ABSTRACT
Equipping current decision-making tools with notions of fairness, equitability, or other ethically motivated outcomes, is one of the top priorities in recent research efforts in machine learning, AI, and optimization. In this paper, we investigate how to allocate limited resources to locally interacting communities in a way to maximize a pertinent notion of equitability. In particular, we look at the dynamic setting where the allocation is repeated across multiple periods (e.g., yearly), the local communities evolve in the meantime (driven by the provided allocation), and the allocations are modulated by feedback coming from the communities themselves. We employ recent mathematical tools stemming from data-driven feedback online optimization, by which communities can learn their (possibly unknown) evolution, satisfaction, as well as they can share information with the deciding bodies. We design dynamic policies that converge to an allocation that maximize equitability in the long term. We further demonstrate our model and methodology with realistic examples of healthcare and education subsidies design in Sub-Saharian countries. One of the key empirical takeaways from our setting is that long-term equitability is fragile, in the sense that it can be easily lost when deciding bodies weigh in other factors (e.g., equality in allocation) in the allocation strategy. Moreover, a naive compromise, while not providing significant advantage to the communities, can promote inequality in social outcomes.
Supplemental Material
- Arrow, K. The theory of discrimination. Working Papers 403, Princeton University, Department of Economics, Industrial Relations Section., 1971.Google Scholar
- Barocas, S., Hardt, M., and Narayanan, A. Fairness and Machine Learning. fairmlbook.org, 2019. http://www.fairmlbook.org.Google Scholar
- Barro-Lee. The Barro-Lee Data Set. Online (2021). http://www.barrolee.com.Google Scholar
- Ben-Porat, O., Sandomirskiy, F., and Tennenholtz, M. Protecting the protected group: Circumventing harmful fairness. Proceedings of the AAAI Conference on Artificial Intelligence 35, 6 (May 2021), 5176--5184.Google ScholarCross Ref
- Bianchin, G., Vaquero, M., Cortes, J., and Dall'Anese, E. Data-driven synthesis of optimization-based controllers for regulation of unknown linear systems. arXiv: 2103.16067 (2021).Google Scholar
- Bianchin, G., Vaquero, M., Cortes, J., and Dall'Anese, E. Online stochastic optimization for unknown linear systems: Data-driven synthesis and controller analysis. arXiv: 2108.13040 (2021).Google Scholar
- Calmon, F., Wei, D., Vinzamuri, B., Natesan Ramamurthy, K., and Varshney, K. R. Optimized pre-processing for discrimination prevention. In Advances in Neural Information Processing Systems (2017), vol. 30.Google Scholar
- Chen, M., and Palmer, A. J. Assessing equity in benefit distribution of government health subsidy in 2012 across East China: benefit incidence analysis. International Journal for Equity in Health 15 (2016).Google ScholarCross Ref
- Chi, J., Shen, J., Dai, X., Zhang, W., Tian, Y., and Zhao, H. Towards return parity in Markov Decision Processes. Online (2021). arXiv preprint, arXiv:2111.10476.Google Scholar
- Coulson, J., Lygeros, J., and Doerfler, F. Data-enabled predictive control: In the shallows of the DeePC. In 2019 18th European Control Conference (ECC) (2019), pp. 307--312.Google ScholarCross Ref
- Creager, E., Madras, D., Pitassi, T., and Zemel, R. Causal modeling for fairness in dynamical systems. In Proceedings of the 37th International Conference on Machine Learning (2020), vol. 119 of Proceedings of Machine Learning Research, pp. 2185--2195.Google ScholarDigital Library
- D'Amour, A., Srinivasan, H., Atwood, J., Baljekar, P., Sculley, D., and Halpern, Y. Fairness is not static: Deeper understanding of long term fairness via simulation studies. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (2020), FAT* '20, p. 525--534.Google ScholarDigital Library
- Diana, E., Gill, W., Kearns, M., Kenthapadi, K., and Roth, A. Minimax group fairness: Algorithms and experiments. In Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society (2021), p. 66--76.Google ScholarDigital Library
- Duflo, E., Dupas, P., and Kremer, M. The Impact of Free Secondary Education: Experimental Evidence from Ghana. Working Paper 28937, National Bureau of Economic Research, June 2021.Google ScholarCross Ref
- Ge, Y., Liu, S., Gao, R., Xian, Y., Li, Y., Zhao, X., Pei, C., Sun, F., Ge, J., Ou, W., and Zhang, Y. Towards long-term fairness in recommendation. In Proceedings of the 14th ACM International Conference on Web Search and Data Mining (2021), p. 445--453.Google ScholarDigital Library
- Gupta, S., Verhoeven, M., Gillingham, R., Schiller, C., Mansoor, A., and Cordoban, J. P. Equity and Efficiency in the Reform of Price Subsidies. International Monetary Fund, 2000.Google Scholar
- Hardt, M., Price, E., Price, E., and Srebro, N. Equality of opportunity in supervised learning. In Advances in Neural Information Processing Systems (2016), vol. 29.Google Scholar
- Heidari, H., Ferrari, C., Gummadi, K. P., and Krause, A. Fairness behind a veil of ignorance: A welfare analysis for automated decision making. In Proceedings of the 32nd International Conference on Neural Information Processing Systems (2018), NIPS'18, p. 1273--1283.Google Scholar
- Heidari, H., and Kleinberg, J. Allocating opportunities in a dynamic model of intergenerational mobility. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (2021), FAccT '21, p. 15--25.Google ScholarDigital Library
- Heidari, H., Loi, M., Gummadi, K. P., and Krause, A. A moral framework for understanding fair ML through economic models of equality of opportunity. In Proceedings of the Conference on Fairness, Accountability, and Transparency (2019), FAT* '19, p. 181--190.Google ScholarDigital Library
- Heidari, H., Nanda, V., and Gummadi, K. On the long-term impact of algorithmic decision policies: Effort unfairness and feature segregation through social learning. In Proceedings of the 36th International Conference on Machine Learning (2019), K. Chaudhuri and R. Salakhutdinov, Eds., pp. 2692--2701.Google Scholar
- Hortay, O., and Rozner, B. P. Allocating renewable subsidies. Economic Analysis and Policy 64 (2019), 236--247.Google ScholarCross Ref
- Hossain, S., Mladenovic, A., and Shah, N. Designing fairly fair classifiers via economic fairness notions. In Proceedings of The Web Conference 2020 (2020), p. 1559--1569.Google ScholarDigital Library
- IEA. Energy subsidies: Tracking the impact of fossil-fuel subsidies. Online (2021). https://www.iea.org/topics/energy-subsidies.Google Scholar
- Kahneman, D., and Tversky, A. Prospect Theory: An Analysis of Decision under Risk. Econometrica 47, 2 (1979), 263--291.Google ScholarCross Ref
- Kallus, N., Mao, X., and Zhou, A. Assessing algorithmic fairness with unobserved protected class using data combination. Online (2019). arXiv preprint, arXiv:1906.00285.Google Scholar
- Kasy, M., and Abebe, R. Fairness, equality, and power in algorithmic decision-making. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (2021), FAccT '21, p. 576--586.Google ScholarDigital Library
- Komives, K., Foster, V., Halpern, J., and Wodon, Q. Water, Electricity, and the Poor: Who Benefits from Utility Subsidies? The World Bank, Washington, DC, 2005. with support from Roohi Abdullah.Google Scholar
- Le Blanc, D. Providing water to the urban poor in developing countries: the role of tariffs and subsidies. UN: Sustainable Development, Innovation Briefs 4 (2007).Google Scholar
- Lin, B., and Xu, M. Good subsidies or bad subsidies? evidence from low-carbon transition in china's metallurgical industry. Energy Economics 83 (2019), 52--60.Google ScholarCross Ref
- Liu, Y., Sandmann, F. G., Barnard, R. C., Pearson, C. A., Pastore, R., Pebody, R., Flasche, S., and Jit, M. Optimising health and economic impacts of COVID-19 vaccine prioritisation strategies in the WHO European Region: a mathematical modelling study. The Lancet, Regional Health Europe 12 (2022).Google Scholar
- Ljung, L. System Identification: Theory for the User. Prentice Hall, 1999. Second Edition.Google ScholarDigital Library
- Matrajt, L., Eaton, J., Leung, T., Dimitrov, D., Schiffer, J. T., Swan, D. A., and Janes, H. Optimizing vaccine allocation for COVID-19 vaccines shows the potential role of single-dose vaccination. Nature Communications 12, 1 (2021), 3449.Google ScholarCross Ref
- Morik, M., Singh, A., Hong, J., and Joachims, T. Controlling fairness and bias in dynamic learning-to-rank. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (2020), p. 429--438.Google ScholarDigital Library
- Mouzannar, H., Ohannessian, M. I., and Srebro, N. From fair decision making to social equality. In Proceedings of the Conference on Fairness, Accountability, and Transparency (2019), FAT* '19, p. 359--368.Google ScholarDigital Library
- Nesterov, Y. Introductory Lectures on Convex Optimization. Kluwer Academic Publisher, 2004.Google ScholarCross Ref
- Nocedal, J., and Wright, S. J. Numerical Optimization. Springer, 2006. Second Edition.Google Scholar
- Notarnicola, I., Simonetto, A., Farina, F., and Notarstefano, G. Distributed personalized gradient tracking with convex parametric models. IEEE Transactions on Automatic Control (in press) (2022).Google ScholarCross Ref
- Oketch, M. Financing higher education in sub-Saharan Africa: some reflections and implications for sustainable development. Higher Education 72 (2016).Google Scholar
- Ortiz-Ospina, E., and Roser, M. Financing healthcare. Our World in Data (2017). https://ourworldindata.org/financing-healthcare.Google Scholar
- Ortiz-Ospina, E., and Roser, M. Life expectancy vs. healthcare expenditure, 1991 to 2014. Our World in Data (2017). Sources: World Bank, World Development Indicators http://data.worldbank.org/data-catalog/world-development-indicators.Google Scholar
- Rawls, J. A theory of justice. Harvard University Press, 1971.Google ScholarCross Ref
- Roser, M. Human development index (HDI). Our World in Data (2014). https://ourworldindata.org/human-development-index.Google Scholar
- Roser, M., and Ortiz-Ospina, E. Financing education. Our World in Data (2016). https://ourworldindata.org/financing-education.Google Scholar
- Simonetto, A., Dall'Anese, E., Monteil, J., and Bernstein, A. Personalized optimization with user's feedback. Automatica 131 (2021), 109767.Google ScholarDigital Library
- Varian, H. R. Equity, envy, efficiency. MIT, 1973.Google Scholar
- Wen, M., Bastani, O., and Topcu, U. Algorithms for fairness in sequential decision making. In Proceedings of The 24th International Conference on Artificial Intelligence and Statistics (13--15 Apr 2021), vol. 130 of Proceedings of Machine Learning Research, PMLR, pp. 1144--1152.Google Scholar
- Yaghini, M., Krause, A., and Heidari, H. A human-in-the-loop framework to construct context-aware mathematical notions of outcome fairness. In Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society (2021), p. 1023--1033.Google ScholarDigital Library
- Zhang, D., and Wang, J. Recommendation fairness: From static to dynamic. Online (2021). https://arxiv.org/abs/2109.03150v3.Google Scholar
Index Terms
- Achievement and Fragility of Long-term Equitability
Recommendations
Equitability and Welfare Maximization for Allocating Indivisible Items
AAMAS '23: Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent SystemsWe study fair allocations of indivisible goods and chores in conjunction with system efficiency, measured by two social welfare functions, namely utilitarian and egalitarian welfare. To model preference, each agent is associated with a cardinal and ...
Fair Optimization -- Methodological Foundations of Fairness in Network Resource Allocation
COMPSACW '14: Proceedings of the 2014 IEEE 38th International Computer Software and Applications Conference WorkshopsNetwork resource allocation problems are concerned with the allocation of limited resources among competing entities so as to respect some fairness rules while looking for the overall efficiency. This paper presents the methodology of fair optimization ...
Equitable Allocations of Indivisible Chores
AAMAS '20: Proceedings of the 19th International Conference on Autonomous Agents and MultiAgent SystemsWe study fair allocation of indivisible chores (i.e., items with non-positive value) among agents with additive valuations. An allocation is deemed fair if it is (approximately) equitable, which means that the disutilities of the agents are (...
Comments