Monday, August 17, 2020

College admissions in the UK during the pandemic: exams replaced by teacher assessments, algorithms, appeals and disorder

Many exams around the world have been cancelled because of corona virus. For college admissions, which operates on a strict schedule (despite the facts that college is going remote in many places), exam scores have been estimated by teachers and/or algorithms. (That is to say, exams, which are noisy measures of aptitude and accomplishment, are being replaced by noisy predictions of exam performance...)

Wired summarizes the situation in the UK, where exam results are especially salient, since universities issue acceptances conditional on grades:

An Algorithm Determined UK Students' Grades. Chaos Ensued  by Amit Katwala

"This year's A-Levels, the high-stakes exams taken in high school, were canceled due to the pandemic. The alternative only exacerbated existing inequities."

...

"The coronavirus pandemic means exams were canceled and replaced with teacher assessments and algorithms. It has created chaos.

"In Scotland, the government was forced to completely change tack after tens of thousands of students were downgraded by an algorithm that changed grades based on a school’s previous performance and other factors. Anticipating similar scenes for today’s A-level results, the government in England has introduced what it’s calling a ‘triple lock’—whereby, via stages of appeals, students will effectively get to choose their grade from a teacher assessment, their mock exam results, or a resit to be taken in the autumn.

*********

So, chaos ensued. The Guardian gives us a taste:

Ofqual 'blindsided' government by revoking A-level appeals processMove by regulator could leave thousands of students without university places by Richard Adams and Heather Stewart

"The government’s attempts to fix this summer’s broken A-level and GCSE results have been thrown into chaos after the exam regulator Ofqual revoked the appeals process for schools and students that it had published only a few hours earlier.

"The move, which was caused by internal disagreements over the appeals process, could leave thousands of A-level candidates in limbo, including those trying to secure places at universities this autumn by overturning the grades awarded to them by Ofqual’s algorithm.

"The guidance was revoked late on Saturday night, less than eight hours after it was published, leaving school leaders and exam boards bewildered.

"Insiders said Ofqual had “blindsided” the Department for Education, after ministers had endorsed the guidance just two hours earlier.

...

"The controversy is set to deepen on Thursday when GCSE results are published for more than 600,000 year 11 pupils in England. As many as 2m assessments submitted by teachers could be downgraded by Ofqual’s model, which relies on a school’s past performance and each pupil’s previous results from primary school.

"Analysis published by the Observer found that 97% of GCSE results will be allocated solely by Ofqual’s algorithm rather than teacher-assessed grades submitted by schools and colleges. Teacher rankings of pupils will be used to distribute grades within each school."

*********

Also from the Guardian:

A-levels row: Oxford college to honour all offers despite resultsWorcester College move comes as thousands of Oxbridge alumni call on others to follow suit.  by Aaron Walawalkar

"An Oxford college said it will accept all students with offers regardless of their A-level results, as thousands of Oxbridge alumni call on others to show equal “kindness and generosity” to downgraded pupils.

...

"It comes after Oxbridge applicants who received lower grades than expected on Thursday were told they may have to wait a year before they can start degree courses after successfully appealing against their results.

"About 39% of A-level results were downgraded by the exam regulator Ofqual’s algorithm, with disadvantaged pupils worst affected.

"Ashe told BBC Radio 4: “For us, at a very simple level, it was clear to us before results came out that these students were not going to have sat exams and therefore we took the view that there wasn’t going to be any new information there that could justify rejecting someone to whom we’d made an offer."


Sunday, August 16, 2020

4th Workshop on Mechanism Design for Social Good, August 17-19, 2020 (MD4SG )'20

 This week: 

4th Workshop on Mechanism Design for Social Good, August 17-19, 2020 (MD4SG '20)

"The fourth workshop on Mechanism Design for Social Good (MD4SG '20) will be held online on August 17-19, 2020. The goal of the workshop is to highlight work where techniques from algorithms, optimization, and mechanism design, along with insights from other disciplines, have the potential to improve access to opportunity for historically underserved and marginalized communities. The workshop will feature keynote presentations, contributed talks, problem pitches and demos, a poster session, and a panel discussion, with a focus on bridging research and policy. To this end, participants will include researchers as well as practitioners in various government and non-government organizations and industry."

Here's the program:

Monday, August 17

 Time Event Authors

10:00-10:15 EST Opening Remarks Program Chairs: Faidra Monachou and Francisco Marmolejo-Cossío

10:15-11:00 EST Keynote: Maximizing the Social Good: Markets without Money Nicole Immorlica

11:00-11:45 EST Session 1: Education Policy and Diversity

11:00-11:15 EST Top Percent Policies and the Return to Postsecondary Selectivity Zachary Bleemer

11:15-11:25 EST Unequal Assignment to Public Schools and the Limits of School Choice Mariana Laverde

11:25-11:35 EST Who Gets Placed Where and Why? An Empirical Framework for Foster Care Placement Alejandro Robinson Cortes

11:35-11:45 EST Explainability in Matching Mechanisms with Diversity Goals: A Case Study of Ethiopian Universities Basiliyos Betru, Meareg Hailemariam and Rediet Abebe

11:45-12:00 EST Short Break

12:00-12:45 EST Keynote Anjana Rajan

12:45-13:30 EST Session 2: Technology, Law and Policy

12:45-13:00 EST Feminicide & Machine Learning: Detecting Gender-based Violence to Strengthen Civil Sector Activism Catherine D'Ignazio, Helena Suarez Val, Silvana Fumega, Harini Suresh, Isadora Cruxen, Wonyoung So, Maria De Los Angeles Martinez and Mariel Garcia-Montes

13:00-13:10 EST The Gender Panopticon: AI, Gender, and Design Justice Sonia Katyal

13:10-13:20 EST Privacy as Privilege Rebecca Wexler

13:20-13:30 EST "Computer says no!": Unpacking the Human in the Loop Requirement in the Context of Welfare Fraud Doaa Abu Elyounes

13:30-14:00 EST Long Break

14:00-15:00 EST Session 3: Labor Markets

14:00-14:15 EST The Role of Referrals in Inequality, Immobility, and Inefficiency in Labor Markets Lukas Bolte, Nicole Immorlica and Matthew Jackson

14:15-14:30 EST All Things Equal? Social Networks as a Mechanism for Discrimination Chika Okafor

14:30-14:40 EST Outside Options, Bargaining, and Wages: Evidence from Coworker Networks Sydnee Caldwell and Nikolaj Harmon

14:40-14:50 EST The Geography of Unemployment Adrien Bilal

14:50-15:00 EST Location Sorting and Endogenous Amenities: Evidence from Amsterdam Milena Almagro and Tomas Dominguez-Iino

15:00-16:00 EST Poster Session

16:00-17:00 EST Networking Session 


Tuesday, August 18

Time Event Authors

10:00-10:45 EST Keynote: Tech in Support of Caregiving: Innovation Opportunities and Ecosystem Challenges Deborah Estrin

10:45-11:20 EST Session 1: Environment, Agriculture and Food Consumption

10:45-11:00 EST Improving Farmers' Income on Online Agri-platforms: Theory and Field Implementation of a Two-stage Auction Retsef Levi, Manoj Rajan, Somya Singhvi and Yanchong Zheng

11:00-11:10 EST Optimal Interventions for Increasing Healthy Food Consumption Among Low-Income Populations Elisabeth Paulson, Retsef Levi and Georgia Perakis

11:10-11:20 EST Simple and Approximately Optimal Contracts for Payment for Ecosystem Services Wanyi Li, Irene Lo and Itai Ashlagi

11:20-12:00 EST Long Break

12:00-12:45 EST Keynote Natalia Aríza Ramirez

12:45-13:25 EST Session 2: Education in Practice

12:45-13:00 EST School Choice in Chile Natalie Epstein, Jose Correa, Rafael Epstein, Juan Escobar, Ignacio Rios, Nicolas Aramayo, Bastian Bahamondes, Carlos Bonet, Martin Castillo Quintana, Andrés Cristi, Boris Epstein and Felipe Subiabre

13:00-13:15 EST Competition under Social Interactions and the Design of Education Policies Claudia Allende

13:15-13:25 EST From Pipeline to Pipelines: How Multiple Definitions of CS Education Distort CS Enrollment Data Stephanie Tena-Meza, Aj Alvero and Miroslav Suzara

13:25-14:00 EST Long Break

14:00-15:00 EST Panel Discussion (in Spanish, with live captioning) Natalia Aríza Ramirez, José R. Correa, Rafael Obregón

15:00-16:00 EST Session 3: Healthcare

15:00-15:15 EST Large-scale clinical trial of an AI-augmented intervention for HIV prevention in youth experiencing homelessness Bryan Wilder, Laura Onasch-Vera, Graham Diguiseppi, Robin Petering, Chyna Hill, Amulya Yadav, Eric Rice and Milind Tambe

15:15-15:30 EST Heterogeneous Donor Circles for Fair Liver Transplant Allocation Shubham Akshat, Sommer Gentry and S. Raghavan

15:30-15:40 EST Predicting no-show appointments in a pediatric hospital in Chile using machine learning Hector Ramirez, Fabian Villena Rodriguez, Jocelyn Dunstan, Victor Riquelme, Juan Pablo Hoyos, Javier Madariaga and Juan Peypouquet

15:40-15:50 EST Socioeconomic Network Heterogeneity and Pandemic Policy Response Mohammad Akbarpour, Cody Cook, Aude Marzuoli, Simon Mongey, Abhishek Nagaraj, Matteo Saccarola, Pietro Tebaldi, Shoshana Vasserman and Hanbin Yang

15:50-16:00 EST The Consequences of Medicare Pricing: An Explanation of Treatment Choice Elena Falcettoni

16:00-17:00 EST Networking Session 


Wednesday, August 19

Time Event Authors

10:00-10:45 EST Keynote Stephanie Dinkins

10:45-11:30 Session 1: Fairness and Inequality

10:45-11:00 EST Measuring Non-Expert Comprehension of Machine Learning Fairness Metrics Debjani Saha, Candice Schumann, Duncan C. McElfresh, John P. Dickerson, Michelle L. Mazurek and Michael Carl Tschantz

11:00-11:10 EST Pipeline Interventions Eshwar Ram Arunachaleswaran, Sampath Kannan, Aaron Roth and Juba Ziani

11:10-11:20 EST Pricing with Fairness Maxime Cohen, Adam Elmachtoub and Xiao Lei

11:20-11:30 EST Public Transit Access and Income Segregation Prottoy Akbar

11:30-12:30 EST Poster Session

12:30-13:15 EST Session 2: Algorithms for Policy and Governance

12:30-12:45 EST Modeling Assumptions Clash with the Real World: Configuring Student Assignment Algorithms to Serve Community Needs Samantha Robertson, Tonya Nguyen and Niloufar Salehi

12:45-12:55 EST The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies Stephan Zheng, Alex Trott, Sunil Srinivasa, Nikhil Naik, Melvin Gruesbeck, David Parkes and Richard Socher

12:55-13:05 EST (Machine) Learning what Governments Value Daniel Bjorkegren, Joshua Blumenstock and Samsun Knight

13:05-13:15 EST Economic Method, Digital Platform: When Mechanism Design Moves Online Salome Viljoen, Jake Goldenfein and Lee McGuigan

13:15-13:45 EST Long Break

13:45-14:30 EST Session 3: Online Platforms and Civic Participation

13:30-13:45 EST Advertising for Demographically Fair Outcomes Lodewijk Gelauff, Ashish Goel, Kamesh Munagala and Sravya Yandamuri

13:45-13:55 EST Neutralizing Self-Selection Bias in Sampling for Sortition Bailey Flanigan, Paul Gölz, Anupam Gupta and Ariel Procaccia

13:55-14:05 EST Auditing Digital Platforms for Discrimination in Economic Opportunity Advertising Sara Kingsley, Clara Wang, Alexandra Mikhalenko, Proteeti Sinha and Chinmay Kulkarni

14:05-14:15 EST Online Policies for Efficient Volunteer Crowdsourcing Vahideh Manshadi and Scott Rodilitz

14:15-14:30 EST Short Break

14:30-15:10 EST Session 4: Information

14:30-14:40 EST Information Design for Congested Social Services: Optimal Need-Based Persuasion Jerry Anunrojwong, Krishnamurthy Iyer and Vahideh Manshadi

14:40-14:50 EST How to get-toilet-paper.com? Provision of Information as a Public Good Robizon Khubulashvili, Mallory Avery, Kristi Bushman, Alexandros Labrinidis, Sera Linardi and Konstantinos Pelechrinis

14:50-15:00 EST Responsible Sourcing: The First Step Is the Hardest Pia Ramchandrani, Hamsa Bastani and Ken Moon

15:00-15:10 EST Anticipation and Consumption Neil Thakral and Linh To

15:10-15:30 EST Closing Remarks & Award Announcements Faidra Monachou and Francisco Marmolejo-Cossío

15:30-17:00 EST Networking Session 

The full program of the workshop will take place on August 17-19, 2020, on Gather.town and Zoom. Please register at this link to attend the talks and other events

Saturday, August 15, 2020

Social distancing during the plague

 The Washington Post has the story (as a tourist destination...)

Restaurants in Italy are reopening ancient ‘wine windows’ used during the plague

The coronavirus pandemic has inspired a comeback for the socially distant invention.

By Natalie B. Compton Add to list

"In the 1300s, Europeans lived in fear of the plague claiming lives across the continent. In 2020, Europeans face the threat of the coronavirus, a pandemic that has killed more than 728,000 worldwide.

..."Between the wrath of the Black Death in the 1300s and the Italian Plague in the 1600s, wine merchants in the Italian region of Tuscany built “wine windows” to protect buyers and sellers from coming into close contact."

"The socially distant precaution was ahead of its time."


Friday, August 14, 2020

Should residency program rank order lists be kept confidential from the Dean?

Here's the report of a survey of residency program directors in radiology. One issue, not confined to radiology, is the confidentiality of their rank order list for the resident match--confidentiality from their own administrative hierarchy.  The problem with having to show your rank order list to your dean is that it interferes with program directors' incentives to rank candidates in order of true preferences:   Thirty-seven percent felt pressure to match applicants from the top of the rank list in order to improve the perceived “success” in the match."  That is, some of these programs are refraining from ranking the most desirable applicants they interviewed because they worry these people will match to other programs.  This will make their program look bad to the dean (who will ask "how come you have to go so far down on your list?")

“What Program Directors Think” V: Results of the 2019 Spring Survey of the Association of Program Directors in Radiology (APDR) Academic Radiology,  8 August 2020, In Press, Corrected Proof

by Anna Rozenshtein MD, MPH1 Brent D. Griffith MD2 Priscilla J.Slanetz MD, MPH3 Carolynn M.DeBenedectis MD4 Jennifer E.Gould MD Jennifer R.Kohr MD6 Tan-Lucien Mohammed MD, MS7Angelisa M.PaladinMD8Paul J.Rochon MD9 Monica Sheth MD10Ernest F.Wiggins III MD11 Jonathan O.Swanson MD12   Academic Radiology Available online 8 August 2020, In Press, Corrected Proof 

"The Association of Program Directors in Radiology (APDR) surveys its membership annually on hot topics and new developments in radiology residency training. Here we report the results of that annual survey.

...

"Radiology Residency Match: Forty-nine percent of respondents reported that the final rank list is known only to the program administration (PD/APD) and the selection committee, while 27% disclosed the rank list to the department administration and 24% to the institution. Thirty-seven percent felt pressure to match applicants from the top of the rank list in order to improve the perceived “success” in the match."

Thursday, August 13, 2020

Gender distribution of medical residents

 Surgeons are still less likely to be women (particularly orthopedic surgeons):

Bennett CL, Baker O, Rangel EL, Marsh RH. The Gender Gap in Surgical Residencies. JAMA Surg. Published online July 29, 2020. doi:10.1001/jamasurg.2020.2171



HT: Irene Wapnir

Wednesday, August 12, 2020

The end of the beginning of the gig economy? Uber, Lyft and California AB5

 The San Francisco Chronicle has the latest news on the most concrete step yet to require Uber and Lyft to shift from a contractor-driver business model to one of driver employees...

Judge says California Uber, Lyft drivers should be employees

by Carolyn Said 

"A San Francisco Superior Court judge on Monday granted California’s request for a preliminary injunction to make the state’s Uber and Lyft drivers into employees. The 34-page order was scathing about the ride-hailing companies’ “prolonged and brazen refusal to comply with California law,” namely AB5, the new gig-work law that makes it harder for companies to claim that workers are independent contractors.

...

"However, it is likely to have little immediate impact.

"Judge Ethan Schulman stayed his injunction for 10 days. The companies will appeal it and seek a longer stay before the 10 days are up. An appeals court likely would hear their emergency motion quickly. Uber said it expects to be granted the longer-term delay and does not anticipate any near-term changes to its business. No matter what, it could not hire tens of thousands of drivers in a matter of days, it said.

...

"AB5 established an ABC test that says workers are employees unless A) they are free from a hiring entity’s control, B) perform work outside the hiring entity’s usual business, and C) have an independent business doing that kind of work.

...

"Along with other gig companies, Uber and Lyft are pursuing a $110 million November ballot measure, Proposition 22, asking California voters to keep drivers as freelancers who are entitled to some earnings guarantees and benefits. DoorDash, Instacart and Postmates, the other Prop. 22 backers, are not named in the California lawsuit but presumably would be affected by whatever precedent it sets. (Uber has purchased Postmates in a deal that will close next year.)"

Tuesday, August 11, 2020

Transplanting two kidneys from older deceased donors to reduce discards

Here's a venerable but newly fashionable idea in transplantation, particularly for deceased donors whose (individual) kidneys have already been rejected when offered.  Transplant both kidneys into the same recipient, to reduce the risk.

Here's a paper from January:
Lee, K.W., Park, J.B., Cha, S.R. et al. Dual kidney transplantation offers a safe and effective way to use kidneys from deceased donors older than 70 years. BMC Nephrol 21, 3 (2020). https://doi.org/10.1186/s12882-019-1664-8


And here's a very recent news article talking about a different, larger study:

By Melissa J. Webb

"Using data from the United Kingdom Transplant Registry, they identified 7,841 kidneys procured from deceased donors aged 60 years or older, finding that 17% of these were discarded.

Considering the remaining kidneys used for transplant (356 for dual; 5,032 for single), the researchers determined that both donors and recipients of dual transplants were older (median, 73 years vs. 66 years and 64 years vs. 61 years, respectively). Donors of kidneys used in dual transplantation also had higher United States Kidney Donor Risk Indices (2.48 vs 1.98 for those used in single transplants).

After adjusting for confounders, the researchers observed similar 5-year graft survival between dual and single transplants (HR = 0.81), as well as a higher median eGFR at 12 months for recipients of dual transplants (40 mL/min/1.73m2 vs. 36 mL/min/1.73m2)."

Monday, August 10, 2020

Reputation among thieves: ransomware and kidnapping

Like everyone else, I occasionally get notifications of data breaches from organizations with which I have digital relations.  Often the breach involved a third party.  Sometimes the breach involves the theft of data accompanied by a demand of ransom--i.e. the victim is invited to pay the cybercriminal, who then promises to destroy the data instead of selling it on the dark web or otherwise using it.

This bears some resemblance to the kidnapping business, and its high-seas version, piracy.

Here's part of an email I recently received informing me of such a breach, and subsequent payment of ransom.

"I’m writing to inform you that Blackbaud, the company that hosts [xxx’s] relationship management system, suffered a security incident in May. Blackbaud is the world’s largest provider of fundraising technology for non-profits and educational institutions, and many organizations have been impacted by this incident.
...
"We were also informed by Blackbaud that in order to protect data and mitigate potential identity theft, it met the cybercriminal’s ransomware demand. Blackbaud has advised us that it received assurances from the cybercriminal and third-party experts that the data was destroyed. Blackbaud has been monitoring the web in an effort to verify the data accessed by the cybercriminal has not been misused. "
************
Why should "assurances from the cybercriminal" be reassuring? (and for how long?).  And what are the roles played by "third-party experts"?

My guess is that, as in the kidnapping biz, intermediaries have emerged to conduct the negotiations, get some sort of assurances, and make it possible for criminal organizations to maintain reputations for honor among thieves.

It is of course possible to regard ransom paying as a repugnant transaction that facilitates ransomware, kidnapping, etc.  In fact the U.S. for some time made it a crime to pay ransom to kidnappers, but relaxed that view over time, as kidnapping became a bigger international business, and there was often a considerable desire (sometimes covered by insurance) to pay ransom when it seemed the best way to recover the kidnapped person alive.

Here are some related posts which touch on that story:

Monday, June 24, 2019  Kidnapping insurance

Tuesday, September 13, 2016 Ransom as a (not so) repugnant transaction

Monday, August 9, 2010 Brokers for pirate ransom

Saturday, December 5, 2009 Market for kidnapping

Sunday, November 30, 2008 Pirate ransom: counterparty risk

Sunday, August 9, 2020

Experimental Economics SITE conference at Stanford, by Zoom, Aug 10 and 11, 2020

 Tomorrow and Tuesday, here's the SITE Experimental Economics conference schedule (Mon Aug. 10 and Tuesday Aug. 11).  


Muriel Introduce SITE confernece
SESSION 1: 9am - 10:30am pacific (12pm eastern, 6pm europe)Lise - in charge
1Overriding in Teams: The Role of Beliefs, Social Image, and GenderMaria Recalde, University of Melbourne(Lise Q)
2Information and the Persistence of the Gender Wage Gap: Early Evidence from California's Salary History BanBenjamin Hansen, University of Oregon(Lise Q)
3Attention as Human CapitalHeather Schofield, University of Pennsylvania(Colin Q)
10:30am - 11am pacific: Break/Discussion (1pm eastern, 7pm europe)
SESSION 2: 11 -12:30pm pacific (2pm eastern, 8pm europe)Muriel - in charge
4Social Learning in Groups: an Experimental StudyMarina Agranov, Caltech(Muriel Q)
5Beliefs in Repeated GamesGuillaume Frechette, NYU(Muriel Q)
6Beyond Ordinal: The Value of Indifferences and Cardinal Information in MatchingClayton Featherstone, University of Pennsylvania(Muriel Q)
12:30pm - 1pm pacific: Break/Discussion (3:30pm eastern, 9:30pm europe)
Day 2 (August 11)
SESSION 3: 9:30am -10:30pm pacific (12:30pm eastern, 6:30pm europe)Lise - in charge
7The Burden of Holding DebtAlejandro Martínez-Marquina, Stanford University
8Claiming Credit: Gender, Memory, and Social NormsJonas Mueller-Gastell, Stanford University
9It’s Not my Fault: Excuse-Seeking Behavior in the Intertemporal DomainMarissa Lepper, University of Pittsburgh
10Do Actions Speak Louder than Motives? Evaluating the Effectiveness of Image-FundraisingPun (pronounced like "Poon") Winichakul, University of Pittsburgh
10:30am - 11am pacific: Break/Discussion (1pm eastern, 7pm europe)
SESSION 4: 11 -12:30pm pacific (2pm eastern, 8pm europe)Muriel - in charge
11Fairness Across the World: Preferences and BeliefsAlexander W. Cappelen, Norwegian School of Economics(Christine Q)
12Cognitive Flexibility or Moral Commitment? Evidence of Anticipated Belief Distortion
Silvia Saccardo, Carnegie Mellon University(Christine Q)
13Digital AddictionHunt Allcott, New York University and Microsoft Research(Colin Q)
Muriel close SITE confernece
12:30pm - 1pm pacific: Break/Discussion (3:30pm eastern, 9:30pm europe)

Saturday, August 8, 2020

Is convalescent plasma useful for treating covid-19?

The reported results on convalescent plasma are so far still quite incomplete, and mixed.  If I had to summarize, I'd say that a growing body of evidence suggests that treating early stage (e.g. just hospitalized) covid-19 patients increases and speeds the chance of recovery, while there is little convincing evidence that convalescent plasma helps more severely ill patients who have begun to have serious complications.

Here is a recent WSJ article:

By Amy Dockser Marcus

"Hospitalized Covid-19 patients who received transfusions of blood plasma rich with antibodies from recovered patients reduced their mortality rate by about 50%, according to researchers running a large national study.
...
"The researchers said they saw signs that the treatment might be working in patients who received high levels of antibodies in plasma early in the course of their illness. They based their conclusions on an analysis of about 3,000 patients."
************

Here's a recent paper in JAMA on a very small randomized trial in China that doesn't find statistically significant effects on patients who 

August 4, 2020
Ling Li, MD, PhD; Wei Zhang, MD; Yu Hu, MD, PhD; Xunliang Tong, MD, PhD; Shangen Zheng, MD; Juntao Yang, PhD; Yujie Kong, MD; Lili Ren, PhD; Qing Wei, MD; Heng Mei, MD, PhD; Caiying Hu, MD; Cuihua Tao, MD; Ru Yang, MD; Jue Wang, MD; Yongpei Yu, PhD; Yong Guo, PhD; Xiaoxiong Wu, MD; Zhihua Xu, MD; Li Zeng, MD; Nian Xiong, MD, PhD; Lifeng Chen, MD; Juan Wang, MD; Ning Man, MD; Yu Liu, PhD; Haixia Xu, MD; E. Deng, MS; Xuejun Zhang, MS; Chenyue Li, MD; Conghui Wang, PhD; Shisheng Su, PhD; Linqi Zhang, PhD; Jianwei Wang, PhD; Yanyun Wu, MD, PhD; Zhong Liu, MD, PhD
  JAMA. 2020; 324(5):460-470. doi: 10.1001/jama.2020.10044

Abstract: This randomized trial compares the effects of convalescent plasma therapy with standard care vs standard care alone on time to clinical improvement among patients with severe or life-threatening COVID-19 disease in China.

"Among patients with severe or life-threatening COVID-19, convalescent plasma therapy added to standard treatment did not significantly improve the time to clinical improvement within 28 days, although the trial was terminated early and may have been underpowered to detect a clinically important difference."
**********

My last donation had high enough antibodies to qualify me for another: I hope these are going to patients for whom they will be useful.

Friday, August 7, 2020

Global kidney exchange between Abu Dhabi and Kerala (India)

Here is an article in the newspaper Malayalam Manorama, in Malayalam, the language spoken in Kerala, about a global kidney exchange between hospitals in Abu Dhabi in the United Arab Emirates, and in Kerala in India.

The url hints at the story: the exchange was between a Kerala hospital and a UAE hospital that both used kidney exchange software provided by Mike Rees's organization, the Alliance for Paired Kidney Donation (APKD), to identify the exchange, which was performed in India:


 
The article says SEHA Kidney Care Staff( Anan Purushothaman, Sheenamma Varghese , Siddiq Anwar) with Dr Mike Rees from Alliance For Paired Donation helped find a compatible  kidney donor in India via the “Global Kidney Paired Exchange”. Dr Feroz Aziz then successfully transplanted the two pairs.

Kim Krawiec, through a friend fluent in Malayalam, gives the following summary:

"The article goes on to say that Najla was in want of a kidney donor. Even though she had 3 of her relatives who were ready to donate none of them were compatible. She was asked to get in touch with the organisation called Alliance for Paired Kidney Donation, where they find donors all around the world using the latest technology. With the help of this organisation and the latest technology, not to mention the doctors and nurses she was able to find a compatible donor. At the same time Najma's mother was able to donate her kidney to the Abu Dhabi donor's husband. Now all are well and back to normal life."


Thursday, August 6, 2020

Is randomization repugnant?


Patrick R. Heck,  Christopher F. Chabris,  Duncan J. Watts, and Michelle N. Meyer
PNAS first published July 27, 2020 https://doi.org/10.1073/pnas.2009030117
Edited by Margaret Levi, Stanford University, Stanford, CA, and approved July 4, 2020 (received for review May 13, 2020)
 
Abstract: We resolve a controversy over two competing hypotheses about why people object to randomized experiments: 1) People unsurprisingly object to experiments only when they object to a policy or treatment the experiment contains, or 2) people can paradoxically object to experiments even when they approve of implementing either condition for everyone. Using multiple measures of preference and test criteria in five preregistered within-subjects studies with 1,955 participants, we find that people often disapprove of experiments involving randomization despite approving of the policies or treatments to be tested.

Wednesday, August 5, 2020

Admissions to polytechnics in Finland, by Kristian Koerselman

Here's a paper that gives a very clear description of the centralized application process for Finnish polytechnics (applied universities), which gives applicants a complicated strategic problem, and results in many applicants re-applying in subsequent years.

Assignments are by a (school proposing) deferred acceptance algorithm, but applicants get extra points in a school's preferences by listing it first, they can only list four programs, and they must choose which exams to take.

Why Finnish polytechnics reject top applicants
Kristian Koerselman
Education Economics, July 2020

"The Finnish polytechnic assignment provides us with an example where applicants are asked to strategize in their applications while having poor prior information on the set of programs that would admit them. These features should in and of themselves already be expected to cause a poor assignment outcome. I highlight the additional role which entrance exams appear to play in creating what should arguably be seen as a misrepresentation of true admission criteria analogous to a misrepresentation of applicants' true preferences. Because applicants take and retake different entrance exams in different years, programs rank the same applicants differently in different years, giving applicants an incentive to reapply even if they were originally assigned to their within-year most preferred feasible program.
...
"Finland provides 9 years of compulsory, comprehensive education, after which almost all students continue in approximately equal proportions to either an academically-oriented high school or to a vocational school. High school concludes with a set of nationally standardized and externally graded matriculation exams. Though students have a reasonable amount of freedom in choosing the subjects they want to take an exam in, they have to take into account that different higher education programs value matriculation exam grades in different subjects differently.

Higher education is provided by polytechnics, also called universities of applied sciences, and by universities. The former mainly offer bachelor programs, and the latter mainly combined bachelor/master programs. About half of each birth cohort ever enrolls in higher education, with total yearly admissions somewhat larger at polytechnics than at universities. Although it is not uncommon for vocational school graduates to apply to a polytechnic, high school graduates are the largest group of polytechnic applicants.
...
"Higher education applications are extremely competitive, with for example only about one-third of polytechnic applicants being admitted nationally each year. Rejected applicants are likely to reapply, often multiple years, and even admitted applicants often reapply. Re-applications are an important reason why the numbers of applicants per seat are so large. Applicants effectively queue into higher education, likely causing them to be admitted at an unnecessarily old age, and therefore also to graduate at an unnecessarily old age. The 2011 polytechnic applicants on which this study is based for example had on average graduated from high school already two and a half years earlier, and many of them would be older still when they were finally admitted to the higher education program they would eventually graduate from.

"All higher education applications are made to a national clearinghouse. Polytechnic admission decisions are generally made centrally by the clearinghouse itself, while university admission decisions are generally not. In this paper, I analyze the 2011 centralized assignment of high school graduates to Finnish polytechnics. In total, 50,894 high school graduates applied to 16,655 seats in 440 programs, divided over 8 fields.

"The application process starts in March, when applicants can apply to up to four programs in order of preference. Applicants must then choose which entrance exams to prepare for and take, typically in May or June. After the entrance exams have been graded, an admission score is calculated for each application. This score is mainly based on applicants' matriculation exam grade point averages and entrance exam results. The weights assigned to different matriculation exam subjects are typically shared within each field, and entrance exams tend to be shared as well. Extra points are awarded for the first listed choice, as well as for factors like relevant labor market experience. The relative weight of the different admission score components in determining the admission score can be seen in Table 2.

"Based on their submitted preference ordering and on their admission scores, applicants are assigned to programs through a centrally run program-proposing deferred acceptance algorithm, each applicant either being admitted to a single program or not being assigned at all. Admitted applicants then either accept their seat or reject it. A much smaller second round of offers is sent out by the programs themselves to make up for first-round rejections. The second round of the process ends at the start of the fall term in September.
...
"Applicants have multiple reasons to strategize in choosing which programs to apply to. Among others, the fact that applicants receive extra points for their first listed choice implies that they will want to list a program first where they have a chance to actually be admitted. Similarly, the four-program limitation means not only that there may be programs acceptable to the applicant which the applicant is not allowed to list, but also that the applicant will need to use the four allowed applications wisely. Third, the applicant faces a strategic choice in which entrance exams to prepare for and take, typically concentrating all effort on a single application. Fourth, the use of a program-proposing algorithm may in and of itself already give applicants an incentive to strategize."
...
"Though applicants receive good indications of their matriculation exam grades before they apply, and may be aware of previous years' admission score cut-offs, they however necessarily learn their entrance exam scores only after choosing where to apply and which entrance exams to take, adding a considerable degree of uncertainty to their application.
...
"When classifying applicants into thirds based on their program-specific matriculation exam GPA, as many as 54% of top third applicants remain unassigned anywhere. Even using the actual admission score, 34% of top third applicants remain unassigned.
...
"Even if applicants do apply to more than one program, their admission chances are relatively low for programs listed second, third and fourth, with the probability of being assigned to a program being 27% for the program listed first, but only between 3 and 4 per cent for programs listed lower. This is partly due to the extra points given for the first listed program, but is probably also related to applicants' strategic choices on which entrance exams to take. "

Tuesday, August 4, 2020

Unbalanced matching markets with short preference lists may not have very unequal stable matchings, by Kanoria, Min, and Qian

Here's a new paper exploring what happens when unbalanced matching markets (with unequal numbers of participants on opposite sides) also have short preference lists, so that each individual has preferences over only a potentially small fraction of possible mates.

Which Random Matching Markets Exhibit a Stark Effect of Competition?
Yash Kanoria, Seungki Min, Pengyu Qian

"We revisit the popular random matching market model introduced by Knuth (1976) and Pittel (1989), and shown by Ashlagi, Kanoria and Leshno (2013) to exhibit a "stark effect of competition", i.e., with any difference in the number of agents on the two sides, the short side agents obtain substantially better outcomes. We generalize the model to allow "partially connected" markets with each agent having an average degree d in a random (undirected) graph. Each agent has a (uniformly random) preference ranking over only their neighbors in the graph. We characterize stable matchings in large markets and find that the short side enjoys a significant advantage only for d exceeding (log^2 n where n is the number of agents on one side: For moderately connected markets with  d=o(log^2 n), we find that there is no stark effect of competition, with agents on both sides getting a sqrt(d)  ranked partner on average. Notably, this regime extends far beyond the connectivity threshold of d=Θ(logn). In contrast, for densely connected markets with d=ω(log2n), we find that the short side agents get logn-ranked partner on average, while the long side agents get a partner of (much larger) rank d/logn on average. Numerical simulations of our model confirm and sharpen our theoretical predictions. Since preference list lengths in most real-world matching markets are much below (log^2 n), our findings may help explain why available datasets do not exhibit a strong effect of competition."

Monday, August 3, 2020

Josh Morrison and health policy activism: kidneys and covid

Here's a profile of Josh Morrison, one of the most interesting health care policy activists I've encountered.  I first met him when he was the general counsel of the kidney exchange organization The Alliance for Paired Kidney Donation, and since then he's created new organizations (with evocative names) and new policies.


"Morrison donated a kidney in 2011, months into his job as a corporate attorney. A few years later he abandoned the law for a more mission-driven career helping people find kidney donors, eventually starting the nonprofit Waitlist Zero in 2014.

"In his telling, his parents “really hated” the idea of being a live organ donor. What he’s planning next terrifies them: Morrison wants to give himself Covid-19 for the sake of science.
...
"The 35-year-old from Brooklyn is the leader of 1Day Sooner, a grassroots organization he co-founded in the spring with a radical idea: Speed up vaccine testing by giving the coronavirus to willing recruits. Including Morrison and his co-founder, 22-year-old Stanford human biology graduate Sophie Rose, more than 30,000 people from 140 countries are signed up — a pool of applicants offering to enlist in what’s known as a human challenge trial.
...
"Human challenge trials involve deliberately infecting small groups of vaccinated volunteers. In a time of social distancing, mask-wearing, and the public’s general leeriness of contracting Covid-19, some researchers, doctors, and ethicists say challenge trials are worthwhile. Unlike traditional Phase 3 clinical trials, which sign up thousands of participants, inject some with a vaccine and others with a placebo, and then wait for people to encounter the virus in everyday life, there’s no waiting on people to catch a virus in a challenge trial. This means it can be completed in weeks instead of months or years, potentially yielding data on vaccine efficacy much more quickly.

"On July 15, human challenge trials for the coronavirus received their biggest endorsement. Adrian Hill, director of the Jenner Institute at the University of Oxford in the U.K., announced that Oxford scientists — already hard at work on a promising coronavirus vaccine — want to launch a challenge trial."

Sunday, August 2, 2020

Foie gras sale ban in CA doesn't rule out interstate commerce

The Mercury News has the lastest twist on the long running saga of foie gras in California:

Foie gras ruling puts it back on Californians’ plates, but not on restaurant menus
Delicacy can be sold, shipped by out-of-state producers; reselling it remains verboten

"Foie gras can again be legally shipped to Californians for consumption at home, according to a new ruling. But diners won’t find the fatty goose and duck livers back on restaurant menus.

"U.S. District Judge Stephen V. Wilson ruled Tuesday in Los Angeles that the sale of foie gras doesn’t violate the law if the seller is located outside of California and the product is brought into the state by a third-party delivery service, the Associated Press reported, adding that the California attorney general’s office is reviewing the decision.

"However, foie gras still can’t legally be resold — which means restaurant sales are prohibited, according to attorneys for both the plaintiffs and defendants.
...
"The issue of whether foie gras can be produced and/or sold in California had been simmering in courts for years.

California’s ban on the production and sale of foie gras (pronounced fwah grah) originally went into effect July 1, 2012, eight years after SB 1520 (by then-Sen. John Burton) was signed into law by then-Gov. Arnold Schwarzenegger, in 2004."
************

See all my posts on foie gras.

Saturday, August 1, 2020

Arranged marriage, in India, on television

The Guardian and the Indian Express have the story about a tv show about arranged marriage, a venerable institution that is becoming controversial.

Here's the Guardian:

Indian Matchmaking: Netflix's 'divisive' dating show causes storm
Series following contestants hoping to be chosen for arranged marriage has divided opinion in India
by Hannah Ellis-Petersen

"For some, Indian Matchmaking represents an unacceptable normalising of the regressive standards forced on Indian women to in order to be seen as a “suitable” wife, while pushing the unspoken issue of caste under the carpet.
...
"The eight-part series follows Taparia as she attempts to find appropriate matches for clients both in India and across the world in order to set up arranged marriages, often on behalf of their client’s parents. It is a show set in a world of upper-class affluence, where Indian families can afford to hire Taparia’s expensive services and even fly her across the world to find them, or their children, a suitable match.

"Arranged marriage remains prevalent in India. As Taparia says in the show, arranged marriage is just described as “marriage” while it is “love marriage” that is spoken of as outside the norm. Newspapers are still full of matrimonial adverts where women are reduced to three-line descriptions of their “fair skinned”, “accomplished” or “modern yet traditional” attributes.

"Indian Matchmaking’s uncritical presentation of its clients’ “criteria” – usually fair-skinned women from a “good” family - has come in for particular criticism.

"Critics have said the show perpetuates damaging ideas around colourism and caste – the Hindu system of hierarchy, which rigidly designates someone’s class and social status. Dalits, India’s lowest class, still undergo rampant discrimination and abuse in society while the upper Brahmin caste hold much of the power and influence. Cross-caste marriage in India can get you killed.

“Indian Matchmaking is really a cesspool of casteism, colourism, sexism, classism,” wrote one Twitter user."
***********

And here's the Indian Express:

Indian Matchmaking: An 8-episode of misguided gender politics, ultimately a betrayal for Indian audiences
By cherry picking its clients and assorting stories it wants to tell, by ticking boxes of caste, religion and class as imperative for an arranged alliance, Indian Matchmaking panders to the West gaze with complying obedience.
by Ishita Sengupta

"Positioned as an outlet to familiarise the world with a practice peculiar to India and Indians; the documentary could have been a first-hand exploration about the evolving origin of a cultural custom and the multifarious ways people go about it. And for millennials back home, it could affirm our rejection of a practice we long recognise as outdated or be a vehicle to convince us of its efficiency in a language we comprehend better than our parents’ monologues. But Indian Matchmaking dilutes an age-old practice by blunting the pointed shards on which it has stood for years. The end result is an eight-episode betrayal for the audience in India and a cut-to-fit documentary about the country and its traditions for the West, confirming every suspicion they nurtured.

"Created by Smriti Mundhra, who previously co-directed A Suitable Girl in 2017, it follows Sima Taparia, one of India’s top matchmakers as she visits her clientele spread across India and abroad. At the very outset, Taparia (“from Mumbai”) insists, “Matches are made in heaven and God has given me the job of making them successful on earth,” thereby placing herself beyond reproach. But in her job of a self-declared messiah (it is never shown how much she earns) intending to bring together people with the supposed divine connection, she falls back on caste, class, complexion, height and sometimes breadth of smiles as plausible criteria for two people to give each other a shot at spending their lives together."
***********

And Livemint:

Opinion | What economic studies say about our marriage market
 29 Jul 2020,  by Anirudh Tagat
"A matchmaking show on Netflix seems to skim over the market deficiencies that scholars have studied in depth"
*************
Update:  and this, from the NY Times, Aug 5--

By Sanjena Sathian

"It’s easy to applaud stories about rejecting old customs in favor of modern ideals. It’s harder, yet worthwhile, to sit with the subtler tension between tradition and modernity. This is what the great marriage plots have always considered: a mannered society, and how to live within it."