Predicting Individual Treatment Effects of Large-scale Team Competitions in a Ride-sharing Economy

KDD '20: The 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining Virtual Event CA USA July, 2020, pp. 2368-2377, 2020.

Cited by: 0|Bibtex|Views147|DOI:https://doi.org/10.1145/3394486.3403286
EI
Other Links: arxiv.org|dl.acm.org|academic.microsoft.com|dblp.uni-trier.de
Weibo:
We present the first predictive analysis of individual treatment effects of team competitions in DiDi, a leading platform of the ridesharing economy

Abstract:

Millions of drivers worldwide have enjoyed financial benefits and work schedule flexibility through a ride-sharing economy, but meanwhile they have suffered from the lack of a sense of identity and career achievement. Equipped with social identity and contest theories, financially incentivized team competitions have been an effective inst...More

Code:

Data:

0
Introduction
  • The rise of the sharing economy has brought dramatic changes to work and life in modern society.
  • While the drivers enjoy all the values of the ride-sharing economy [9], they commonly complain about new barriers to job satisfaction and retention, such as working alone, having few bonds with colleagues, no clear career paths, and a lack of a sense of achievement (e.g., [18]).
  • Drivers are able to (1) build team identity and social bonds with teammates; (2) create a sense of achievement by winning a competition; and (3) increase their satisfaction and performance at work [2].
  • The increase in driver productivity often outweighs the cost of organizing and providing financial incentives for these competitions, which creates a win-win situation for both the drivers and the platform
Highlights
  • The rise of the sharing economy has brought dramatic changes to work and life in modern society
  • We train effective machine learning models on large-scale data collected from hundreds of historical experiments, which combine a comprehensive set of features of individual drivers, teams, contest designs, and experimental environments, and we evaluate the models on out-sample experiments
  • We examine the most predictive features nominated from both models
  • We present the first predictive analysis of individual treatment effects of team competitions in DiDi, a leading platform of the ridesharing economy
  • Our findings present many new insights and useful implications for the research and business practices of team competition, the sharing economy, and online field experiments in general
  • While the the best-performing models have already improved the baseline by 24% and generated lots of insights, the accuracy numbers do not look perfect
  • Future directions of the work include testing these insights with field experiments, investigating the causal links between the heterogeneous factors and the individual treatment effect (ITE), and generalizing the procedure to other sharing economy platforms
Results
  • 6.1 Which Features Predict Treatment Effects?

    The authors examine the most predictive features nominated from both models.
  • The largest factor by Lasso for the individual treatment effect is the proportion of snowstorm days during a competition ( < .01).
  • This is easy to understand as severe weather conditions would limit travel activities and driving efficiency
Conclusion
  • The authors present the first predictive analysis of individual treatment effects of team competitions in DiDi, a leading platform of the ridesharing economy.
  • The authors' findings present many new insights and useful implications for the research and business practices of team competition, the sharing economy, and online field experiments in general.
  • Future directions of the work include testing these insights with field experiments, investigating the causal links between the heterogeneous factors and the ITE, and generalizing the procedure to other sharing economy platforms
Summary
  • Introduction:

    The rise of the sharing economy has brought dramatic changes to work and life in modern society.
  • While the drivers enjoy all the values of the ride-sharing economy [9], they commonly complain about new barriers to job satisfaction and retention, such as working alone, having few bonds with colleagues, no clear career paths, and a lack of a sense of achievement (e.g., [18]).
  • Drivers are able to (1) build team identity and social bonds with teammates; (2) create a sense of achievement by winning a competition; and (3) increase their satisfaction and performance at work [2].
  • The increase in driver productivity often outweighs the cost of organizing and providing financial incentives for these competitions, which creates a win-win situation for both the drivers and the platform
  • Objectives:

    The objective of this study is not to prove the causal effect of team competition but to predict individual driver’s performance in out-of-sample/future competitions.
  • Results:

    6.1 Which Features Predict Treatment Effects?

    The authors examine the most predictive features nominated from both models.
  • The largest factor by Lasso for the individual treatment effect is the proportion of snowstorm days during a competition ( < .01).
  • This is easy to understand as severe weather conditions would limit travel activities and driving efficiency
  • Conclusion:

    The authors present the first predictive analysis of individual treatment effects of team competitions in DiDi, a leading platform of the ridesharing economy.
  • The authors' findings present many new insights and useful implications for the research and business practices of team competition, the sharing economy, and online field experiments in general.
  • Future directions of the work include testing these insights with field experiments, investigating the causal links between the heterogeneous factors and the ITE, and generalizing the procedure to other sharing economy platforms
Tables
  • Table1: Summary of Statistics Number Item
  • Table2: Feature Examples (More Details in Supplement
  • Table3: Model Performance, Evaluated by RMSE
  • Table4: Performance of Three Prototype Contests under the Original Design and Simulated New Designs
  • Table5: Examples of Features with Detailed Description
Download tables as Excel
Related work
  • This study is related to the following lines of literature: Sharing economy. A growing literature investigates the socioeconomic effects on and consequences of ride-sharing platforms, such as Uber and Lyft [36]. Inspired by the findings in [17] that economic gains positively influence people’s intention to participate, a stream of work quantifies the positive effect of financial incentives, such as subsidy [14], on improving supply-demand efficiency. Our study adds to this literature by investigating the effect of rewarded team competitions on service provision in a ride-sharing economy.

    Team competition. Team competitions have been increasingly applied in online communities, such as crowdsourcing [27], education [28], online games [11], and charitable giving [10]. It has been shown that team competitions are effective in improving key metrics, such as participation [28]. Data-mining researchers have developed team matching algorithms to ensure team formation of high efficiency, effectiveness, and fairness, taking into account factors such as demographics, social networks, and tasks (e.g., [1, 4, 35]).
Funding
  • This work is funded in part by the DiDi Research Partnership with the University of Michigan
  • Jieping Ye and Lingyu Zhang’s work is in part funded by the National Key Research and Development Program of China under grant 2018AAA0101100
  • Qiaozhu Mei’s work is in part supported by the National Science Foundation under grant no. 1633370
Reference
  • Wei Ai, Roy Chen, Yan Chen, Qiaozhu Mei, and Webb Phillips. 2016. Recommending teams promotes prosocial lending in online microfinance. Proceedings of the National Academy of Sciences 113, 52 (2016), 14944–14948.
    Google ScholarLocate open access versionFindings
  • Wei Ai, Yan Chen, Qiaozhu Mei, Jieping Ye, and Lingyu Zhang. 2019. Putting teams into the gig economy: A field experiment at a ride-sharing platform. (2019). University of Michigan Working Paper.
    Google ScholarLocate open access versionFindings
  • George A Akerlof and Rachel E Kranton. 2000. Economics and identity. The Quarterly Journal of Economics 115, 3 (2000), 715–753.
    Google ScholarLocate open access versionFindings
  • Aris Anagnostopoulos, Luca Becchetti, Carlos Castillo, Aristides Gionis, and Stefano Leonardi. 2012. Online team formation in social networks. In Proceedings of the 21st international conference on World Wide Web. ACM, 839–848.
    Google ScholarLocate open access versionFindings
  • Joshua D Angrist and Jörn-Steffen Pischke. 2008. Mostly harmless econometrics: An empiricist’s companion. Princeton University Press.
    Google ScholarFindings
  • Susan Athey and Guido Imbens. 201Recursive partitioning for heterogeneous causal effects. Proceedings of the National Academy of Sciences 113, 27 (2016), 7353–7360.
    Google ScholarLocate open access versionFindings
  • Stephen P Borgatti, Ajay Mehra, Daniel J Brass, and Giuseppe Labianca. 2009. Network analysis in the social sciences. Science 323, 5916 (2009), 892–895.
    Google ScholarLocate open access versionFindings
  • Léon Bottou, Jonas Peters, Joaquin Quiñonero-Candela, Denis X Charles, D Max Chickering, Elon Portugaly, Dipankar Ray, Patrice Simard, and Ed Snelson. 2013. Counterfactual reasoning and learning systems: The example of computational advertising. The Journal of Machine Learning Research 14, 1 (2013), 3207–3260.
    Google ScholarLocate open access versionFindings
  • M Keith Chen, Peter E Rossi, Judith A Chevalier, and Emily Oehlsen. 201The value of flexible work: Evidence from Uber drivers. Journal of Political Economy 127, 6 (2019), 2735–2794.
    Google ScholarLocate open access versionFindings
  • Roy Chen, Yan Chen, Yang Liu, and Qiaozhu Mei. 2017. Does team competition increase pro-social lending? Evidence from online microfinance. Games and Economic Behavior 101 (2017), 311–333.
    Google ScholarLocate open access versionFindings
  • Ziqiang Cheng, Yang Yang, Chenhao Tan, Denny Cheng, Alex Cheng, and Yueting Zhuang. 2019. What makes a good team? A large-scale study on the effect of team composition in Honor of Kings. In The World Wide Web Conference. 2666–2672.
    Google ScholarFindings
  • Robert B Cialdini and Melanie R Trost. 1998. Social influence: Social norms, conformity and compliance. The Handbook of Social Psychology (1998), 151–192.
    Google ScholarLocate open access versionFindings
  • Gang Fang, Izabela E Annis, Jennifer Elston-Lafata, and Samuel Cykert. 2019. Applying machine learning to predict real-world individual treatment effects: Insights from a virtual patient cohort. Journal of the American Medical Informatics Association 26, 10 (2019), 977–988.
    Google ScholarLocate open access versionFindings
  • Zhixuan Fang, Longbo Huang, and Adam Wierman. 2019. Prices and subsidies in the sharing economy. Performance Evaluation (2019), 102037.
    Google ScholarLocate open access versionFindings
  • Jerome H Friedman. 2002. Stochastic gradient boosting. Computational Statistics and Data Analysis 38, 4 (2002), 367–378.
    Google ScholarLocate open access versionFindings
  • Ruocheng Guo, Jundong Li, and Huan Liu. 2020. Learning individual causal effects from networked observational data. In Proceedings of the 13th International Conference on Web Search and Data Mining. 232–240.
    Google ScholarLocate open access versionFindings
  • Juho Hamari, Mimmi Sjöklint, and Antti Ukkonen. 2016. The sharing economy: Why people participate in collaborative consumption. Journal of the association for information science and technology 67, 9 (2016), 2047–2059.
    Google ScholarLocate open access versionFindings
  • Nathan Heller. 2017. Is the gig economy working? The New Yorker (2017). https://www.newyorker.com/magazine/2017/05/15/is-the-gig-economy-working
    Locate open access versionFindings
  • Benjamin E. Hermalin. 1997. Toward an economic theory of leadership: Leading by example. The American Economic Review 88, 5 (1997), 1188–1206.
    Google ScholarLocate open access versionFindings
  • Arthur E Hoerl and Robert W Kennard. 1970. Ridge regression: Biased estimation for nonorthogonal problems. Technometrics 12, 1 (1970), 55–67.
    Google ScholarLocate open access versionFindings
  • Ron Kohavi, Alex Deng, Brian Frasca, Toby Walker, Ya Xu, and Nils Pohlmann. 2013. Online controlled experiments at large scale. In Proceedings of the 19th ACM SIGKDD. ACM, 1168–1176.
    Google ScholarLocate open access versionFindings
  • Robert E Kraut, Susan R Fussell, Susan E Brennan, and Jane Siegel. 2002. Understanding effects of proximity on collaboration: Implications for technologies to support remote collaborative work. Distributed work (2002), 137–162.
    Google ScholarLocate open access versionFindings
  • Maggie Makar, Adith Swaminathan, and Emre Kıcıman. 2019. A distillation approach to data efficient individual treatment effect estimation. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 33. 4544–4551.
    Google ScholarLocate open access versionFindings
  • Benny Moldovanu, Aner Sela, and Xianwen Shi. 2007. Contests for status. Journal of political Economy 115, 2 (2007), 338–363.
    Google ScholarLocate open access versionFindings
  • Lev Muchnik, Sinan Aral, and Sean J Taylor. 2013. Social influence bias: A randomized experiment. Science 341, 6146 (2013), 647–651.
    Google ScholarLocate open access versionFindings
  • Alain Pinsonneault, Henri Barki, R Brent Gallupe, and Norberto Hoppen. 1999. Electronic brainstorming: The illusion of productivity. Information Systems Research 10, 2 (1999), 110–133.
    Google ScholarLocate open access versionFindings
  • Markus Rokicki, Sergej Zerr, and Stefan Siersdorfer. 2015. Groupsourcing: Team competition designs for crowdsourcing. In Proceedings of the 24th international conference on world wide web. 906–915.
    Google ScholarLocate open access versionFindings
  • Charles D Scales Jr, Tannaz Moin, Arlene Fink, Sandra H Berry, Nasim Afsar-Manesh, Carol M Mangione, and B Price Kerfoot. 2016. A randomized, controlled trial of team-based competition to increase learner participation in qualityimprovement education. International Journal for Quality in Health Care 28, 2 (2016), 227–232.
    Google ScholarLocate open access versionFindings
  • Uri Shalit, Fredrik D Johansson, and David Sontag. 2017. Estimating individual treatment effect: generalization bounds and algorithms. In Proceedings of the 34th International Conference on Machine Learning-Volume 70. 3076–3085.
    Google ScholarLocate open access versionFindings
  • VS Subrahmanian and Srijan Kumar. 2017. Predicting human behavior: The next frontiers. Science 355, 6324 (2017), 489–489.
    Google ScholarLocate open access versionFindings
  • Adith Swaminathan and Thorsten Joachims. 2015. The self-normalized estimator for counterfactual learning. In Advances in Neural Information Processing Systems. 3231–3239.
    Google ScholarLocate open access versionFindings
  • Robert Tibshirani. 1996. Regression shrinkage and selection via the Lasso. Journal of the Royal Statistical Society. Series B (Methodological) 58, 1 (1996), 267–288.
    Google ScholarLocate open access versionFindings
  • Milan Vojnović. 2015. Contest theory: Incentive mechanisms and ranking methods. Cambridge University Press.
    Google ScholarFindings
  • Jürgen Wegge, Carla Roth, Barbara Neubach, Klaus-Helmut Schmidt, and Ruth Kanfer. 2008. Age and gender diversity as determinants of performance and health in a public organization: the role of task complexity and group size. Journal of Applied Psychology 93, 6 (2008), 1301.
    Google ScholarLocate open access versionFindings
  • Lingyu Zhang, Tianshu Song, Yongxin Tong, Zimu Zhou, Dan Li, Wei Ai, Lulu Zhang, Guobin Wu, Yan Liu, and Jieping Ye. 2019. Recommendation-based team formation for on-demand taxi-calling platform. In Proceedings of the 28th ACM International Conference on Information and Knowledge Management. 59–68.
    Google ScholarLocate open access versionFindings
  • Zhe Zhang and Beibei Li. 2017. A quasi-experimental estimate of the impact of p2p transportation platforms on urban consumer patterns. In Proceedings of the 23rd ACM SIGKDD. ACM, 1683–1692.
    Google ScholarLocate open access versionFindings
Full Text
Your rating :
0

 

Tags
Comments