Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF

Web and Internet Economics 16th
International Conference WINE 2020

Beijing China December 7 11 2020
Proceedings Xujin Chen
Visit to download the full and correct content document:
https://textbookfull.com/product/web-and-internet-economics-16th-international-confer
ence-wine-2020-beijing-china-december-7-11-2020-proceedings-xujin-chen/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Web and Internet Economics 10th International

Conference WINE 2014 Beijing China December 14 17 2014
Proceedings 1st Edition Tie-Yan Liu
https://textbookfull.com/product/web-and-internet-economics-10th-
international-conference-wine-2014-beijing-china-
december-14-17-2014-proceedings-1st-edition-tie-yan-liu/

Conference WINE 2016 Montreal Canada December 11 14
2016 Proceedings 1st Edition Yang Cai
international-conference-wine-2016-montreal-canada-
december-11-14-2016-proceedings-1st-edition-yang-cai/

Conference WINE 2018 Oxford UK December 15 17 2018
Proceedings George Christodoulou
international-conference-wine-2018-oxford-uk-
december-15-17-2018-proceedings-george-christodoulou/

Conference WINE 2019 New York NY USA December 10 12
2019 Proceedings Ioannis Caragiannis
international-conference-wine-2019-new-york-ny-usa-
december-10-12-2019-proceedings-ioannis-caragiannis/
Theory and Applications of Models of Computation 16th
International Conference TAMC 2020 Changsha China
October 18 20 2020 Proceedings Jianer Chen
https://textbookfull.com/product/theory-and-applications-of-
models-of-computation-16th-international-conference-
tamc-2020-changsha-china-october-18-20-2020-proceedings-jianer-
chen/
Bio inspired Information and Communication Technologies

12th EAI International Conference BICT 2020 Shanghai
China July 7 8 2020 Proceedings Yifan Chen
https://textbookfull.com/product/bio-inspired-information-and-
communication-technologies-12th-eai-international-conference-
bict-2020-shanghai-china-july-7-8-2020-proceedings-yifan-chen/
Information Systems Security 16th International

Conference ICISS 2020 Jammu India December 16 20 2020
Proceedings Salil Kanhere
https://textbookfull.com/product/information-systems-
security-16th-international-conference-iciss-2020-jammu-india-
december-16-20-2020-proceedings-salil-kanhere/
Combinatorial Optimization and Applications 14th

International Conference COCOA 2020 Dallas TX USA
December 11 13 2020 Proceedings Weili Wu
https://textbookfull.com/product/combinatorial-optimization-and-
applications-14th-international-conference-cocoa-2020-dallas-tx-
usa-december-11-13-2020-proceedings-weili-wu/
Distributed Computing and Internet Technology 16th

International Conference ICDCIT 2020 Bhubaneswar India
January 9 12 2020 Proceedings Dang Van Hung
https://textbookfull.com/product/distributed-computing-and-
internet-technology-16th-international-conference-
icdcit-2020-bhubaneswar-india-january-9-12-2020-proceedings-dang-
Xujin Chen
Nikolai Gravin
Martin Hoefer
Ruta Mehta (Eds.)
ARCoSS
LNCS 12495
Web and
Internet Economics
16th International Conference, WINE 2020
Beijing, China, December 7–11, 2020
Proceedings
Lecture Notes in Computer Science 12495
Founding Editors
Gerhard Goos, Germany
Juris Hartmanis, USA
Editorial Board Members

Elisa Bertino, USA Gerhard Woeginger , Germany
Wen Gao, China Moti Yung, USA
Bernhard Steffen , Germany
Advanced Research in Computing and Software Science

Subline of Lecture Notes in Computer Science
Subline Series Editors

Giorgio Ausiello, University of Rome ‘La Sapienza’, Italy
Vladimiro Sassone, University of Southampton, UK
Subline Advisory Board

Susanne Albers, TU Munich, Germany
Benjamin C. Pierce, University of Pennsylvania, USA
Bernhard Steffen , University of Dortmund, Germany
Deng Xiaotie, Peking University, Beijing, China
Jeannette M. Wing, Microsoft Research, Redmond, WA, USA
More information about this subseries at http://www.springer.com/series/7409
Xujin Chen Nikolai Gravin
• •
Martin Hoefer Ruta Mehta (Eds.)

•
Web and
Internet Economics
16th International Conference, WINE 2020
Beijing, China, December 7–11, 2020
Proceedings
123
Editors
Xujin Chen Nikolai Gravin
Academy of Mathematics Shanghai University of Finance
and Systems Science and Economics
Chinese Academy of Sciences Shanghai, China
Beijing, China
Ruta Mehta
Martin Hoefer University of Illinois at Urbana-Champaign
Goethe University Frankfurt Urbana, IL, USA
Frankfurt am Main, Germany
ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-030-64945-6 ISBN 978-3-030-64946-3 (eBook)
https://doi.org/10.1007/978-3-030-64946-3
LNCS Sublibrary: SL3 – Information Systems and Applications, incl. Internet/Web, and HCI
© Springer Nature Switzerland AG 2020

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, expressed or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This volume contains all regular papers and abstracts presented at the 16th Conference
on Web and Internet Economics (WINE 2020). WINE 2020 was held as an Internet
event during December 7–11, 2020, organized at Peking University, Beijing, China.
Over the last 16 years, the WINE conference series has become a leading inter-
disciplinary forum for the exchange of ideas and scientific progress across continents
on incentives and computation arising in diverse areas, such as theoretical computer
science, artificial intelligence, economics, and applied mathematics. WINE 2020 built
on the success of the previous editions of WINE (named Workshop on Internet and
Network Economics until 2013) which were held annually from 2005 to 2019.
The Program Committee was composed of 45 active researchers from the field.
They reviewed 136 submissions and decided to accept 42 papers. Each paper had at
least three reviews, with additional reviews solicited as needed. We are very grateful to
all members of the Program Committee for their insightful reviews and discussions. We
thank EasyChair for providing a virtual platform to organize the review process. We
also thank Springer for providing the proceedings and offering support for Best Paper
and Best Student Paper Awards.
In addition to the contributed talks, the program included four invited talks by
leading researchers in the field: Eric Budish (University of Chicago, USA), Jose Correa
(Universidad de Chile, Chile), Yiling Chen (Harvard University, USA), and Con-
stantinos Daskalakis (MIT, USA).
Our special thanks go to the general chair Xiaotie Deng and the local organization
team, as well as the poster chair Umang Bhaskar.
October 2020 Xujin Chen

Nikolai Gravin
Martin Hoefer
Ruta Mehta
Organization
General Chair
Xiaotie Deng Peking University, China
Program Committee Chairs

Xujin Chen Chinese Academy of Sciences, China
Nikolai Gravin Shanghai University of Finance and Economics, China
Martin Hoefer Goethe University Frankfurt, Germany
Ruta Mehta University of Illinois at Urbana-Champaign, USA
Steering Committee
Xiaotie Deng Peking University, China
Paul Goldberg Oxford University, UK
Christos Papadimitriou Columbia University, USA
Paul Spirakis The University of Liverpool, UK
Rakesh Vohra University of Pennsylvania, USA
Andrew Yao Tsinghua University, China
Yinyu Ye Stanford University, USA
Program Committee
Bo An Nanyang Technological University, Singapore
Siddharth Barman Indian Institute of Science, India
Xiaohui Bei Nanyang Technological University, Singapore
Simina Branzei Purdue University, USA
Yang Cai Yale University, USA
Shuchi Chawla University of Wisconsin-Madison, USA
Jing Chen Stony Brook University, USA
Xi Chen Columbia University, USA
Yukun Cheng Suzhou University of Science and Technology, China
Jose Correa Universidad de Chile, Chile
Edith Elkind University of Oxford, UK
Dimitris Fotakis National Technical University of Athens, Greece
Vasilis Gkatzelis Drexel University, USA
Kira Goldner Columbia University, USA
Tobias Harks University of Augsburg, Germany
Zhiyi Huang The University of Hong Kong, China
Thomas Kesselheim University of Bonn, Germany
Max Klimm Humboldt University Berlin, Germany
viii Organization
Yuqing Kong Peking University, China

Tie-Yan Liu Microsoft Research Asia, China
Irene Lo Stanford University, USA
Brendan Lucier Microsoft Research New England, USA
Luca Moscardelli University of Chieti-Pescara, Italy
Sigal Oren Ben-Gurion University, Israel
Georgios Piliouras Singapore University of Technology and Design,
Singapore
Qi Qi Hong Kong University of Science and Technology,
China
Daniela Saban Stanford University, USA
Rahul Savani The University of Liverpool, UK
Marco Scarsini LUISS Rome, Italy
Alkmini Sgouritsa The University of Liverpool, UK
Xiaoming Sun Chinese Academy of Sciences, China
Vasilis Syrgkanis Microsoft Research New England, USA
Sam Taggart Oberlin College, USA
Christos Tzamos University of Wisconsin-Madison, USA
Adrian Vetta McGill University, Canada
Carmine Ventre King’s College London, UK
Zihe Wang Shanghai University of Finance and Economics, China
Matt Weinberg Princeton University, USA
Lirong Xia Rensselaer Polytechnic Institute, USA
Jialin Zhang Chinese Academy of Sciences, China
Song Zuo Google Research, China
Additional Reviewers
Andreas Abels Yurong Chen

Mete Şeref Ahunbay Zhihuai Chen
Michele Aleandri Yun Kuen Cheung
Maxwell Allman Je-Ok Choi
Siddhartha Banerjee Avi Cohen
Umang Bhaskar Vincent Conitzer
Davide Bilò Andrés Cristi
Vittorio Bilò Bart de Keijzer
Georgios Birmpas Argyrios Deligkas
Peter Biro Yuan Deng
Shant Boodaghians Jack Dippel
Alexander Braun Shahar Dobzinski
William Brown Valerio Dose
Johannes Brustle Alon Eden
Linda Cai Guillaume Escamocher
Rahul Chandan Meryem Essaidi
Wei Chen Angelo Fanelli
Organization ix
Yiding Feng Daniel Moroz

Diodato Ferraioli Rolf H. Möhring
Matheus V. X. Ferreira Vishnu Narayan
Aris Filos-Ratsikas Ndiamé Ndiaye
Rupert Freeman Eric Neyman
Rafael Frongillo Kim Thang Nguyen
Sai Ganesh Nagarajan Rad Niazadeh
Evangelia Gergatsouli Argyris Oikonomou
Daniele Giachini Dario Paccagnan
Yiannis Giannakopoulos Panagiotis Patsilinakos
Sreedurga Gogulapati Dominik Peters
Lukas Graf Chara Podimata
Hadi Hosseini Luciano Pomatto
Neng Huang Emmanouil Pountourakis
Jan Hazła Bary Pradelski
Zhihao Jiang Li Qian
Alkis Kalavasis Goran Radanovic
Anthimos-Vardis Kandiros Nidhi Rathi
Jamie Kang Bhaskar Ray Chaudhury
Artem Kaznatcheev Rebecca Reiffenhäuser
Suleyman Kerimov Rojin Rezvan
Pieter Kleer Fedor Sandomirskiy
Vasilis Kontonis Alvaro Sandroni
Anand Krishna Daniel Schmand
Stefanos Leonardos Marc Schroder
Bo Li Steffen Schuldenzucker
Minming Li Ariel Schvartzman
Weian Li Erel Segal-Halevi
Thanasis Lianeas Ella Segev
Shengxin Liu Xiaohan Shan
Tian Liu Xinkai Shu
Tracy Liu Sujoy Sikdar
Yang Liu Ryann Sim
Zhengyang Liu Sahil Singla
Xinhang Lu Stratis Skoulakis
Thodoris Lykouris Warut Suksompong
Weidong Ma Ranjani Sundaram
Yishay Mansour Mashbat Suzuki
Simon Mauras Zhihao Tang
Alejandro Melo Ponce Yifeng Teng
Kaleigh Mentzer Clayton Thomas
Divyarthi Mohan Artem Tsikiridis
Faidra Monachou Marc Uetz
Gianpiero Monaco Rohit Vaish
Barnabé Monnot Grigoris Velegkas
Ilan Morgenstern Xavier Venel
x Organization
José Verschae Xiang Yan

Cosimo Vinci Zhenzhen Yan
Alexandros Voudouris Wei Yu
Bo Waggoner Huishuai Zhang
Chenhao Wang Jia Zhang
Hao Wang Mengqian Zhang
Kangning Wang Peng Zhang
Xiangning Wang Qiankun Zhang
Yining Wang Yong Zhang
Xiaowei Wu Zhijie Zhang
Yifan Wu Mingfei Zhao
Mingyu Xiao Shuran Zheng
Mobin Yahyazadeh Yichi Zhou
Contents
Matching
Almost Envy-Free Repeated Matching in Two-Sided Markets. . . . . . . . . . . . 3

Sreenivas Gollapudi, Kostas Kollias, and Benjamin Plaut
Dynamic Weighted Matching with Heterogeneous Arrival

and Departure Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Natalie Collina, Nicole Immorlica, Kevin Leyton-Brown,
Brendan Lucier, and Neil Newman
A Fine-Grained View on Stable Many-To-One Matching Problems

with Lower and Upper Quotas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Niclas Boehmer and Klaus Heeger
The Ad Types Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Riccardo Colini-Baldeschi, Julián Mestre, Okke Schrijvers,
and Christopher A. Wilkens
Multidimensional Stable Roommates with Master List . . . . . . . . . . . . . . . . . 59

Robert Bredereck, Klaus Heeger, Dušan Knop, and Rolf Niedermeier
Markets
Optimal Nash Equilibria for Bandwidth Allocation . . . . . . . . . . . . . . . . . . . 77

Benjamin Plaut
Counteracting Inequality in Markets via Convex Pricing . . . . . . . . . . . . . . . 89

Ashish Goel and Benjamin Plaut
Markets for Efficient Public Good Allocation with Social Distancing. . . . . . . 102
Devansh Jalota, Marco Pavone, Qi Qi, and Yinyu Ye
Mechanism Design and Pricing
Two Strongly Truthful Mechanisms for Three Heterogeneous Agents

Answering One Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Grant Schoenebeck and Fang-Yi Yu
A Generic Truthful Mechanism for Combinatorial Auctions . . . . . . . . . . . . . 133

Hanrui Zhang
xii Contents
The Price of Anarchy of Two-Buyer Sequential Multiunit Auctions. . . . . . . . 147

Mete Şeref Ahunbay and Adrian Vetta
Revenue-Optimal Deterministic Auctions for Multiple Buyers with Ordinal

Preferences over Fixed-Price Items . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Will Ma
Robust Revenue Maximization Under Minimal Statistical Information . . . . . . 177

Yiannis Giannakopoulos, Diogo Poças,
and Alexandros Tsigonias-Dimitriadis
Revenue Monotonicity Under Misspecified Bidders . . . . . . . . . . . . . . . . . . . 191

Makis Arsenis, Odysseas Drosis, and Robert Kleinberg
On the Power and Limits of Dynamic Pricing in Combinatorial Markets . . . . 206

Ben Berger, Alon Eden, and Michal Feldman
Competitively Pricing Parking in a Tree. . . . . . . . . . . . . . . . . . . . . . . . . . . 220

Max Bender, Jacob Gilbert, Aditya Krishnan, and Kirk Pruhs
Routing, Scheduling, Load Balancing
The Price of Anarchy for Instantaneous Dynamic Equilibria . . . . . . . . . . . . . 237

Lukas Graf and Tobias Harks
Data-Driven Models of Selfish Routing: Why Price of Anarchy Does

Depend on Network Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
Francisco Benita, Vittorio Bilò, Barnabé Monnot, Georgios Piliouras,
and Cosimo Vinci
Competition Alleviates Present Bias in Task Completion . . . . . . . . . . . . . . . 266

Aditya Saraf, Anna R. Karlin, and Jamie Morgenstern
Improving Approximate Pure Nash Equilibria in Congestion Games . . . . . . . 280

Vipin Ravindran Vijayalakshmi and Alexander Skopalik
The Curse of Rationality in Sequential Scheduling Games . . . . . . . . . . . . . . 295

Cong Chen and Yinfeng Xu
Sequential Solutions in Machine Scheduling Games . . . . . . . . . . . . . . . . . . 309

Cong Chen, Paul Giessler, Akaki Mamageishvili, Matúš Mihalák,
and Paolo Penna
Nash Social Welfare in Selfish and Online Load Balancing . . . . . . . . . . . . . 323

Vittorio Bilò, Gianpiero Monaco, Luca Moscardelli, and Cosimo Vinci
Contents xiii
Fairness
Simultaneously Achieving Ex-ante and Ex-post Fairness . . . . . . . . . . . . . . . 341

Haris Aziz
Optimal Bounds on the Price of Fairness for Indivisible Goods. . . . . . . . . . . 356

Siddharth Barman, Umang Bhaskar, and Nisarg Shah
Fair Division with Binary Valuations: One Rule to Rule Them All . . . . . . . . 370
Daniel Halpern, Ariel D. Procaccia, Alexandros Psomas,
and Nisarg Shah
Consensus Halving for Sets of Items . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384

Paul W. Goldberg, Alexandros Hollender, Ayumi Igarashi,
Pasin Manurangsi, and Warut Suksompong
Learning
Learning Strong Substitutes Demand via Queries . . . . . . . . . . . . . . . . . . . . 401

Paul W. Goldberg, Edwin Lock, and Francisco Marmolejo-Cossío
A Cardinal Comparison of Experts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416

Itay Kavaler and Rann Smorodinsky
Minimum-Regret Contracts for Principal-Expert Problems . . . . . . . . . . . . . . 430

Caspar Oesterheld and Vincent Conitzer
Bayesian Repeated Zero-Sum Games with Persistent State,

with Application to Security Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
Vincent Conitzer, Yuan Deng, and Shaddin Dughmi
Abstracts
Large Random Matching Markets with Localized Preference Structures Can

Exhibit Large Cores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
Ross Rheingans-Yoo
The Influence of One Strategic Agent on the Core of Stable Matchings . . . . . 463
Ron Kupfer
How Many Citizens Have Already Voted? The Effect of (Interim) Turnout
Rate Polls in Elections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
Akaki Mamageishvili and Oriol Tejada
Online Hypergraph Matching with Delays . . . . . . . . . . . . . . . . . . . . . . . . . 465

Marco Pavone, Amin Saberi, Maximilian Schiffer, and Matthew Tsao
xiv Contents
Market Equilibrium in Multi-tier Supply Chain Networks. . . . . . . . . . . . . . . 467

Tao Jiang, Young-San Lin, and Thành Nguyen
Decision Scoring Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468

Caspar Oesterheld and Vincent Conitzer
Bayesian Learning in Dynamic Nonatomic Routing Games . . . . . . . . . . . . . 469

Emilien Macault, Marco Scarsini, and Tristan Tomala
Privacy Rights and Data Security: GDPR and Personal

Data Driven Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 470
T. Tony Ke and K. Sudhir
Closing the Gap: Mitigating Bias in Online Résumé-Filtering . . . . . . . . . . . . 471

Jad Salem and Swati Gupta
Catastrophe by Design in Population Games: Destabilizing Wasteful

Locked-In Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
Stefanos Leonardos, Iosif Sakos, Costas Courcoubetis,
and Georgios Piliouras
Assortment Planning for Two-Sided Sequential Matching Markets . . . . . . . . 475

Itai Ashlagi, Anilesh K. Krishnaswamy, Rahul Makhijani,
Daniela Saban, and Kirankumar Shiragur
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477

Matching
Almost Envy-Free Repeated Matching
in Two-Sided Markets
Sreenivas Gollapudi1 , Kostas Kollias1 , and Benjamin Plaut2(B)

1
Google Research, Mountain View, USA
sgollapu@google.com, kostaskollias@google.com
2
Stanford University, Stanford, USA
bplaut@stanford.edu
Abstract. A two-sided market consists of two sets of agents, each of

whom have preferences over the other (Airbnb, Upwork, Lyft, Uber, etc.).
We propose and analyze a repeated matching problem, where some set
of matches occur on each time step, and our goal is to ensure fairness
with respect to the cumulative allocations over an infinite time horizon.
Our main result is a polynomial-time algorithm for additive, symmetric
(vi (j) = vj (i)), and binary (vi (j) ∈ {a, 1}) valuations that both (1) guar-
antees envy-freeness up to a single match (EF1) and (2) selects a maximum
weight matching on each time step. Thus for this class of valuations, fair-
ness can be achieved without sacrificing economic efficiency. This result
holds even for dynamic valuations, i.e., valuations that change over time.
Although symmetry is a strong assumption, we show that this result can-
not be extended to asymmetric binary valuations: (1) and (2) together are
impossible even when valuations do not change over time, and for dynamic
valuations, even (1) alone is impossible. To our knowledge, this is the first
analysis of envy-freeness in a repeated matching setting.
1 Introduction
Recent years have seen a dramatic increase in electronic marketplaces, both
in quantity and scale. Many of these are two-sided markets, meaning that the
market makes matches between two sets of agents (homeowners and guests for
Airbnb, employers and workers for Upwork, drivers and riders for Lyft and Uber,
etc.), each of whom has preferences over the other. This is in contrast to tra-
ditional resource allocation (cake cutting, Fisher markets, auctions, etc.) where
only one side of the market has preferences. Although envy-freeness and relax-
ations thereof have been studied extensively in one-sided resource allocation (this
research area is typically referred to as “fair division”), we are aware of just one
paper considering envy-freeness for two-sided preferences [22].1
There are two primary motivations for our work. The first is to simply study
fair division for two-sided preferences. The second is that in some ways, two-
sided electronic marketplaces like Airbnb, Upwork, Lyft, and Uber are actually
1
Although [22] is conceptually similar, it is in different setting of recommendation
algorithms, and so it is technically quite different.
c Springer Nature Switzerland AG 2020
X. Chen et al. (Eds.): WINE 2020, LNCS 12495, pp. 3–16, 2020.
https://doi.org/10.1007/978-3-030-64946-3_1
4 S. Gollapudi et al.
in a better position to impose fairness than one-sided marketplaces. The reason

is that most one-sided markets are decentralized, in the sense that a seller offers
different goods at different prices, buyers peruse the wares at their leisure, and
make individual decisions about what they wish to purchase. On the contrary,
most two-sided markets operate by way of matches mediated by a centralized
platform, giving the platform the ability to affect the outcomes of the system.
Indeed, on Lyft and Uber, an automated central authority has almost com-
plete control over the matches, giving the algorithm tremendous power over the
outcomes for each individual agent. The power dynamic between the platform
and the participating agents makes it even more important to ensure that the
matching algorithms are fair to each agent.
1.1 Repeated Matching

A crucial element of marketplaces is repeated matching. Agents do not receive
all of their matches at once; typically, an agent can only process a few matches
at a time (a driver can only fit so many riders in the car, a worker can only
handle so many contracts at once). Only after an agent completes some of her
current matches can she be given new matches. This motivates a model where on
each time step, an irrevocable matching decision must be made, and we expect
fairness with respect to the cumulative matching at each time step. We consider
an infinite time horizon and a finite set of agents, so we must allow the same
pair of agents to be matched multiple times; thus each agent’s cumulative set of
matches will be a multiset.
A vital aspect of any repeated setting is that preferences can change. In
some cases, preferences may change in direct response to the matches an agent
receives: if a driver is matched with a rider who wishes to go to location X, once
that ride is completed, the driver will prefer riders whose pickups are close to
X. In other cases, agents may desire variety among matches: Airbnb guests may
not wish to vacation in the same area every time. Additionally, preferences may
simply drift over time. We refer to valuations that change over time as dynamic
valuations.
1.2 Fairness Notions

For one-sided fair division with indivisible items, full envy-freeness is impossible:
for two agents and a single item, one must receive the item and the other agent
will be envious. The same issue applies in our setting: if every agent on one side
of the market is interested in the same agent on the other side, no algorithm can
guarantee envy-freeness.
One solution is to consider relaxations of envy-freeness, such as envy-freeness
up to one good (EF1). An outcome is EF1 if whenever agent i envies agent j,
there exists a good in j’s bundle such that i would not envy j after removing
that good.2 This property has been studied widely for one-sided preferences, but
2
Note that this good is not actually removed: this is simply a thought experiment
used in the definition of EF1.
Almost Envy-Free Repeated Matching in Two-Sided Markets 5
to our knowledge has not been considered for two-sided markets. We can define
EF1 equivalently for two-sided preferences: simply replace “there exists a good”
with “there exists a match”, etc.
The less obvious question is how to adapt EF1 to the repeated setting. In this
paper, we assume time is divided into discrete steps, where on each step, some
set of matches occur. Each match consists of two agents, one from each side of
the market. We would like the cumulative matching after each time step to be
EF1. We will also assume that each side of the market has the same number of
agents (if not, add “dummy” agents to the smaller side).
We consider two different versions of this model. In the first version, each time
step consists of just a single match, so we are effectively requiring the cumulative
matching to be EF1 at every point in time. We call this EF1-over-time.
However, asking for fairness at every point in time may be too strong.
Furthermore, in most real-world applications, matches would be happening in
parallel anyway. Conversely, EF1-over-time poses no restrictions on how many
matches agents receive. In real life, agents often have similar “capacities” (e.g.,
most cars have a similar number of seats) and thus should arguably receive
matches at similar rates. These concerns motivate a second version of the model,
where on each time step, each agent is matched exactly once (i.e., we select a
perfect matching). Thus each time step represents a “round” of matches (which
may happen in parallel). We still require that the cumulative matching is EF1
after time step, and we call this EF1-over-rounds.
1.3 Our Results
We use vi (j) to denote agent i’s value for agent j, and assume valuations are
additive. We say that valuations are symmetric if vi (j) = vj (i) for all agents
i, j, and binary if there exists a ∈ [0, 1) such that vi (j) ∈ {a, 1} for all i, j.3
It is worth noting that for symmetric valuations, negative values for vi (j) are
subsumed in the following sense: if the algorithm ever says to match agents i
and j where vi (j) < 0, we simply ignore this and never make the match, which
gives both agents value 0 for the “match”.4
We now describe our results. Due to space constraints, all proofs are deferred
to the full version of the paper [16].
EF1-Over-Rounds for Dynamic Symmetric, and Binary Valuations. Our main

result is that for dynamic, symmetric, and binary valuations, we give an algorithm
which both satisfies EF1-over-rounds, and selects a maximum weight matching on
each time step (Theorem 3.1) This holds even when valuations are dynamic.5 This
3
If we wish to add dummy agents to one side of the market, the most natural case
would be a = 0, in order to express that dummy agents have no value.
4
In order for this argument to be complete when considering EF1-over-rounds, this
“match” must still count as part of the perfect matching for that time step.
5
We allow valuations to change arbitrarily between time steps. Furthermore, our algo-
rithm does not need to know how valuations will change in response to a given match.
shows that for this class of valuations, fairness can be achieved without sacrificing
economic efficiency. Our algorithm runs in time O(n2.5 ) per time step.
The class of symmetric and binary valuations is somewhat restricted, but is
important to keep several things in mind. First, it is often hard to elicit more
complex valuations. Agents can easily answer binary questions such as “Would
you be happy with this match?”, but may not be able to provide a real number
value for potential matches. Second, the best interpretation of our result (in our
opinion) is that agents’ preferences are not truly binary, but that our algorithm
is guaranteeing EF1 with respect to a binary projection of the preferences. That
is, ask each agent to label each possible match as “good” or “bad”, and guarantee
EF1 with respect to those preferences. This interpretation is reinforced by the
fact that the cumulative matchings computed by our algorithm will be EF1
uniformly across all possible values a, and agents can even have different values
of a. See Sect. 3.1 for a discussion of this.
Symmetry, however, is a significant assumption on the agents’ preferences.
There are reasons to believe real-world preferences are largely symmetric: a rider
is likely to prefer a driver closer to her, and vice versa. However, a natural
question is whether this assumption is necessary.
Counterexamples. Our next set of results shows that the symmetry assumption
is in fact necessary. First, we show that for dynamic binary valuations, EF1-
over-rounds alone is impossible (Theorem 4.1). Second, for non-dynamic (i.e.,
valuations do not change over time) binary valuations, it is impossible to satisfy
EF1-over-rounds while guaranteeing a maximum weight matching for each time
step (Theorem 4.2). These impossibility results suggest that EF1-over-rounds
may be too much to ask for in the setting of two-sided repeated matching.
However, our counterexamples do not rule out the possibility of EF1-over-time,
even for general additive valuations. We leave this as our primary open question.
Beyond Symmetric Valuations. Despite this negative result, we show that it is

possible to relax the symmetry assumption, at least in the context of EF1-over-
time. We show that for {0, 1} binary valuations6 with an assumption that we call
“only symmetric cycles”, EF1-over-time can be guaranteed. We formally define
“only symmetric cycles” in the full version of the paper [16], but this assumption
is strictly weaker than full symmetry of valuations.
Beyond Binary Valuations. In a similar vein, we show that when one side of the
market has two agents, the binary assumption can also be relaxed. Specifically,
we give an algorithm which is EF1-over-time for any additive valuations.
1.4 Related Work

There are two primary bodies of related work: (one-sided) fair division, and
matching markets.
6
For this result, we assume that vi (j) ∈ {0, 1} for each agent, as opposed vi (j) ∈ {a, 1}
for any a ∈ [0, 1).
Fair Division. Fair division has a long history. In fact, the Bible documents
Abraham and Lot’s use of the cut-and-choose protocol to fairly divide land.
The formal study of fair division was started by [27] in 1948, and envy-freeness
was proposed in 1958 [15] and further developed by [13]. A full overview of the
fair division literature is outside the scope of this paper (we refer the interested
reader to [5,21]), and we discuss only the work most relevant to our own.
There are two main differences between our work and that of traditional fair
division. First, we study two-sided preferences instead of one-sided preferences.
Second, we study a repeated setting, where we must make an irrevocable decision
on each time step; most fair division research considers a “one-shot” model where
all the goods are allocated at once.
We briefly overview some key results in the one-sided one-shot model of fair
division for indivisible items7 . The EF1 property was proposed for this model
by [6]. EF1 allocations always exist, and can be computed in polynomial time,
even for general combinatorial valuations [20]8 . This sweeping positive result for
the one-sided one-shot model lies in stark contrast to our negative results for
the two-sided repeated model. It was later shown that for additive valuations,
maximizing the product of valuations yields an allocation that is both EF1 and
Pareto optimal [8].
Envy-Freeness up to Any Good (EFX). The same paper suggested a new fairness
notion that is strictly stronger than EF1, which they called EFX9 . The first for-
mal results regarding EFX allocations were given by [23]. A major breakthrough
recently proved the existence of EFX allocations for additive valuations and three
agents [9], but despite ongoing effort, the question of existence remains unsolved
for more than three agents (or more complex valuations). This is perhaps the
most significant open problem in the fair division of indivisible items.
In many contexts (especially for additive valuations), it is common to modify
the requirement to be that whenever i envies j, removing any good which i
values positively from j’s bundle is sufficient to eliminate the envy [8].10 Under
the latter definition, for {0, 1} binary valuations, EF1 and EFX coincide. This
is because the only positively valued goods are the maximum value goods. In
this sense, our positive results for {0, 1} binary valuations immediately extend
to EFX as well.
Repeated Fair Division. There are a smattering of recent works studying envy-
freeness for one-sided preferences in a repeated (i.e., not one-shot) setting; see [1]
7
Indivisible items, such as cars, must each go entirely to a single agent. In contrast,
divisible items, such as a cakes, can be split between multiple players. Fair division
studies both of these settings, but the indivisible case is more relevant to our work.
8
The algorithm of [20] was originally developed with a different property in mind.
9
An allocation is envy-free up to any good (EFX) if whenever i envies j, removing
any good from j’s bundle eliminates the envy.
10
The reason this is less common when considering non-additive valuations is that for
general combinatorial valuations, it is less clear what “values positively” means.
for a short survey. One example is [2], which focuses on minimizing the maximum
envy (i.e., the maximum difference between an agent’s value for her own bundle
and her value for another agent’s bundle) at each time step. Despite the growing
interest in repeated one-sided fair allocation, the literature on the analogous
two-sided problem remains sparse.
Matching Markets. The other relevant field is (bipartite) matching markets11 .

In a one-to-one matching market, each agent receives exactly one match. Perhaps
the most famous result for one-to-one matching is that of Gale and Shapley,
whose algorithm finds a stable matching [14]. More relevant to us is the model of
many-to-many matching markets, where each agent can receive multiple matches;
stability has often been the primary criterion in this model as well [11,19,25,26].
There is also some work on stability in dynamic matching markets [10,18].
In contrast, fairness in many-to-many matching has received considerably
less attention. In fact, Gale and Shapley’s algorithm for one-to-one matching is
known to compute the stable matching which is the worst possible for one side
of the market, and best possible for the other.
Fair Ride-Hailing. There is a growing body of work surrounding the ethics of

crowdsourced two-sided markets, especially ride-hailing (e.g., Lyft and Uber)
[3,4,7,12,17]. We are aware of just two works studying fairness for two-sided
markets from an algorithmic perspective: [29] and [28], both of which focus on
ride-hailing. The former paper considers ride-sharing, where multiple passengers
are matched with a single driver. The authors focus on fairness with respect
to the savings achieved by each passenger. This paper is primarily theoretical,
like ours, but is specific to ride-hailing, unlike ours. The latter paper studies a
fairness notion based on the idea that “spread over time, all drivers should receive
benefits proportional to the amount of time they are active in the platform”. The
model considered in this paper is more general than just ride-hailing, however
the paper is primarily experimental, and the experiments are in the ride-hailing
setting.
Consequently, we are not aware of any prior work studying algorithms with
provable fairness guarantees for repeated two-sided matching markets. In this
way, our work can be viewed as simultaneously building on the fair division
literature (by considering two-sided preferences) and building on the matching
market literature (by studying envy-freeness for repeated two-sided markets).
The paper proceeds as follows. Section 2 describes the formal model. Section 3
presents our main result: an algorithm for symmetric and binary valuations such
that (1) the sequence of cumulative matchings is EF1-over-rounds, (2) a max-
imum weight matching is chosen for each time step, and (3) this holds even
for dynamic valuations (Theorem 3.1). Section 4 presents our counterexamples:
Theorem 3.1 cannot be extended to non-symmetric binary valuations. Specifi-
cally, without symmetry, (1) and (3) together and (1) and (2) together are both
11
For a broad overview of this topic, see [24].
impossible (Theorems 4.1 and 4.2, respectively). The rest of the results (and all
of the proofs) can be found in the full version of the paper [16].
2 Model
Let N and M be two sets of agents. We assume that |N | = |M | = n; if this
is not the case, we can add “dummy” agents (i.e., agents i such that vi (j) =
vj (i) = 0 for all j) to the smaller side of the market until both sides have the
same number of agents. We will typically use odd numbers for the elements
of N and even numbers for the elements of M , i.e., N = {1, 3, . . . , 2n − 1} and
M = {2, 4, . . . , 2n}. A matching X assigns a multiset of agents in N to each agent
in M , and a multiset of agents in M to each agent in N . For each i ∈ N ∪ M , we
will use Xi to denote agent i’s bundle, i.e., the multiset of agents she is matched
to. Throughout the paper, we will use standard set notation for operations on
the multisets Xi . For example, Xi ∪ {j} increments the multiplicity of j in Xi ,
i.e., the number of times j occurs in Xi . In order for X to be a valid matching,
the multiplicity of j in Xi must be equal to the multiplicity of i in Xj for each
i ∈ N, j ∈ M.
Each agent i also has a valuation function vi , which assigns a real number
to each possible bundle she might receive. We will use v to denote the valuation
profile which assigns valuation vi to agent i. We say that vi is additive if for any
bundle Xi ,
vi (X) = vi ({j})
j∈X
Since X is a multiset, the sum over j ∈ X includes each j a number of times

equal to its multiplicity. For example, if X = {j, j} and vi ({j}) = 1, then
vi (X) = 2. With slight abuse of notation, we will write vi ({j}) = vi (j). We say
that a valuation vi is binary if there exists a ∈ [0, 1) such that vi (j) ∈ {a, 1}
for all i, j ∈ N or i, j ∈ M . We say that a valuation profile v is symmetric if
vi (j) = vj (i) for all i ∈ N, j ∈ M .
We say that i envies j under X if vi (Xi ) < vi (Xj ). We only consider envy
within the same side of the market: it is unclear what it would mean for some
i ∈ N to envy j ∈ M . We can express this by setting vi (j) = 0 for i, j ∈ N or
i, j ∈ M .
Definition 2.1. A matching X is envy-free up to one match (EF1) if whenever
i envies j, there exists k ∈ Xj such that vi (Xi ) ≥ vi (Xj \ {k}).
Note that the set subtraction Xj \ {k} decreases the multiplicity of k by 1;
it does not remove k altogether. Also, we will say that X is EF1 with respect to
N (resp., M ) if the above holds for every pair i, j ∈ N (resp., i, j ∈ M ).
One important tool we will use is the envy graph:
Definition 2.2. The envy graph of a matching X is a graph with a vertex for
each agent, and a directed edge from agent i to agent j if i envies j under X.
We will especially be interested in cycles in the envy graph, and will use the
terms “cycle in the envy graph” and “envy cycle” interchangeably.
2.1 Repeated Matching

We consider a repeated setting, where on each time step t, some set of matches
occur. Each “match” (alternatively, pairing) consists of one agent in N and one
agent in M .
Let xt denote the set of matches which occur at time t. Each agent will
receive at most one match per time step. If agent i is matched to an agent j at
time t, let xti = {j}; otherwise, let xti = ∅. For an infinite sequence x1 , x2 , x3 . . . ,
let X t denote the cumulative matching up to and including time t. Formally, for
each i ∈ N ∪ M ,
t ∅ if t = 0
Xi =
Xit−1 ∪ xti if t > 0
In words, Xit is the set of matches i has received up to and including time t.
Our main result holds even when valuations are allowed to vary over time.
Specifically, a dynamic valuation vi will have a value vit (j) for each agent j on
each time step t (as before, we write vit (j) = 0 for i, j ∈ N or i, j ∈ M ). A profile
of dynamic valuations is symmetric if vit (j) = vjt (i) for all i, j, t. For a pair of
agents i, j (with i = j allowed), vi (Xjt ) is given by
t

vi (Xjt ) = vit (xtj )
t =1

where vit (∅) = 0. That is, i’s value for a bundle Xj is as if i had received exactly
those matches at exactly those times. It is important for this definition to include
both i = j and i = j, so that we can evaluate envy between agents.
We make no assumptions on how valuations change between time steps: they
can even change adversarially, since our algorithm will not use any knowledge
about future valuations when making matching decisions.
We consider two definitions of EF1 in the repeated matching setting. In both
cases, we require the cumulative matching at the end of each time step to be
EF1. The difference is that for EF1-over-time, each time step consists a single
match, and for EF1-over-rounds, each time step consists of a “round” of matches
where all agents receive exactly one match (i.e., a perfect matching between N
and M ).
Definition 2.3. The sequence X = X 0 , X 1 , X 2 . . . is EF1-over-time if for all

t ≥ 0, each xt contains a single match, and X t is EF1.
Definition 2.4. The sequence X = X 0 , X 1 , X 2 . . . is EF1-over-rounds if for

all t ≥ 0, xt is a perfect matching, and X t is EF1.
Formally, these notions are incomparable: EF1-over-time has a stronger fair-

ness requirement (the cumulative matching should be EF1 after every match,
not just after every round of matches), but does not require agents to receive the
same number of matches. However, EF1-over-rounds does imply EF2-over-time
(where we may remove two matches in order to eliminate the envy): expand each
“round” into n time steps, each containing one match, in an arbitrary order. We
know that at the end of each round of n time steps, the cumulative matching
is EF1. Within each round, each agent only gains one additional match, and we
can always remove that match to return to an EF1 state.
Our goal will be to show the existence of (and efficiently compute) a sequence
x1 , x2 , x3 . . . such that the induced sequence X is EF1-over-time and/or EF1-
over-rounds. For brevity, if an algorithm is guaranteed to produce a sequence X
that is EF1-over-time (resp., EF1-over-rounds), we simply say that the algorithm
is EF1-over-time (resp., EF1-over-rounds).
3 EF1 for Dynamic, Symmetric, and Binary Valuations

In this section, we consider binary and symmetric valuations that may change
over time. For this class of valuations, we give a polynomial-time algorithm that
produces a sequence which is EF1-over-rounds, and chooses a maximum weight
matching for each time step. This leads to the following theorem:
Theorem 3.1. For dynamic, binary, and symmetric valuations, Algorithm 1 is

EF1-over-rounds, and the matching xt for each time step t is a maximum weight
matching (with respect to the valuations on that time step). Furthermore, the
algorithm runs in time O(n2.5 ) per time step.
3.1 Algorithm Setup

Before we discuss the algorithm, we need the following definition, which will
imply EF1 (Lemma 3.2):
Definition 3.1. We say that a pair of agents (i, j) is c-envy-bounded if

vi (Xj ) − vi (Xi ) ≤ c, and we say a matching X is c-envy-bounded if every pair
(i, j) is c-envy-bounded.
A quick note: recall that our goal is to choose a sequence of pairings x1 , x2 . . . ,

and that these pairings fully specify the sequence of cumulative matchings
X . Consequently, when giving pseudocode for our algorithms (throughout the
paper), we do not explicitly update X : we assume that whenever some xt is

changed, every X t for t ≥ t is automatically updated. We feel that this leads
to more concise and intuitive pseudocode.
Algorithm 1 is very simple. For each time step t, we initialize xt to be an arbi-
trary maximum weight matching for the current valuations, and make changes to
this matching until we are satisfied. Specifically, while there exist agents i, j such
that (i, j) is not (1−a)-envy-bounded in the cumulative matching, we swap their
matches in xt . When no such pair of agents exists, we exit the while loop and
confirm the matches. Throughout all of our algorithms, we will use the function
MakeMatch to indicate that we are confirming the matches in xt .
It is important to note that the algorithm is not going back in time and
changing pairings already made: once a pairing is confirmed with MakeMatch, it
Algorithm 1. An EF1-over-rounds algorithm for agents with dynamic, sym-

metric, and binary valuations.
1: function EF1Matching(N, M, v)
2: for each t ∈ N≥0 do
3: {xt } ← MaxWeightMatching(N, M, v)
4: while ∃ agents i, j s.t. vi (Xjt ) − vi (Xit ) > 1 − a do
5: (xti , xtj ) ← (xtj , xti )
6: MakeMatch(xt )
is never changed. The algorithm starts with a tentative matching, and changes
tentative matches until it is satisfied for the current time step (see Fig. 1), at
which point the matches are confirmed with MakeMatch. The algorithm then
proceeds to the next time step and never changes pairings from previous time
steps. Note also that the algorithm uses no information about valuations for
future time steps.
N: 1 3 1 3
M: 2 4 2 4
Fig. 1. A hypothetical swap performed by Algorithm 1. On the left we see a tentative

perfect matching: ({1, 2}, {3, 4}). The blue arrow indicates that if this matching were
to be chosen, the pair (3, 1) would not be (1 − a)-envy-bounded. Thus agents 1 and 3
swap their (tentative) matches, and the new tentative matching is ({1, 4}, {3, 2}). The
matching ({1, 2}, {3, 4}) is never confirmed by MakeMatch: it is merely a stepping stone
in the process of computing the eventual matches to be chosen for this time step. For
the case of more than four agents, this process could repeat (although not indefinitely;
see the full version of the paper for this proof [16]. (Color figure online)
Our central correctness lemma will be the following:
Lemma 3.1. Let t ≥ 1 be any time step, and suppose that X t−1 is (1 − a)-
envy-bounded and has no envy cycles. Then X t is (1 − a)-envy-bounded and
has no envy cycles. Furthermore, the chosen matching xt is a maximum weight
matching (with respect to the valuations on that time step).
Before diving into the proof of Lemma 3.1 (and the runtime analysis), we
briefly show that (1 − a)-envy-boundedness will actually give us the result we
want:
Lemma 3.2. Suppose valuations are binary, and suppose X is (1 − a)-envy-

bounded. Then X is EF1.
Proof. Suppose i envies j under X t . If vi (Xjt ) = a|Xjt |, then vi (Xit ) ≥ vi (Xjt ),

which contradicts i envying j. Thus vi (Xjt ) ≥ 1 + a(|Xjt | − 1). Thus there exists
k ∈ Xjt such that vi (Xjt \ {k}) = vi (Xjt ) − 1. Therefore vi (Xi ) − vi (Xjt \ {k}) ≥
1 + (a − 1) = a ≥ 0, which proves the claim.
The role of a. Before diving into the main proof, we briefly discuss the role of
a. For Theorem 3.1, we assume that there exists a ∈ [0, 1) such that vi (j) ∈ {a, 1}
for all i, j ∈ N or i, j ∈ M . For ease of notation, we assume that all agents have
the same value of a, but this is in fact not necessary. In fact, Algorithm 1 will
be EF1-over-rounds simultaneously for all values of a.
Lemma 3.3. Assume (i, j) is (1 − a)-envy-bounded and that |Xit | = |Xjt |. Let
a ∈ [0, 1), and define a new valuation vi such that vi (k) = a whenever vi (k) = a,
and vi (k) = 1 otherwise. Then (i, j) is (1 − a )-envy-bounded with respect to vi .
Note that the assumption of |Xit | = |Xjt | is always satisfied when working
with EF1-over-rounds, since we will match every agent once on each time step.
Therefore we can actually just choose an arbitrary value of a ∈ [0, 1) and run
Algorithm 1. Lemma 3.3 implies that the resulting sequence of matchings will be
EF1-over-rounds simultaneously for all values of a, even if different agents have
different values of a. That said, if we need to include dummy agents in order to
equalize the sizes of N and M , a = 0 probably makes the most sense.
4 Counterexamples
A natural question is whether Theorem 3.1 can be extended to all dynamic binary
valuations (i.e., not necessarily symmetric). The answer is unfortunately no,
which we show in two different ways. First, for dynamic binary valuations, EF1-
over-rounds alone is impossible (Theorem 4.1). Second, for non-dynamic binary
valuations, it is impossible to guarantee both EF1-over-rounds and maximum
weight matching for each time step (Theorem 4.2).
t=1 t=2
1 3 1 3
2 4 2 4
Fig. 2. An instance with dynamic and binary valuations where EF1-over-rounds is

impossible.
Theorem 4.1 uses the instance in Fig. 2. Essentially, after the first round,
either agents 1 and 3 form an envy cycle, or agents 2 and 4 form an envy cycle.
After the second round of matching, one of the agents in the envy cycle will
become even more envious, violating EF1.
Theorem 4.1. For dynamic and binary valuations, there is no algorithm which
is EF1-over-rounds.
1, 3 5
2 4 6
Fig. 3. An instance with binary valuations where guaranteeing both EF1-over-rounds

and maximum weight matching is impossible.
Theorem 4.2 uses the instance in Fig. 3. For some intuition, note that are two
cycles of desire: (1, 4, 5, 6) and (3, 4, 5, 6). Like in the previous counterexample,
these cycles will cause problems, but here we have the additional consideration
that agents 1 and 3 are competing for agent 4. We show that the frequency with
which agents 4 and 5 are matched is at least the frequency with which either
agent 1 or agent 3 is matched with agent 4. For example, if agents 1 and 3 have
each been matched to agent 4 twice, then agents 4 and 5 will have been matched
4 times. This leads to agents 1 and 3 increasingly envying agent 5, until EF1 is
violated.
The assumption of maximum weight is necessary only to prevent agents 2
and 5 from ever being matched: if agents 2 and 5 can be matched, the above
argument can be circumvented.
Theorem 4.2. For binary valuations, there is no algorithm which is EF1-over-

rounds and also chooses a maximum weight matching for each time step, even
for non-dynamic valuations.
5 Conclusion
In this paper, we proposed a model of envy-freeness for repeated two-sided

matching. For binary and symmetric valuations, we gave an algorithm that (1)
satisfies EF1-over-rounds, (2) chooses a maximum weight matching for each time
step, and (3) works even for dynamic valuations (Sect. 3). Furthermore, without
symmetry, (1) + (2) together and (1) + (3) together are each impossible. All
proofs can be found in the full version of the paper, along with several additional
results [16].
Our negative results for even binary valuations suggest that EF1-over-rounds
may be too much to ask for. However, our results do not rule out the possibility
of EF1-over-time, even for general additive valuations. More broadly, future work
could investigate other possible fairness notions for this setting.
Another possible future direction concerns more general study of two-sided
preferences. Envy-freeness is an example of a topic that has been widely studied
for one-sided resource allocation, but not for two-sided markets. We wonder if
there are other such topics that are worthy of study for two-sided preferences.
References
1. Aleksandrov, M., Walsh, T.: Online Fair Division: A Survey. arXiv preprint
arXiv:1911.09488 (2019)
2. Benade, G., Kazachkov, A.M., Procaccia, A.D., Psomas, C.A.: How to make envy
vanish over time. In: Proceedings of the 2018 ACM Conference on Economics and
Computation, EC 2018, pp. 593–610. ACM, New York (2018)
3. Bokányi, E., Hannák, A.: Evaluating Algorithm Fairness in an Agent-based Taxi
Simulation (2018, working paper)
4. Bokányi, E., Hannák, A.: Ride-share Matching Algorithms Generate Income
Inequality (2019, working paper)
5. Brandt, F., Conitzer, V., Endriss, U., Lang, J., Procaccia, A.D.: Handbook of
Computational Social Choice. Cambridge University Press, Cambridge (2016)
6. Budish, E.: The combinatorial assignment problem: approximate competitive equi-
librium from equal incomes. J. Polit. Econ. 119(6), 1061–1103 (2011)
7. Calo, R., Rosenblat, A.: The taking economy: uber, information, and power.
Columbia Law Rev. (2017)
8. Caragiannis, I., Kurokawa, D., Moulin, H., Procaccia, A.D., Shah, N., Wang, J.:
The unreasonable fairness of maximum Nash welfare. In: Proceedings of the 2016
ACM Conference on Economics and Computation, pp. 305–322. ACM (2016)
9. Chaudhury, B.R., Garg, J., Mehlhorn, K.: EFX exists for three agents. arXiv
preprint arXiv:2002.05119 (2020)
10. Damiano, E., Lam, R.: Stability in dynamic matching markets. Games Econ.
Behav. 52, 34–53 (2005)
11. Echenique, F., Oviedo, J.: A theory of stability in many-to-many matching markets.
Theor. Econ. 1(2), 233–273 (2006)
12. Fieseler, C., Bucher, E., Hoffmann, C.P.: Unfairness by design? The perceived
fairness of digital labor on crowdworking platforms. J. Bus. Ethics 156(4), 987–
1005 (2019)
13. Foley, D.K.: Resource allocation and the public sector. Yale Econ. Essays 7(1),
45–98 (1967)
14. Gale, D., Shapley, L.S.: College admissions and the stability of marriage. Am.
Math. Mon. 69(1), 9–15 (1962)
15. Gamow, G., Stern, M.: Puzzle-math. Viking Adult (1958). http://www.worldcat.
org/isbn/0670583359
16. Gollapudi, S., Kollias, K., Plaut, B.: Almost Envy-free Repeated Matching in Two-
sided Markets. arXiv preprint arXiv:2009.09336 (2020)
17. Hannák, A., Wagner, C., Garcia, D., Mislove, A., Strohmaier, M., Wilson, C.:
Bias in online freelance marketplaces: evidence from Taskrabbit and Fiverr. In:
Proceedings of the 2017 ACM Conference on Computer Supported Cooperative
Work and Social Computing, CSCW 2017, pp. 1914–1933. ACM, New York (2017)
18. Haruvy, E., Ünver, M.U.: Equilibrium selection and the role of information in
repeated matching markets. Econ. Lett. 94(2), 284–289 (2007)
19. Konishi, H., Ünver, M.U.: Credible group stability in many-to-many matching
problems. J. Econ. Theor. 129(1), 57–80 (2006)
20. Lipton, R.J., Markakis, E., Mossel, E., Saberi, A.: On approximately fair allocations
of indivisible goods. In: Proceedings of the 5th ACM conference on Electronic
commerce, pp. 125–131. ACM (2004)
21. Moulin, H.: Fair Division and Collective Welfare. MIT Press, Cambridge (2004)
22. Patro, G.K., Biswas, A., Ganguly, N., Gummadi, K.P., Chakraborty, A.: FairRec:
two-sided fairness for personalized recommendations in two-sided platforms. Proc.
Web Conf. 2020, 1194–1204 (2020)
23. Plaut, B., Roughgarden, T.: Almost envy-freeness with general valuations. In: Pro-
ceedings of the 29th Annual ACM-SIAM Symposium on Discrete Algorithms, pp.
2584–2603. Society for Industrial and Applied Mathematics (2018)
24. Roth, A.E., Sotomayor, M.: Two-Sided Matching: A Study in Game-Theoretic
Modeling and Analysis. Cambridge University Press, Cambridge (1990)
25. Sotomayor, M.: Three remarks on the many-to-many stable matching problem.
Math. Soc. Sci. 38(1), 55–70 (1999)
26. Sotomayor, M.: Implementation in the many-to-many matching market. Games
Econ. Behav. 46(1), 199–212 (2004)
27. Steinhaus, H.: The problem of fair division. Econometrica 16(1), 101–104 (1948)
28. Sühr, T., Biega, A.J., Zehlike, M., Gummadi, K.P., Chakraborty, A.: Two-sided
fairness for repeated matchings in two-sided markets: a case study of a ride-hailing
platform. In: Proceedings of the 25th ACM SIGKDD International Conference on
Knowledge Discovery & Data Mining, KDD 2019, pp. 3082–3092. ACM, New York
(2019)
29. Wolfson, O., Lin, J.: Fairness versus optimality in ridesharing. In: 2017 18th IEEE
International Conference on Mobile Data Management (MDM), pp. 118–123. IEEE
(2017)
Another random document with
no related content on Scribd:
This mystical error is distinctly characterized in the first chapter of
this gospel, and is there met by the direct assertions, that in Jesus
Christ, the Word, and the God, was not only life, but that the life
itself was the light of men;――and that John the Baptist “was not
the Light, but was only sent to bear witness of the Light;” and
again, with all the tautological earnestness of an old man, the aged
writer repeats the assertion that “this was the true Light, which
enlightens every man that comes into the world.” Against these
same sectaries, the greater part of the first chapter is directed
distinctly, and the whole tendency of the work throughout, is in a
marked manner opposed to their views. With them too, John had
had a local connection, by his residence in Ephesus, where, as it is
distinctly specified in the Acts of the Apostles, Paul had found the
peculiar disciples of John the Baptist long before, on his first visit to
that city; and had successfully preached to some of them, the
religion of Christ, which before was a strange and new thing to them.
The whole tendency and scope of this gospel, indeed, as directed
against these two prominent classes of heretics, both Gnostics and
Sabians, are fully and distinctly summed up in the conclusion of the
twentieth chapter;――“These things are written, that ye might
believe that Jesus is the Christ, the Son of God, and that in
believing on him, ye might have life through his name.”
As to the place where this gospel was written, there is a very

decided difference of opinion among high authorities, both ancient
and modern,――some affirming it to have been composed in
Patmos, during his exile, and others in Ephesus, before or after his
banishment. The best authority, however, seems to decide in favor of
Ephesus, as the place; and this view seems to be most generally
adopted in modern times. Even those who suppose it to have been
written in Patmos, however, grant that it was first given to the
Christian world in Ephesus,――the weight of early authority being
very decided on this latter point. This distinction between the place of
composition and the place of publication, is certainly very reasonable
on some accounts, and is supported by ancient authorities of
dubious date; but there are important objections to the idea of the
composition of both this and the Apocalypse, in the same place,
during about one year, which was the period of his exile. There seem
to be many things in the style of the gospel which would show it to
be a work written at a different period, and under different
circumstances from the Apocalypse; and some Biblical critics, of
high standing, have thought that the gospel bore marks in its style,
which characterized it as a production of a much older man than the
author of the energetic, and almost furious denunciations of the
Apocalypse, must have been. In this case, where ancient authority is
so little decisive, it is but fair to leave the point to be determined by
evidence thus connected with the date, and drawn from the internal
character of the composition itself,――a sort of evidence, on which
the latest moderns are far more capable of deciding than the most
ancient, and the sagest of the Fathers. The date itself is of course
inseparably connected with the determination of the place, and like
that, must be pronounced very uncertain. The greatest probability
about both these points is, that it was written at Ephesus, after his
return from Patmos; for the idea of its being produced before his
banishment, during his first residence in Asia, has long ago been
exploded; nor is there any late writer of authority on these points,
who pretends to support this unfounded notion.
his first epistle.
All that has been said on the character and the objects of the
gospel, may be exactly applied to this very similar production. So
completely does it resemble John’s gospel, in style, language,
doctrines and tendencies, that even a superficial reader might be
ready to pronounce, on a common examination, that they were
written in the same circumstances and with the same object. This
has been the conclusion at which the most learned critics have
arrived, after a full investigation of the peculiarities of both,
throughout; and the standard opinion now is, that they were both
written at the same time and for the same persons. Some reasons
have been given by high critical authority, for supposing that they
were both written at Patmos, and sent together to Ephesus,――the
epistle serving as a preface, dedication, and accompaniment of the
gospel, to those for whom it was intended, and commending the
prominent points in it to their particular attention. This beautiful and
satisfactory view of the object and occasion of the epistle, may
certainly be adopted with great propriety and justice; but in regard to
the places of its composition and direction, a different view is much
more probable, as well as more consistent with the notion, already
presented above, of the date and place of the gospel. It is very
reasonable to suppose that the epistle was written some years after
John’s return to Ephesus,――that it was intended, (along with the
gospel, for the churches of Asia generally, to whom John hoped to
make an apostolic pastoral visit, shortly,) to confirm them in the faith,
as he announces in the conclusion. There is not a single
circumstance in gospel or epistle, which should lead any one to
believe that they were directed to Ephesus in particular. On the
contrary, the total absence of anything like a personal or local
direction to the epistle, shows the justice of its common title, that it is
a “general epistle,” a circular, in short, to all the churches under his
special apostolic supervision,――for whose particular dangers,
errors and necessities, he had written the gospel just sent forth, and
to whom he now minutely commended that work, in the very opening
words of his letter, referring as palpably and undeniably to his
gospel, as any words can express. “Of that which ‘was from the
beginning, of the Word,’ which I have heard, which I have seen with
my eyes, which I have looked upon, which my hands have
handled,――of the Word of Life” &c.; particularizing with all the
minute verbosity of old age, his exact knowledge of the facts which
he gives in his gospel, assuring them thus of the accuracy of his
descriptions. The question concerns his reputation for fidelity as a
historian; and it is easy to see therefore, why he should labor thus to
impress on his readers his important personal advantages for
knowing exactly all the facts he treats of, and all the doctrines which
he gives at such length in the discourses of Christ. Again and again
he says, “I write,” and “I have written,” recapitulating the sum of the
doctrines which he has designed to inculcate; and he particularizes
still farther that he has written to all classes and ages, from the
oldest to the youngest, intending his gospel for the benefit of all. “I
have written to you, fathers,”――“unto you, young men,”――“unto
you, little children,” &c. What else can this imply, than a dedication of
the work concerning “the WORD,” to all stations and ages,――to the
whole of the Christian communities, to whom he commits and
recommends his writings;――as he writes “to the
fathers――because they know him who was from the
beginning,”――in the same way “to the young men, because they
are constant, and the Word of God dwells in them,” and “that the
doctrine they have received may remain unchangeable in them,” and
“on account of those who would seduce them.” He recapitulates
all the leading doctrines of his gospel,――the Messiahship, and the
Divinity of Jesus,――his Unity, and identity with the divine
abstractions of the Gnostic theology. Here too, he inculcates and
renewedly urges the great feeling of Christian brotherly love, which
so decidedly characterizes the discourses of Jesus, as reported in
his gospel. So perfect was the connection of origin and design,
between the gospel and this accompanying letter, that they were
anciently placed together, the epistle immediately following the
gospel; as is indubitably proved by certain marks in ancient
manuscripts.
It was mentioned, in connection with a former part of John’s life, that this epistle is
quoted by Augustin and others, under the title of the epistle to the Parthians. It seems very
probable that this may have been also addressed to those churches in the east, about
Babylon, which had certainly suffered much under the attacks of these same mystical
heretics. It is explained, however, by some, that this was an accidental corruption in the
copying of the Greek.――The second epistle was quoted by Clemens Alexandrinus, under
the title of “the epistle to the virgins,” προς παρθενους, which, as some of the modern critics
say, must have been accidentally changed to παρθους, by dropping some of the syllables,
and afterwards transferred to the first (!) as more appropriate;――a perfectly unauthorized
conjecture, and directly in the face of all rules of criticism.
the second and third epistles.
These are both evidently private letters from John to two of his
intimate personal friends, of whose circumstances nothing whatever
being known, except what is therein contained, the notice of these
brief writings must necessarily be brief also. They are both honorably
referred to, as entertainers of the servants of Jesus Christ as they
travel from place to place, and seem to have been residents in some
of the Asian cities within John’s apostolic circuit, and probably
received him kindly and reverently into their houses on his tours of
duty; and them he was about to visit again shortly. The second
epistle is directed to a Christian female, who, being designated by
the very honorable title of “lady,” was evidently a person of rank; and
from the remark towards the conclusion, about the proper objects of
her hospitality, it is plain that she must have been also a person of
some property. Mention is made of her children as also objects of
warm affection to the aged apostle; and as no other member of her
family is noticed, it is reasonable to conclude that she was a widow.
The contents of this short letter are a mere transcript, almost
verbatim, of some important points in the first, inculcating Christian
love, and watchfulness against deceivers;――(no doubt the
Gnostical heretics,――the Cerinthians and Nicolaitans.) He
apologizes for the shortness of the letter, by saying that he hopes
shortly to visit her; and ends by communicating the affectionate
greetings of her sister’s children, then residents in Ephesus, or
whatever city was then the home of John. The third epistle is
directed to Gaius, (that is, Caius, a Roman name,) whose hospitality
is commemorated with great particularity and gratitude in behalf of
Christian strangers, probably preachers, traveling in his region.
Another person, named Diotrephes, (a Greek by name, and probably
one of the partizans of Cerinthus,) is mentioned as maintaining a
very different character, who, so far from receiving the ministers of
the gospel sent by the apostle, had even excluded from Christian
fellowship those who did exercise this hospitality to the messengers
of the apostle. John speaks threateningly of him, and closes with the
same apology for the shortness of the letter, as in the former. There
are several persons, named Gaius, or Caius, mentioned in apostolic
history; but there is no reason to suppose that any of them was
identified with this man.
For these lucid views of the objects of all these epistles, I am mainly indebted to Hug’s
Introduction, to whom belongs the merit of expressing them in this distinctness, though
others before him have not been far from apprehending their simple force. Michaelis, for
instance, is very satisfactory, and much more full on some points. In respect to the place
whence they were written, Hug appears to be wholly in the wrong, in referring them to
Patmos, just before John’s return. Not the least glimmer of a reason appears, why all the
writings of John should be huddled together in his exile. I can make nothing whatever of the
learned commentator’s reason about the deficiency of “pen, ink and paper,” (mentioned in
Epistle ii. 12, and iii. 13.) as showing that John must still have been in “that miserable
place,” Patmos. The idea seems to require a great perversion of simple words, which do not
seem to be capable of any other sense than that adopted in the above account.
the traditions of his life in ephesus.
To this period of his life, are referred those stories of his miracles
and actions, with which the ancient fictitious apostolic narratives are
so crowded,――John being the subject of more ancient traditions
than any other apostle. Some of those are so respectable and
reasonable in their character, as to deserve a place here, although
none of them are of such antiquity as to deserve any confidence, on
points where fiction has often been so busy. The first which follows,
is altogether the most ancient of all apostolic stories, which are not in
the New Testament; and even if it is a work of fiction, it has such
merits as a mere tale, that it would be injustice to the readers of this
book, not to give them the whole story, from the most ancient and
best authorized record.
It is related that John, after returning from banishment, was often

called to the neighboring churches to organize them, or to heal
divisions, and to ordain elders. On one occasion, after ordaining a
bishop, he committed to his particular care and instruction a fine
young man, whom he saw in the congregation, charging the bishop,
before the whole church, to be faithful to him. The bishop accordingly
took the young man into his house, watched over him, and instructed
him, and at length baptized him. After this, viewing the young man as
a confirmed Christian, the bishop relaxed his watchfulness, and
allowed the youth greater liberties. He soon got into bad company, in
which his talents made him conspicuous, and proceeding from one
step to another, he finally became the leader of a band of robbers. In
this state of things, John came to visit the church, and presently
called upon the bishop to bring forward his charge. The bishop
replied that he was dead,――dead to God;――and was now in the
mountains, a captain of banditti. John ordered a horse to be brought
immediately to the church door, and a guide to attend him; and
mounting, he rode full speed in search of the gang. He soon fell in
with some of them, who seized him, to be carried to their head
quarters. John told them that this was just what he wanted, for he
came on purpose to see their captain. As they drew near, the captain
stood ready to receive them; but on seeing John, he drew back, and
began to make off. John pursued with all the speed his aged limbs
would permit, crying out, “My son, why do you run from your own
father, who is unarmed and aged? Pity me, my son, and do not fear.
There is yet hope of your life. I will intercede for you; and, if
necessary, will cheerfully suffer death for you, as the Lord did for us.
Stop,――believe what I say; Christ hath sent me.” The young man
stopped, looked on the ground, and then throwing down his arms,
came trembling, and with sobs and tears, begged for pardon. The
apostle assured him of the forgiveness of Christ; and conducting him
back to the church, there fasted and prayed with him, and at length
procured his absolution.
Another story, far less probable, is related in the ancient

martyrologies, and by the counterfeit Abdias. Craton, a philosopher,
to make a display of contempt for riches, had persuaded two wealthy
young men, his followers, to invest all their property in two very
costly pearls; and then, in the presence of a multitude, to break
them, and pound them to dust. John happening to pass by, at the
close of the transaction, censured this destruction of property, which
might better have been given in alms to the poor. Craton told him, if
he thought so, he might miraculously restore the dust to solid pearls
again, and have them for charitable purposes. The apostle gathered
up the particles, and holding them in his hand, prayed fervently, that
they might become solid pearls, and when the people said “Amen,” it
took place. By this miracle, Craton, and all his followers, were
converted to Christianity; and the two young men took back the
pearls, sold them, and then distributed the avails in charity.
Influenced by this example, two other young men of distinction,
Atticus and Eugenius, sold their estates, and distributed the avails
among the poor. For a time, they followed the apostle, and
possessed the power of working miracles. But, one day, being at
Pergamus, and seeing some well-dressed young men, glittering in
their costly array, they began to regret that they had sold all their
property, and deprived themselves of the means of making a figure
in the world. John read in their countenances and behavior the state
of their minds; and after drawing from them an avowal of their regret,
he bid them bring him each a bundle of straight rods, and a parcel of
smooth stones from the sea shore. They did so,――and the apostle,
after converting the rods into gold, and the stones into pearls, bid
them take them, and sell them, and redeem their alienated estates, if
they chose. At the same time, he plainly warned them, that the
consequence would be the eternal loss of their souls. While he
continued his long and pungent discourse, a funeral procession
came along. John now prayed, and raised the dead man to life. The
resuscitated person began to describe the invisible world, and so
graphically painted to Atticus and Eugenius the greatness of their
loss, that they were melted into contrition. The apostle ordered them
to do penance thirty days,――till the golden rods should become
wood, and the pearls become stones. They did so, and were
afterwards very distinguished saints.
Another story, of about equal merit, is told by the same authority.

While John continued his successful ministry at Ephesus, the
idolaters there, in a tumult, dragged him to the temple of Diana, and
insisted on his sacrificing to the idol. He warned all to come out of
the temple, and then, by prayer, caused it to fall to the ground, and
become a heap of ruins. Then, addressing the pagans on the spot,
he converted twelve thousand of them in one day. But Aristodemus,
the pagan high priest, could not be convinced, till John had drunken
poison without harm, by which two malefactors were killed instantly,
and also raised the malefactors to life. This resuscitation he
rendered the more convincing to Aristodemus, by making him the
instrument of it. The apostle pulled off his tunic, and gave it to
Aristodemus. “And what is this for?” said the high priest. “To cure you
of your infidelity,” was the reply. “But how is your tunic to cure me of
infidelity?” “Go,” said the apostle, “and spread it upon the dead
bodies, and say: ‘The apostle of our Lord Jesus Christ hath sent me
to resuscitate you, in his name, that all may know, that life and death
are the servants of Jesus Christ, my Lord.’” By this miracle the high
priest was fully convinced; and afterwards convinced the proconsul.
Both of them were baptized,――and persecution, from that time,
ceased. They also built the church dedicated to St. John, at
Ephesus.
For this series of fables I am indebted again to the kindness of Dr. Murdock, in whose
manuscript lectures they are so well translated from the original romances, as to make it
unnecessary for me to repeat the labor of making a new version from the Latin. The sight of
the results of abler efforts directly before me, offers a temptation to exonerate myself from a
tedious and unsatisfactory effort, which is too great to be resisted, while researches into
historical truth have a much more urgent claim for time and exertion.
The only one of all these fables that occurs in the writings of the Fathers, is the first,
which may be pronounced a tolerably respectable and ancient story. It is narrated by
Clemens Alexandrinus, (about A. D. 200.) The story is copied from Clemens Alexandrinus
by Eusebius, from whom we receive it, the original work of Clemens being now lost.
Chrysostom also gives an abridgement of the tale. (I. Paraenes ad Theodosius) Anastasius
Sinaita, Simeon Metaphrastes, Nicephorus Callistus, the Pseudo-Abdias, and the whole
herd of monkish liars, give the story almost verbatim from Clemens; for it is so full in his
account as to need no embellishment to make it a good story. Indeed its completeness in all
these interesting details, is one of the most suspicious circumstances about it; in short, it is
almost too good a story to be true. Those who wish to see all the evidence for and against
its authenticity, may find it thoroughly examined in Lampe’s Prolegomena to a Johannine
Theology (I. v. 4‒10.) It is, on the whole, the best authorized of all the stories about the
apostles, which are given by the Fathers, and may reasonably be considered to have been
true in the essential parts, though the minute details of the conversations, &c., are probably
embellishments worked in by Clemens Alexandrinus, or his informants.
The rest of these stories are, most unquestionably, all unmitigated falsehoods; nor does
any body pretend to find the slightest authority for a solitary particular of them. They are
found no where but in the novels of the Pseudo-Abdias, and the martyrologies. (Abdiae
Babyloniae episcopi et Apostolorum discipuli de Historia, lib. V., St. John.)
his death.
Respecting the close of his life, all antiquity is agreed that it was
not terminated by martyrdom, nor by any violent death whatever, but
by a calm and peaceful departure in the course of nature, at a very
great age. The precise number of years to which he attained can not
be known, because no writer who lived within five hundred years of
his time has pretended to specify his exact age. It is merely
mentioned on very respectable ancient authority, that he survived to
the beginning of the reign of Trajan. This noblest of the successors
of Julius, began his splendid reign in A. D. 98, according to the most
approved chronology; so that if John did not outlive even the first
year of Trajan, his death is brought very near the close of the first
century; and from what has been reasonably conjectured about his
age, compared with that of his Lord, it may be supposed that he
attained upwards of eighty years,――a supposition which agrees
well enough with the statement of some of the Fathers, that he died
worn out with old age.
Jerome has a great deal to say also, about the age of John at the time when he was
called, arguing that he must have been a mere boy at the time, because tradition asserts
that he lived till the reign of Trajan. Lampe very justly objects, however, that this proof
amounts to nothing, if we accept another common tradition, that he lived to the age of 100
years; which, if we count back a century from the reign of Trajan, would require him to have
attained mature age at the time of the call. Neither tradition however, is worth much. Our old
friend Baronius, too, comes in to enlighten the investigation of John’s age, by what he
considers indubitable evidence. He says that John was in his twenty-second year when he
was called, and passing three years with Christ, must have been twenty-five years old at the
time of the crucifixion; “because,” says the sagacious Baronius, “he was then initiated into
the priesthood.” An assertion which Lampe with indignant surprise stigmatizes as showing
“remarkable boldness,” (insignis audacia,) because it contains two very gross
errors,――first in pretending that John was ever made a priest, (sacerdos,) and secondly in
confounding the age required of the Levites with that of the priests when initiated. For
Baronius’s argument resting wholly on the very strange and unfounded notion, that John
was made a priest, is furthermore supported on the idea that the prescribed age for entering
the priesthood was twenty-five years; but in reality, the age thus required was thirty years,
so that if the other part of this idle story was true, this would be enough to overthrow the
conclusion. Lampe also alludes to the absurd idea of the painters, in representing John as a
young man, even while writing his gospel; while in reality all writers agree that that work was
written by him in his old age. This idea of his perpetual youth, once led into a blunder some
foolish Benedictine monks, who found in Constantinople an antique agate intaglio,
representing a young man with a cornucopia, and an eagle, and with a figure of victory
placing a crown on his head. This struck their monkish fancies at once, as an
unquestionable portrait of John, sent to their hands by a miraculous preservation.
Examination however, has shown it to be a representation of the apotheosis of Germanicus.
But even here, the monkish inventors have found room for new
fables; and though the great weight of all ancient testimony deprives
them of the opportunity to enter into the horrible details of a bloody
and agonizing death, they can not refuse themselves the pleasure of
some tedious absurdities, about the manner of his death and burial,
which are barely worth a partial sketch, to show how determined the
apostolic novelists are to follow their heroes to the very last, with the
glories of a fancifully miraculous departure.
The circumstances of his death are described in the

martyrologies, and by Abdias, in this manner. He had a vision
acquainting him with his approaching exit, five days before it
happened. On a Lord’s-day morning, he went to the great church at
Ephesus, bearing his name, and there performed public worship as
usual, at day-break. About the middle of the forenoon, he ordered a
deacon, and some grave diggers, with their tools, to accompany him
to the burying ground. He then set them to digging his grave, while
he, after ordering the multitude to depart, spent the time in prayer.
He once looked into the grave, and bid them dig it deeper. When it
was finished, he took off his outer garment, and spread it in the
grave. Then, standing over it, he made a speech to those present,
(which is not worth repeating,) then gave thanks to God for the
arrival of the time of his release,――and placing himself in the grave,
and wrapping himself up, he instantly expired. The grave was filled
up; and afterwards miracles took place at it, and a kind of manna
issued from it, which possessed great virtues.
There is no need, however, of such fables, to crown with the false

honors of a vain prodigy, the calmly glorious end of the “Last of the
Apostles.” It is enough for the Christian to know, that, with the long,
bright course of almost a century behind him, and with the mighty
works of his later years around him, John closed the solemn
apostolic drama, bearing with him in his late departure the last light
of inspiration, and the last personal “testimony of Jesus, which is the
spirit of prophecy.” Blessed in his works thus following him, he died
in the Lord, and now rests from his labors on the breast of that loved
friend, who cherished so tenderly the youthful Son of
Thunder;――on the bosom of his Redeemer and his Lord,――
“The bosom of his Father and his God.”
PHILIP.
In all the three gospel lists, this apostle is placed fifth in order, the
variations in the arrangements of the preceding making no difference
in his position. In the first chapter of Acts, however, a different
arrangement is made of his name, as will be hereafter mentioned.
The mere mention of his name on the list, is all the notice taken of
him by either of the three first evangelists, and it is only in the gospel
of John, that the slightest additional circumstance can be learned
about him. From this authority it is ascertained that he was of
Bethsaida, the city of Andrew and Peter, and probably also the home
or frequent visiting-place of the sons of Zebedee, by the younger of
whom he is so particularly commemorated. Immediately after the
narration of the introduction of Andrew, John and Peter, to Jesus, in
the first chapter of this gospel, it is said that Jesus next proceeded
from Bethabara into Galilee, and there finds Philip; but the particular
place is not mentioned, though Bethsaida being immediately after
mentioned as his home, very probably was the place of the meeting.
Andrew and Peter, on their return home, had doubtless had no small
talk among their acquaintances, about the wonderful person
announced as the Messiah, to whom they had been introduced, and
had thus satisfied themselves that he was really the divine character
he was said to be. Philip too, must have heard of him in this way,
before he saw him; so that when Jesus met him, he was prepared at
once to receive the call which Jesus immediately gave
him,――“Follow me.” From the circumstance that he was the first
person who was summoned by Jesus, in this particular formula of
invitation to the discipleship, some writers have, not without reason,
claimed for Philip the name and honors of the Protoclete, or “first-
called;” though Andrew has commonly been considered as best
entitled to this dignity, from his being the first mentioned by name, as
actually becoming acquainted with Jesus. Philip was so devoutly
engaged, at once, in the cause of his new Master, that he, like
Andrew, immediately sought out others to share the blessings of the
discipleship; and soon after meeting one of his friends, Nathanael,
he expressed the ardor of his faith in his new teacher, by the words
in which he invited him to join in this honorable fellowship,――“We
have found him of whom Moses, in the law, and all the prophets did
write,――Jesus of Nazareth, the son of Joseph.” The result of this
application will be related in the life of the person most immediately
concerned. After this, no notice whatever is taken of Philip except
where incidental remarks made by him in the conversations of
Jesus, are recorded by John. Thus, at the feeding of the five
thousand, upon Jesus’s asking whether they had the means of
procuring food for the multitude, Philip answered, that “two hundred
pence would not buy enough for them, that every one might take a
little,”――thus showing himself not at all prepared by his previous
faith in Jesus, for the great miracle which was about to happen;
though Jesus had asked the question, as John says, with the actual
design of trying the extent of his confidence in him. He is afterwards
mentioned in the last conversations of Jesus, as saying to him,
“Show us the Father, and it sufficeth us,”――here too, betraying also
a most unfortunate deficiency, both of faith and knowledge, and
implying also a vain desire to gratify his eyes with still more
miraculous displays of the divine power of his Master; though, even
in this respect, he probably was no worse off than all the rest of the
disciples, before the resurrection of Jesus.
Protoclete.――Hammond claims this peculiar honor for Philip, with great zeal. (See
his notes on John i. 43.)
Of his apostleship not one word is recorded in the New Testament,

for he is no where mentioned in the Acts, except as being one of the
apostles assembled in the upper chamber after the ascension; nor
do the epistles contain the slightest allusion to him. Some of the
most ancient authorities among the Fathers, however, are distinct in
their mention of some circumstances of his later life; but all these
accounts are involved in total discredit, by the fact that they make
him identical with Philip the deacon, whose active and zealous
labors in Samaria, and along the coast of Palestine, from Gaza,
through Ashdod to Caesarea, his home, are minutely related in the
Acts, and have been already alluded to, in that part of the life of
Peter which is connected with these incidents. It has always been
supposed, with much reason, in modern times, that the offices of an
apostle and a deacon were so totally distinct and different, that they
could never both be borne by one and the same person; but the
Fathers, even the very ancient ones, seem to have had not the
slightest idea of any such incompatibility; and therefore uniformly
speak of Philip the apostle, as the same person with Philip, one of
the seven deacons, who is mentioned by Luke, in the Acts of the
Apostles, as having lived at Caesarea, in Palestine, with his
daughters, who were virgins and prophetesses. Testimony more
distinct than this, can no where be found, among all the Fathers, on
any point whatever; and very little that is more ancient. Yet how does
it accord with the notions of those who revere these very Fathers as
almost immaculate in truth, and in all intellectual, as well as moral
excellence? What is the evidence of these boasted Fathers worth,
on any point in controversy about apostolic church government, or
doctrine, or criticism, if the modern notion of the incompatibility of the
two offices of apostle and deacon is correct?
The testimony of the Fathers on this point, is simply this. Eusebius (Church History, III.
31,) quotes Polycrates, bishop of Ephesus, who, in his letter to Victor, bishop of Rome,
(written A. D. 195, or 196,) makes mention of Philip in these exact words: “Philip, who was
one of the twelve apostles, died in Hierapolis;” (in Phrygia;) “and so did two of his
daughters, who had grown old in virginity. And another of his daughters, after having
passed her life under the influence of the Holy Spirit, was buried at Ephesus.” This certainly
is a most perfect identification of Philip the apostle with Philip the deacon; for it is this latter
person who is particularly mentioned in Acts, xxi. 8, 9, as “having four daughters who did
prophesy.” He is there especially designated as “Philip the evangelist, one of the seven,”
while Polycrates expressly declares, that this same person “was one of the twelve.”
Eusebius also, in the preceding chapter, quotes Clemens Alexandrinus as mentioning Philip
among those apostles who were married, because he is mentioned as having had
daughters; and Clemens even adds that these were afterwards married, which directly
contradicts the previous statement of Polycrates, that three of them died virgins, in old age.
Yet Eusebius quotes all this stuff, with approbation.
Papias, (A. D. 140,) bishop of Hierapolis, the very place of the death and burial of Philip,
is represented by Eusebius as having been well acquainted with the daughters of Philip,
mentioned in Acts, as the virgin prophetesses. Papias says that he himself “heard these
ladies say that their father once raised a dead person to life, in their time.” But it deserves
notice, that Papias, the very best authority on this subject, is no where quoted as calling this
Philip “an apostle;” though Eusebius, on his own authority, gives this name to the Philip of
whom Papias speaks. It is therefore reasonable to conclude, that this blunder, betraying
such a want of familiarity with the New Testament history, originated after the time of
Papias, whose intimate acquaintance with Philip’s family would have enabled him to say, at
once, that this was the deacon, and not the apostle; though it is not probable that he was
any less deplorably ignorant of the scriptures than most of the Fathers were.
Now what can be said of the testimony of the Fathers on points where they can not refer,
either to their own personal observation, or to informants who have seen and heard what
they testify? The only way in which they can be shielded from the reproach of a gross
blunder and a disgraceful ignorance of the New Testament, is, that they were right in
identifying these two Philips, and that modern theologians are wrong in making the
distinction. On this dilemma I will not pretend to decide; for though so little reverence for the
judgment and information of the Fathers has been shown in this book, there does seem to
me to be some reason for hesitation on this point, where the Fathers ought to have been as
well informed as any body. They must have known surely, whether, according to the notions
of those primitive ages of Christianity, there was any incompatibility between the apostleship
and the deaconship! If their testimony is worth anything on such points, it ought to weigh so
much on this, as to cause a doubt whether they are not right, and the moderns wrong.
However, barely suggesting this query, without attempting a decision, as Luther says, “I will
afford to other and higher spirits, occasion to reflect.”
This is all the satisfaction that the brief records of the inspired or
uninspired historians of Christianity can give the inquirer, on the life
of this apostle;――so unequal were the labors of the first ministers of
Christ, and their claims for notice. Philip, no doubt, served the
purpose for which he was called, faithfully; but in these brief
sketches, there are no traces of any genius of a high character, that
could distinguish him above the thousands that are forgotten, but
whose labors, like those of the minutest animals in a mole-hill,
contribute an indispensable portion to the completion of the mass, in
whose mighty structure all their individual efforts are swallowed up
forever.
And though the ancient Polycrates may have blundered

grievously, in respect to the apostle’s personal identity, his hope of
the glorious resurrection of those whom he supposed to have died in
Asia will doubtless be equally well rewarded, if, to the amazement of
the Fathers, the apostle Philip should rise at last from the dust of
Babylon, or the ashes of Jerusalem, while his namesake, the
evangelist, shall burst from his tomb in Hierapolis. “For,” as
Polycrates truly says, “in Asia, some great lights have gone down,
which shall rise again on that day of the Lord’s approach, when he
shall come from the heavens in glory, and shall raise up all his
saints;――Philip, one of the twelve apostles, who sleeps at
Hierapolis, with his venerable virgin daughters,――John, who lay in
the bosom of the Lord, and who is laid at Ephesus,――Polycarp, at
Smyrna,――Thraseas, at Eumenia,――Sagaris, at
Laodicea,――Papirius and Melito, at Sardis――all await the
visitation of the Lord from the heavens, in which he shall raise them
from the dead.”
NATHANAEL, BAR-THOLOMEW.
his name and call.
In respect to this apostle, there occurs a primary question about

his name, which is given so differently in different sacred authorities,
as to induce a strong suspicion that the two names refer to two
totally distinct persons. The reasons for applying the two words,
Nathanael and Bartholomew, to the same person, are the
circumstances,――that none of the three first evangelists mention
any person named Nathanael, and that John never mentions the
name Bartholomew,――that Bartholomew and Nathanael are each
mentioned on these different authorities, among the chosen disciples
of Jesus,――that Bartholomew is mentioned by the three first
evangelists, on all the lists, directly after Philip, who is by John
represented as his intimate friend,――and that Bartholomew is not
an individual name, but a word showing parentage merely,――the
first syllable being often prefixed to Syriac names, for this purpose;
and Bar-Tholomew means the “son of Tholomew,” or “Tholomai;” just
as Bar-Jonah means the “son of Jonah;” nor was the former any
more in reality the personal, individual name of Nathanael, than the
latter was of Peter; but some circumstance may have occurred to
make it, in this instance, often take the place of the true individual
name.
A few very brief notices are given of this apostle by John, who
alone alludes to him, otherwise than by a bare mention on the list. It
is mentioned in his gospel that Nathanael was of Cana, in Galilee, a
town which stood about half-way between lake Gennesaret and the
Mediterranean sea; but the circumstances of his call seem to show
that he was then with Philip, probably at or near Bethsaida. Philip,
after being summoned by Jesus to the discipleship, immediately
sought to bring his friend Nathanael into an enjoyment of the honors
of a personal intercourse with Jesus, and invited him to become a
follower of the Messiah, foretold by Moses and the prophets, who
had now appeared, as Jesus of Nazareth, the son of Joseph. On
hearing of that mean place, as the home of the promised King of
Israel, Nathanael, with great scorn, replied, in inquiry, “Can any good
thing come out of Nazareth?” To this sneering question, Philip
answered by the simple proposition, “Come and see;”――wisely
judging that no argument could answer his friend’s prejudice so well
as an actual observation of the character and aspect of the
Nazarene himself. Nathanael, accordingly, persuaded by the
earnestness of his friend, came along with him, perhaps, partly to
gratify him, but, no doubt, with his curiosity somewhat moved to
know what could have thus brought Philip into this devout regard for
a citizen of that dirty little town; and he therefore readily
accompanied him to see what sort of prophet could come out of
Nazareth.
The words with which Jesus greeted Nathanael, even before he

had been personally introduced, or was prepared for any salutation,
are the most exalted testimonial of his character that could be
conceived, and show at once his very eminent qualifications for the
high honors of the apostleship. When Jesus saw Nathanael coming
to him, he said, “Behold a true son of Israel, in whom is no
guile!”――manifesting at once a confidential and intimate knowledge
of his whole character, in thus pronouncing with such ready decision,
this high and uncommon tribute of praise upon him, as soon as he
appeared before him. Nathanael, quite surprised at this remarkable
compliment from one whom he had never seen until that moment,
and whom he supposed to be equally ignorant of him, replied with
the inquiry, “Whence knowest thou me?” Jesus answered, “Before
Philip called thee, when thou wast under the fig-tree, I saw thee.”
The fig-trees of Palestine, presenting a wide, leafy cover, and a
delightful shade, were often used in the warm season as places of
retirement, either in company, for conversation, or in solitude, for
meditation and prayer, as is shown in numerous passages of the
Rabbinical writings; and it was, doubtless, in one of these
occupations that Nathanael was engaged, removed, as he
supposed, from all observation, at the time to which Jesus referred.
But the eye that could pierce the stormy shades of night on the
boisterous waves of Galilee, and that could search the hearts of all
men, could also penetrate the thick, leafy veil of the fig-tree, and
observe the most secret actions of this guileless Israelite, when he
supposed the whole world to be shut out, and gave himself to the
undisguised enjoyment of his thoughts, feelings, and actions, without
restraint. Nathanael, struck with sudden but absolute conviction, at
this amazing display of knowledge, gave up all his proud scruples
against the despised Nazarene, and adoringly exclaimed, “Rabbi!
thou art the Son of God,――thou art the King of Israel.” Jesus,
recognizing with pleasure the ready faith of this pure-minded
disciple, replied, “Because I said unto thee, ‘I saw thee under the fig-
tree,’――believest thou? Thou shalt see yet greater things than
these.” Then turning to Philip as well as to Nathanael, he says to
them both, “I solemnly assure you, hereafter ye shall see heaven
open, and the angels of God ascending and descending upon the
Son of Man.”
On the day but one after this occurrence, as John records, Jesus
was in Cana of Galilee, the residence of Nathanael, and was present
at a wedding which took place there. From the circumstance that the
mother of Jesus was there also, it would seem likely that it was the
marriage of some of their family friends; otherwise the conjecture
might seem allowable, that the presence of Jesus and his disciples
on this occasion, was in some way connected with the introduction of
Nathanael to Jesus; and that this new disciple may have been some
way concerned in this interesting event. The manner in which the
occurrence is announced,――it being next specified, that two days
after the occurrences recorded in the end of the first chapter, Jesus
was present at a marriage in Cana of Galilee,――would seem to
imply very fairly, that Jesus had been in some other place
immediately before; and it is probable therefore, that he
accompanied Nathanael home from Bethsaida, or whatever place
was the scene of his calling to the discipleship, along with Philip. The
terms of the statement are not, however, absolutely incompatible
with the idea of this first introduction of these two disciples to Jesus,
in Cana itself, which may have been the part of Galilee into which
Jesus is said to have gone forth, after leaving Bethabara; although,
the reasons above given make it probable that Bethsaida was the
scene. After this first incident, no mention whatever is made of
Nathanael, either under his proper name, or his paternal appellation,
except that when the twelve were sent forth in pairs, he was sent
with his friend Philip, that those who had been summoned to the
work together, might now go forth laboring together in this high
commission. One solitary incident is also commemorated by John, in
which this apostle was concerned, namely, the meeting on the lake
of Gennesaret, after the resurrection, where his name is mentioned
among those who went out on the fishing excursion with Peter. His
friend Philip is not there mentioned, but may have been one of the
“two disciples,” who are included without their names being given.
From this trifling circumstance, some have concluded that Nathanael
was a fisherman by trade, as well as the other four who are
mentioned with him; and certainly the conjecture is reasonable, and
not improbable, except from the circumstance, that his residence
was at Cana, which is commonly understood to have been an inland
town, and too far from the water, for any of its inhabitants to follow

Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF

Uploaded by

Copyright:

Available Formats

You might also like

Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Full Chapter Web and Internet Economics 16Th International Conference Wine 2020 Beijing China December 7 11 2020 Proceedings Xujin Chen PDF

Uploaded by

Copyright:

Available Formats

Web and Internet Economics 16th

International Conference WINE 2020

Web and Internet Economics 10th International

Web and Internet Economics 12th International

Web and Internet Economics 14th International

Web and Internet Economics 15th International

Bio inspired Information and Communication Technologies

Information Systems Security 16th International

Combinatorial Optimization and Applications 14th

Distributed Computing and Internet Technology 16th

Editorial Board Members

Advanced Research in Computing and Software Science

Subline Series Editors

Subline Advisory Board

Martin Hoefer Ruta Mehta (Eds.)

ISSN 0302-9743 ISSN 1611-3349 (electronic)

© Springer Nature Switzerland AG 2020

October 2020 Xujin Chen

Program Committee Chairs

Yuqing Kong Peking University, China

Andreas Abels Yurong Chen

Yiding Feng Daniel Moroz

José Verschae Xiang Yan

Almost Envy-Free Repeated Matching in Two-Sided Markets. . . . . . . . . . . . 3

Dynamic Weighted Matching with Heterogeneous Arrival

A Fine-Grained View on Stable Many-To-One Matching Problems

The Ad Types Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Multidimensional Stable Roommates with Master List . . . . . . . . . . . . . . . . . 59

Optimal Nash Equilibria for Bandwidth Allocation . . . . . . . . . . . . . . . . . . . 77

Counteracting Inequality in Markets via Convex Pricing . . . . . . . . . . . . . . . 89

Mechanism Design and Pricing

Two Strongly Truthful Mechanisms for Three Heterogeneous Agents

A Generic Truthful Mechanism for Combinatorial Auctions . . . . . . . . . . . . . 133

The Price of Anarchy of Two-Buyer Sequential Multiunit Auctions. . . . . . . . 147

Revenue-Optimal Deterministic Auctions for Multiple Buyers with Ordinal

Robust Revenue Maximization Under Minimal Statistical Information . . . . . . 177

Revenue Monotonicity Under Misspecified Bidders . . . . . . . . . . . . . . . . . . . 191

On the Power and Limits of Dynamic Pricing in Combinatorial Markets . . . . 206

Competitively Pricing Parking in a Tree. . . . . . . . . . . . . . . . . . . . . . . . . . . 220

Routing, Scheduling, Load Balancing

The Price of Anarchy for Instantaneous Dynamic Equilibria . . . . . . . . . . . . . 237

Data-Driven Models of Selfish Routing: Why Price of Anarchy Does

Competition Alleviates Present Bias in Task Completion . . . . . . . . . . . . . . . 266

Improving Approximate Pure Nash Equilibria in Congestion Games . . . . . . . 280

The Curse of Rationality in Sequential Scheduling Games . . . . . . . . . . . . . . 295

Sequential Solutions in Machine Scheduling Games . . . . . . . . . . . . . . . . . . 309

Nash Social Welfare in Selfish and Online Load Balancing . . . . . . . . . . . . . 323

Simultaneously Achieving Ex-ante and Ex-post Fairness . . . . . . . . . . . . . . . 341

Optimal Bounds on the Price of Fairness for Indivisible Goods. . . . . . . . . . . 356

Consensus Halving for Sets of Items . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384

Learning Strong Substitutes Demand via Queries . . . . . . . . . . . . . . . . . . . . 401

A Cardinal Comparison of Experts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416

Minimum-Regret Contracts for Principal-Expert Problems . . . . . . . . . . . . . . 430

Bayesian Repeated Zero-Sum Games with Persistent State,

Large Random Matching Markets with Localized Preference Structures Can

Online Hypergraph Matching with Delays . . . . . . . . . . . . . . . . . . . . . . . . . 465

Market Equilibrium in Multi-tier Supply Chain Networks. . . . . . . . . . . . . . . 467

Decision Scoring Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468