Two-Sided Fairness for Repeated Matchings in Two-Sided Markets: A Case Study of a Ride-Hailing Platform

pp. 3082-3092, 2019.

Cited by: 10|Bibtex|Views131|DOI:https://doi.org/10.1145/3292500.3330793
EI
Other Links: dl.acm.org|dblp.uni-trier.de|academic.microsoft.com
Weibo:
This paper explored the problem of two-sided fairness for producers and consumers in a sharing economy platform

Abstract:

Ride hailing platforms, such as Uber, Lyft, Ola or DiDi, have traditionally focused on the satisfaction of the passengers, or on boosting successful business transactions. However, recent studies provide a multitude of reasons to worry about the drivers in the ride hailing ecosystem. The concerns range from bad working conditions and work...More

Code:

Data:

0
Introduction
  • Two-sided sharing economy platforms, such as Uber, Lyft, or AirBnB, have brought in disruptions in multiple business landscapes [25, 31].
  • There are three stakeholders in the two-sided economy: (i) providers of goods and services, (ii) customers who pay for them, and (iii) the platform which performs the matching between the providers and customers.
  • With increasingly many people depending on the sharing economy to earn a living, it becomes crucial to ask the question of what a fair distribution of income on such platforms is and what power and means the platform has in shaping these distributions.
  • Allon et al [1] studied how drivers make labor decisions in such on-demand services e.g., when to work and for how long, and attempted to design better financial incentives for drivers to increase their number of work hours
Highlights
  • Two-sided sharing economy platforms, such as Uber, Lyft, or AirBnB, have brought in disruptions in multiple business landscapes [25, 31]
  • We propose a novel framework to think about fairness in the matching mechanisms of ride hailing platforms
  • According to a 2016 report by the United States Department of Commerce, the total spending in the sharing economy was 5% of the total spending, and this ratio is predicted to grow to 50% by 2025 [29]
  • With increasingly many people depending on the sharing economy to earn a living, it becomes crucial to ask the question of what a fair distribution of income on such platforms is and what power and means the platform has in shaping these distributions
  • We focus on studying whether it is possible to optimize the matchings for equitable income distributions
  • This paper explored the problem of two-sided fairness for producers and consumers in a sharing economy platform
Methods
  • TO MATCH DRIVERS AND CUSTOMERS

    the authors consider different strategies to match drivers and customers in ride-hailing platforms.

    5.1 Nearest Driver First (NDF)

    Under this strategy, each customer, making a request in a given round, chooses a driver available in the round that is closest to the customer.
  • Ties are broken randomly
  • The pseudocode for this algorithm is provided in the Reproducibility Supplement A.
  • It is worth making the following observation : Any divergence from the Nearest Driver strategy will necessarily decrease the average utility of the drivers in a given round under the assumed utility functions if ties are broken perfectly.
  • Assuming that variables Mi,j encode a proper matching where all requesting customers get matched to a driver, the value of the expression Mi,j · d (C j
Results
  • The authors analyze the performance of the ILP with Objectives 2 and 3, using λ = {0, 0.5, 1.0}.
  • The authors describe the dataset and the preprocessing methods in the Reproducibility Appendix A.1.
  • The filtered dataset consists of 28k matching rounds with a total of 1462 drivers and 231k job requests.
  • The average duration of a job is 14.39 rounds, which corresponds to 22 minutes.
  • The average distance from pickup location to the destination of the customer is 8.89 km.
  • The number of active rounds
Conclusion
  • This paper explored the problem of two-sided fairness for producers and consumers in a sharing economy platform.
  • The authors have explored the space of various mechanisms for achieving these fairness goals.
  • The authors experimented with an optimization problem balancing equality of both consumers and producers, which led to an increase in equality while preserving the utility volumes.
  • The authors showed that implementing an optimization objective which is a direct translation of equality goals in practice still requires a very careful thought about initialization of utility values
Summary
  • Introduction:

    Two-sided sharing economy platforms, such as Uber, Lyft, or AirBnB, have brought in disruptions in multiple business landscapes [25, 31].
  • There are three stakeholders in the two-sided economy: (i) providers of goods and services, (ii) customers who pay for them, and (iii) the platform which performs the matching between the providers and customers.
  • With increasingly many people depending on the sharing economy to earn a living, it becomes crucial to ask the question of what a fair distribution of income on such platforms is and what power and means the platform has in shaping these distributions.
  • Allon et al [1] studied how drivers make labor decisions in such on-demand services e.g., when to work and for how long, and attempted to design better financial incentives for drivers to increase their number of work hours
  • Methods:

    TO MATCH DRIVERS AND CUSTOMERS

    the authors consider different strategies to match drivers and customers in ride-hailing platforms.

    5.1 Nearest Driver First (NDF)

    Under this strategy, each customer, making a request in a given round, chooses a driver available in the round that is closest to the customer.
  • Ties are broken randomly
  • The pseudocode for this algorithm is provided in the Reproducibility Supplement A.
  • It is worth making the following observation : Any divergence from the Nearest Driver strategy will necessarily decrease the average utility of the drivers in a given round under the assumed utility functions if ties are broken perfectly.
  • Assuming that variables Mi,j encode a proper matching where all requesting customers get matched to a driver, the value of the expression Mi,j · d (C j
  • Results:

    The authors analyze the performance of the ILP with Objectives 2 and 3, using λ = {0, 0.5, 1.0}.
  • The authors describe the dataset and the preprocessing methods in the Reproducibility Appendix A.1.
  • The filtered dataset consists of 28k matching rounds with a total of 1462 drivers and 231k job requests.
  • The average duration of a job is 14.39 rounds, which corresponds to 22 minutes.
  • The average distance from pickup location to the destination of the customer is 8.89 km.
  • The number of active rounds
  • Conclusion:

    This paper explored the problem of two-sided fairness for producers and consumers in a sharing economy platform.
  • The authors have explored the space of various mechanisms for achieving these fairness goals.
  • The authors experimented with an optimization problem balancing equality of both consumers and producers, which led to an increase in equality while preserving the utility volumes.
  • The authors showed that implementing an optimization objective which is a direct translation of equality goals in practice still requires a very careful thought about initialization of utility values
Tables
  • Table1: Performance of all Matching-Mechanisms after 30 Days; Green values indicate a positive change vs. NDF, Red values indicate a negative change vs. NDF
Download tables as Excel
Funding
  • This research was supported in part by a European Research Council (ERC) Advanced Grant for the project “Foundations for Fair Social Computing", funded under the European Union’s Horizon 2020 Framework Programme (grant agreement no. 789373)
Reference
  • Gad Allon, Maxime Cohen, and Park Sinchaisri. 2018. The Impact of Behavioral and Economic Drivers on Gig Economy Workers. SRN 3274628 (2018).
    Google ScholarLocate open access versionFindings
  • Anthony B Atkinson. 1975. The economics of inequality. (1975).
    Google ScholarFindings
  • Asia J Biega, Krishna P Gummadi, and Gerhard Weikum. 2018. Equity of Attention: Amortizing Individual Fairness in Rankings. In ACM SIGIR.
    Google ScholarLocate open access versionFindings
  • Eszter Bokanyi and Aniko Hannak. 2019. Evaluating Algorithm Fairness in an
    Google ScholarFindings
  • Felix Brandt, Vincent Conitzer, Ulle Endriss, Jérôme Lang, and Ariel D Procaccia.
    Google ScholarLocate open access versionFindings
  • 201Handbook of computational social choice. Cambridge University Press.
    Google ScholarFindings
  • [6] Ryan Calo and Alex Rosenblat. 201The taking economy: Uber, information, and power. Columbia Law Review (2017).
    Google ScholarLocate open access versionFindings
  • [7] Abhijnan Chakraborty, Aniko Hannak, Asia J Biega, and Krishna P Gummadi.
    Google ScholarFindings
  • 2017. Fair Sharing for Sharing Economy Platforms. (2017).
    Google ScholarFindings
  • [8] Abhijnan Chakraborty, Gourab K Patro, Niloy Ganguly, Krishna P Gummadi, and Patrick Loiseau. 2019. Equality of voice: Towards fair representation in crowdsourced top-k recommendations. In ACM FAT*. 129–138.
    Google ScholarLocate open access versionFindings
  • [9] Harshal A Chaudhari, John W Byers, and Evimaria Terzi. 2018. Putting Data in the Driver’s Seat: Optimizing Earnings for On-Demand Ride-Hailing. In WSDM.
    Google ScholarFindings
  • [10] Le Chen, Alan Mislove, and Christo Wilson. 2015. Peeking Beneath the Hood of Uber. In ACM IMC.
    Google ScholarLocate open access versionFindings
  • [11] Ettore Damiano and Ricky Lam. 2005. Stability in dynamic matching markets. Games and Economic Behavior 52, 1 (2005).
    Google ScholarLocate open access versionFindings
  • [12] Patricia Everaere, Maxime Morge, and Gauthier Picard. 2013. Minimal concession strategy for reaching fair, optimal and stable marriages. In AAMAS.
    Google ScholarFindings
  • [13] David Gale and Lloyd S Shapley. 1962. College admissions and the stability of marriage. The American Mathematical Monthly 69, 1 (1962).
    Google ScholarLocate open access versionFindings
  • [14] Carla Green and Sam Levin. 2017. Homeless, assaulted, broke: drivers left behind as Uber promises change at the top. https://theguardian.com/usnews/2017/jun/17/uber-drivers-homeless-assault-travis-kalanick.
    Findings
  • [15] Jonathan Hall, Cory Kendrick, and Chris Nosko. 2015. The effects of Uber’s surge pricing: A case study. Univ. of Chicago Booth School of Business (2015).
    Google ScholarFindings
  • [16] Anikó Hannák, Claudia Wagner, David Garcia, Alan Mislove, Markus Strohmaier, and Christo Wilson. 2017. Bias in Online Freelance Marketplaces: Evidence from TaskRabbit and Fiverr. In ACM CSCW.
    Google ScholarLocate open access versionFindings
  • [17] Ernan Haruvy and M Utku Ünver. 2007. Equilibrium selection and the role of information in repeated matching markets. Economics Letters 94, 2 (2007).
    Google ScholarLocate open access versionFindings
  • [18] Kazuo Iwama, Shuichi Miyazaki, and Hiroki Yanagisawa. 2010. Approximation algorithms for the sex-equal stable marriage problem. ACM Trans. on Algorithms 7, 1 (2010), 2.
    Google ScholarLocate open access versionFindings
  • [19] Yongzheng Jia, Wei Xu, and Xue Liu. 2017. An optimization framework for online ride-sharing markets. In IEEE ICDCS.
    Google ScholarLocate open access versionFindings
  • [20] Michael Luca. 2016.
    Google ScholarFindings
  • [21] F Masarani and Sadik S Gokturk. 1989. On the existence of fair matching algorithms. Theory and Decision 26, 3 (1989).
    Google ScholarLocate open access versionFindings
  • [22] Hughston McBain. 1944. Are customers always right. The Rotarian (1944).
    Google ScholarFindings
  • [23] Erhun Ozkan and Amy Ward. 2017. Dynamic matching for real-time ridesharing.
    Google ScholarFindings
  • [24] Alvin E Roth. 1986. On the allocation of residents to rural hospitals: a general property of two-sided matching markets. Econometrica: Journal of the Econometric Society (1986).
    Google ScholarLocate open access versionFindings
  • [25] Juliet Schor. 2016. DEBATING THE SHARING ECONOMY. Journal of SelfGovernance & Management Economics 4, 3 (2016).
    Google ScholarLocate open access versionFindings
  • [26] Anthony F Shorrocks. 1980. The class of additively decomposable inequality measures. Econometrica: Journal of the Econometric Society (1980).
    Google ScholarLocate open access versionFindings
  • [27] Ashudeep Singh and Thorsten Joachims. 2018. Fairness of Exposure in Rankings. In ACM SIGKDD.
    Google ScholarLocate open access versionFindings
  • [28] Arun Sundararajan. 2014. Trusting the ‘sharing economy’to regulate itself. The New York Times 3 (2014).
    Google ScholarLocate open access versionFindings
  • [29] Rudy Telles Jr. 2016. Digital matching firms: A new definition in the “sharing economy” space. ESA issue brief 01-16 (2016).
    Google ScholarLocate open access versionFindings
  • [30] Siddharth Tiwari. 2017. ‘Underhand tactics’ boost sales in e-commerce. sundayguardianlive.com/investigation/9881-underhand-tactics-boost-sales-ecommerce.
    Google ScholarFindings
  • [31] Georgios Zervas, Davide Proserpio, and John W Byers. 2017. The rise of the sharing economy: Estimating the impact of Airbnb on the hotel industry. Journal of Marketing Research 54, 5 (2017).
    Google ScholarLocate open access versionFindings
  • [32] Stephen M Zoepf, Stella Chen, Paa Adu, and Gonzalo Pozo. 2018. The Economics of Ride-Hailing: Driver Revenue, Expenses, and Taxes. Memory: 256GB DDR3, 1866 MHz, ECC
    Google ScholarLocate open access versionFindings
  • 2 http://www.gurobi.com
    Findings
Full Text
Your rating :
0

 

Tags
Comments