us here:
pauseregulations.com/
4
21
52
This was such a fun conversation to have - even if it is 50% as fun to listen to, it's worth the listen!
(also @aleks_madry - great podcast title!)
Our next (and final for this season) guest on the Before AGI podcast is @m_sendhil ! We discuss algorithmic bias, AI in healthcare, and AI in public policy, as well as his journey from computer science to behavioral economics (and back!).
[Due to technical difficulties, this episode is audio-only.]
2
8
49
15,212
1/2
Fresh starts, temptation bundling, want vs should...
Some of my favorite behavior change ideas come from @katy_milkman.
Her new book (out today!) is a must-read: insightful and well-written
amazon.com/dp/B08KPFPSNS/
2
9
47
Are you working on AI and behavior change?
Submit to the AAAI workshop on it: bit.ly/3ctxphk #AAAI2021 #ai4bc Deadline: 11/9/20
Invited speakers include @Susan_Athey @munmun10 Eric Tchetgen, Jon Kleinberg (& me!)
2
17
44
Interesting thread for prospective econ PhD students nitter.app/dclingi/status/9072429โฆ
2
12
45
Amazing - and tragic- finding
Using GPS data that tracks millions of smartphone users, we document voting wait times in polling places around the US during the 2016 Pres Election. We find a large racial gap: voters in all-black neighborhoods wait 29% longer than those in all-white areas (see figure below).
18
43
Fighting poverty w/ unlikely data including from space-aka can you spot north korea here? nyti.ms/1RCA4Tn
2
61
44
SAY THE NAME
cc: @Michael_Levin @dmorey
1
1
36
15,003
Outrageous. Reminded of this excellent paper advances.sciencemag.org/contโฆ
"Our results suggest that the origins of public misinformedness and polarization are more likely to lie in the content of ordinary news or the avoidance of news altogether as they are in overt fakery."
Sinclair's local TV stations across the country are set to air the discredited "Plandemic" researcher's conspiracy theory about Dr. Fauci this weekend. The segment that is set to air suggests Fauci is responsible for the creation of the coronavirus. cnn.com/2020/07/24/media/sinโฆ
2
8
43
Must read for all economists, not just those who study race.
Great to economic thinking being updated to recognize the social construction of racial categories.
Here's a short new paper on what constructivist theories of race mean for empirical discrimination research
ekrose.github.io/files/constโฆ
I survey a huge literature from (mostly) outside econ about racial perceptions and discrimination law on the way to two basic points...๐งต๐
9
44
This is such a good question - even if put aside the huge social consequences, pure self-interest seems to suggest landlords shouldn't evict.
What's the bottleneck to the Coase theorem here? @CassSunstein @R_Thaler
Here's what I don't understand about the eviction wave that's underway. Especially in weaker housing markets, where everybody's losing their income, what's the incentive not to work with a tenant? Are landlords sure they can find another?
5
3
44
Obama has darker skin in attack ads linking him to crime
Bias in the Flesh poq.oxfordjournals.org/cgi/cโฆ My collab w/ @SolomonMg on #Racial #Discrimination in #Election #Campaigns
5
63
41
Wonderful article
"Povertyย is not so much about the choices people make, but rather what they find themselves forced to choose between." - @JoWolffBSG newstatesman.com/2019/11/mytโฆ
11
41
9/ As a field, we put a lot of energy into hypothesis testing: thoughtful data collection, meticulous experiments, creative natural experiments.
But where do these hypotheses come from?
We think ML can improve this pre-scientific stage of our work. /end
2
4
41
10,249
This is not quite a cash transfer but it can serve the same purpose for small business @AndrewYang @PaulFNiehaus #UBI
nytimes.com/2020/03/19/businโฆ
1
16
37
My latest @NYTimes column tackles the uncontroversial topic of discrimination
nyti.ms/1vHpLBA
16
50
40
Always so hard to judge which parts of our preferences are reflective values and what parts are simply automatic biases....
Consuming brands instead of generics also seems like interesting case of status quo bias. . .but what if people derive substantial utility from believing brands are better? @M_Sendhil nyti.ms/2kcj55
4
18
38
Poverty changes how you answer this question. Nice summary of research with Shah, @jiayingzhao and Shafir bit.ly/2sHRJsj
18
36
Inspiring...wish I'd done something this good while in grad school! immigrantdoctors.org/
1
15
39
Astounding creativity here sadanduseless.com/recreated-โฆ
5
34
3/3: Many people missed the most hopeful part of the paper: the original manufacturer was eager to fix the problem and has made progress. I believe others will be equally as well.
1
5
34
2/3: Thanks to @ChicagoBooth and contributions to our (soon to be launched) Center for Applied AI, we have the resources to do this pro bono, as a social good.
6
35
Video is great -including nice illustration of DRM (false memories) amongst the poor. newsy.com/stories/poverty-s-โฆ
18
36
Stop worrying about top 1% and start worrying about the bottom 20% My @nytimes column. nyti.ms/1fODhAZ
10
52
37
A terrific list
Algorithms are used differently than human decision makers:
- people assume algorithms are objective or error-free
- algorithms more likely to be implemented with no process for recourse
- algorithms used at scale
- algorithms are cheap
read more: fast.ai/2018/08/07/hbr-bias-โฆ
ALT People are more likely to assume algorithms are objective or error-free (even if theyโre given the option of a human override) Algorithms are more likely to be implemented with no appeals process in place. Algorithms are often used at scale. Algorithmic systems are cheap.
6
34
This is so moving. Powerfully brings home how unforgiving life can be when you are not economically privileged
When you're in a hole, why would you dig yourself deeper? This week: how scarcity affects our lives. #trypod
20
34
Very sympathetic to the pain of data access - limited access serves neither science nor the public interest. I'm part of a team (w/ @khay @oziadias +others) to launch an open resource that makes data available to all researchers. They will announce a launch soon!
I think youโre right, but Sendhil underplays that aspect of his research to the point of ignoring it entirely. Also, that part of things generally never gets written up in the final published paper, so techniques for doing it effectively are not well disseminated.
5
35
7/ What did we find? Well, read the @BeckerFriedman write up here: bfi.uchicago.edu/insight/finโฆ
or the paper here:
nber.org/papers/w31017
1
6
32
7,143
It's terrific that there's so much socially responsible research happening at business schools and that @FT is recognizing it
๐We are excited to have been one of the recipients of the @FT Responsible Business Education Awards 2022!
Our work highlights "large-scale racial bias in the algorithms used by insurers to predict millions of US patientsโ healthcare needs."
Read more: on.ft.com/3FKenR1
3
33
Great post on how behavioral science can and importantly CANT promote healthy behavior bit.ly/2lcToSz @lorimelichar
21
29
Very nice writeup about scarcity bit.ly/1d9t82P in @harvardmagazine
2
23
33
What the fight against Ebola can teach us about how to fight covid19 nyti.ms/3altTUn
(Includes surprising findings from this RCT:
odube.net/wp-content/uploadsโฆ)
@BilalSiddiqi @jhaushofer
1
15
30
A bit of clarification on potential confusion caused by my article. @noahpinion @rajivatbarnard @jwdink
3
14
29
8/ Why we are excited about this?
First, this is a disturbing fact about judges. A personโs face shouldn't matter at all; and definitely not this much.
Second, the technique is more broadly useful: anywhere there is rich dimensional data about human behavior.
4
1
28
8,664
Are the poor more rational? In one important sense, yes bit.ly/1CAGDAh
5
42
30
Cool - a good reminder that each day (and life) is finite.
The irony of me tweeting this from my computer is not lost on me.
We found this great tool that nudges you to do something other than waste time on your computer. Install it in Chrome - whenever you open a new tab, it tells you how much sunlight is left in your day: sunshine.fyiย Very nice, @bnchrchย ๐
2
6
30
This is a great chance to work on a project that combines behavioral science, experimentation and machine learning. Not to mention with some awesome people in @katy_milkman @lyleungar and @jannspiess
BCFG has a job opening for a postdoc with background in machine learning or econometrics. For more please visit bit.ly/2pZLk7G
11
25
We could use your help to a good cause.
Collecting field studies on racial discrimination against Blacks in United States. Focus is on those using randomization or pseudo-randomization.
Some (tragic) examples already here.
What are we missing?
Colleagues and I are trying to collect recent examples of racial discrimination that have been documented in the field using quasi-random (eg correspondence studies) methods. Especially interested in examples outside of economics. Any suggestions? Thanks!
@m_sendhil
12
27
If you're knowledgable about NLP are looking to work on a project that can really impact society this is a great opportunity...
Interested in connecting NLP/text analysis to research on social policy and inequality? Come work with us! (@aadukia @alexeble @UChicago_RCC @DoraSzasz @emileigharrison) bit.ly/3kVhe00
ALT NLP/Text Analysis Computational Scientist Job Posting based at University of Chicago (page 1 of 2). Interested in connecting NLP/text analysis to research on social policy and inequality? Come work with us! (@aadukia @alexeble @UChicago_RCC @DoraSzasz @emileigharrison) https://bit.ly/3kVhe00
ALT NLP/Text Analysis Computational Scientist Job Posting based at University of Chicago (page 2 of 2). Interested in connecting NLP/text analysis to research on social policy and inequality? Come work with us! (@aadukia @alexeble @UChicago_RCC @DoraSzasz @emileigharrison) https://bit.ly/3kVhe00
1
9
26
Looks like a cool paper - excited to read. The idea that scarcity shifts attention and so can have *positive* and negative effects is important. It showed up in these papers too but is often overlooked
science.sciencemag.org/conteโฆ
journals.sagepub.com/doi/absโฆ
New NBER working paper presenting causal evidence of scarcity on decision-making w/ Guenther Fink & Kelsey Jack #endowmenteffect #scarcity
๐dietmarfehr.github.io/researโฆ
A summary in 9 tweets๐ Spoiler: Scarcity *improves* decision-making @CFCamerer @alexoimas @TimHarford @m_sendhil
2
7
27
Excited that the Journal of Economic Perspectives did a "symposium" on criminal justice! E
Even more excited that the editors clearly worked to include more diverse perspectives than one typically finds in economics!
@AEAjournals @TimothyTTaylor
7
27
Had a great time on @NickMHobson 's podcast talking behavioral science apps. He's as interesting as his mustache would lead you to believe.
The other Pique (@getpiqueapp) cofounders were there too-Mike Norton & @1becweeks
behaviorist.biz/bspodcast/piโฆ
1
4
26
Behavioral finance pioneer Terry Odean teaching online course on money management: bit.ly/makingsmart
1
15
26
2/2 Faculty includes
Susan Athey @Susan_Athey
Josh Blumenstock @jblumenstock
Amitabh Chandra @amitabhchandra2
Victor Chernozhukov @VC31415
Matt Gentzkow
Avi Goldfarb @avicgoldfarb
Guido Imbens
& organizers Ziad Obermeyer @ziadoo Jann Spiess @jannspiess + me and Jens Ludwig
2
3
24
I think this might be our most careful paper on this question bit.ly/2Di6mqE
1
13
23
Excited to meet the next generation of computational social scientists!
Apply if you haven't already...
Super excited to announce that @m_sendhil will visit the Summer Institute in Computational Social Science this summer @DukeU alongside @duncanjwatts @davidlazer @dfreelon Cynthia Rudin, and others. Only 15 days left to apply: bit.ly/2nFDTkn
8
24
Are we giving patients choices or burdening them with choices?
NEW: In doctors' offices, patients are not in the best position to make good decisions - evidence & implications
papers.ssrn.com/sol3/papers.โฆ
1
14
20