Asia-Pacific Working Group
The Asia-Pacific group ran during the 2020-2021 academic year and covered a variety of topics, much like the very first MD4SG working group. The working group is no longer active, but you can connect with community members on the #asia-pacific channel on the MD4SG Slack. If you have not already joined the MD4SG Slack, please email membership@md4sg.com.
Working Group Organizer
Matthew Olckers | Postdoctoral Researcher | UNSW |
Working Group Members
Haris Aziz | Associate Professor | UNSW |
Zhengyang Bao | Graduate Student | Monash University |
Allan Hernandez Chanto | Lecturer | University of Queensland |
Shivam Gupta | Student | Indian Institute of Technology Kanpur |
Nam Ho-Nguyen | Lecturer | University of Sydney |
Memunat Ibrahim | Graduate Student | ANU |
Nathan Ives | Country Manager, Cambodia | Causal Design |
Alexandros Karakostas | Postdoctoral Researcher | University of Queensland |
Ashley Khor | Graduate Student | University of Pittsburgh |
Alexander Lam | Graduate Student | UNSW |
Barton Lee | Graduate Student | UNSW |
Jaeho Lee | Postdoctoral Researcher | KAIST |
Xiaoxia Lei | Graduate Student | Shanghai Jiao Tong University |
Irene Lo | Assistant Professor | Stanford University |
Liv Nemes-Nemeth | Honours Student | University of Sydney |
Alexandru Nichifor | Senior Lecturer | University of Melbourne |
Siqi Pan | Lecturer | University of Melbourne |
Fahimeh Ramezani | Postdoctoral Researcher | UNSW |
Christine Rizkallah | Lecturer | UNSW |
Sanket Shah | Graduate Student | Harvard University |
Nicholas Teh | Student | National University of Singapore |
Emil Temnyalov | Lecturer | University of Technology Sydney |
Kentaro Tomoeda | Senior Lecturer | University of Technology Sydney |
Vivek Trivedi | Graduate Student | University of Queensland |
Gabriel Tseng | Machine Learning Engineer | Okra Solar |
Allen Vong | Research Fellow | University of Macau |
Toby Walsh | Professor | UNSW |
Summary of the 2021 sessions
Facility Location Problems
Led by Alexander Lam.
Reading: "Approximate Mechanism Design without Money" by Ariel Procaccia and Moshe Tennenholtz.
Highlights:
- "Approximation can be used to obtain strategyproofness without resorting to payments."
- From an optimization perspective, how bad can it be if only a small percentage of the agents can lie?
Explaining Machine Learning
Led by Liv Nemes-Nemeth.
Reading: "On The Reasons Behind Decisions" by Adnan Darwiche and Auguste Hirth.
Highlights:
- An approach to explain how a machine learning algorithm reached a decision on a particular instance. Uses theory on prime implicants.
- This approach doesn't explicitly take into account correlations between features.
Liquid Democracy
Led by Nicholas Teh.
Reading: "Liquid Democracy: An Algorithmic Perspective" by Anson Kahng, Simon Mackenzie and Ariel Procaccia.
Highlights:
- Shows a counterintuitive result that delegating voting responsibility to better informed voters does not necessarily lead to a more accurate decision.
Healthcare Rationing
Led by Haris Aziz.
Reading: "Fair Allocation of Vaccines, Ventilators and Antiviral Treatments: Leaving No Ethical Value Behind in Health Care Rationing" by Parag A. Pathak, Tayfun Sönmez, M. Utku Ünver and M. Bumin Yenmez.
Highlights:
- Priority systems can create ethical problems for healthcare rationing. A system of reserves is more flexible and can overcome many of these problems.
- The paper provides a general theory of reserve systems.
Targeting using Community Information
Led by Zhengyang Bao.
Reading: "Targeting High Ability Entrepreneurs using Community Information: Mechanism Design in the Field" by Reshmaan Hussam, Natalia Rigol and Benjamin Roth.
Highlights:
- Entrepreneurs hold remarkably good information about their peers.
- The peer information can be used to determine which entrepreneurs have high marginal returns to capital.
- When the information is used for targeting loans, peer prediction mechanisms can deter favoritism towards friends and relatives.
Social Drivers of Inequality
Led by Matthew Olckers.
Reading: "Inequality's Economic and Social Roots: The Role of Social Networks and Homophily" by Matthew O. Jackson.
Highlights:
- Divided networks can lead to inequality and immobility. For example, job referrals favor people whose friends have jobs, and it's easier to apply to university when you already know someone who has been through the process.
- Policies to combat inequality should take into account how benefits may spread through social networks.
The Economics of Privacy
Led by Allen Vong.
Reading: "The Economics of Privacy" by Alessandro Acquisti, Curtis Taylor and Liad Wagman.
Highlights:
- Privacy can be beneficial or costly depending on the context.
- The digital economy has limited consumer's ability to understand how their data will be used and what the consequences will be.
Summary of the 2020 sessions
Market Design in Education
Led by Kentaro Tomoeda.
Reading: What Really Matters in Designing School Choice Mechanisms by Parag Pathak.
Highlights:
- In designing school choice mechanisms, what's important in theory may not be important in practice... getting involved with implementing a mechanism highlights these cases.
- Strategyproofness (best response is to report the truth) does not guarantee truthful reporting in practice.
Diversity in Education Markets
Led by Emil Temnyalov.
Reading: "Explicit vs. Statistical Targeting in Affirmative Action: Theory and Evidence from Chicago's Exam Schools" (2020) by by Umut Dur, Parag Pathak and Tayfun Sönmez.
Highlights:
- Subtle details of mechanisms can have big impacts on outcomes. These details can be used as a lever for policymakers.
Fairness and Predictive Policing
Led by Ashley Khor.
Readings: A Snapshot of the Frontiers of Fairness in Machine Learning by Alexandra Chouldechova and Aaron Roth and page 131-141 of Predictive Policing: The Argument for Public Transparency by Erik Bakke.
Highlights:
- Concern that profit driven organisations generate most of the predictive solutions. Should there be more citizen involvement and oversight? The article by Erik Bakke suggests that transparency should be a mandatory feature of predictive policing.
- Open question on what the normative properties a predictive policing algorithm should satisfy.
Defining Fairness
Led by Jaeho Lee.
Reading: Prediction-Based Decisions and Fairness: A Catalogue of Choices, Assumptions, and Definitions by Shira Mitchell, Eric Potash, Solon Barocas, Alexander D'Amour, and Kristian Lum.
Highlights:
- The current ML pipeline imposes unintended fairness threats that are often hard to detect or quantify. Even the clearest cases have subtle issues.
- Impossibility results show that some definitions of fairness cannot be satisfied simultaneously.
Applications of Remote Sensing
Led by Gabriel Tseng.
Reading: Machine learning can help get COVID-19 aid to those who need it most by Joshua Blumenstock.
Highlights:
- Many regions, especially in Africa, have a lack of training data for remote sensing, a challenge which Gabriel has faced in his own work. Gabriel described how he tackled this challenge when his team was asked to deliver maps of farms for the Togolese government in rapid time.
- Can remote sensing allow us to sidestep the information assymetry at the heart of many mechanism design problems?
Foster Care
Led by Allan Hernández-Chanto.
Reading: Review of the Foster Care System by the Queensland Family & Child Commission
Highlights:
- The foster care system has unique characteristics that differ from popular applications of market design, such as school choice. The system is complex, has many stakeholders, involves a dynamic matching problem, and has multiple avenues to exit the system.
- Allan described the foster care system in Queensland. He noted that it has become increasingly difficult for the state to recruit foster parents.