Privacy, Prediction, and Allocation

Ben Jacobsen , Nitin Kohli

Authors on Pith no claims yet

Pith reviewed 2026-05-10 08:38 UTC · model grok-4.3

classification 💻 cs.CR

keywords allocationprivacytargetingunit-levelworksindividual-levelprivatesettings

0 comments

The pith

Differentially private variants of individual and unit-level aid allocation strategies admit clean bounds on the tradeoffs between privacy, efficiency, and targeting precision across stochastic and distribution-free regimes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The authors combine techniques from differentially private optimization with standard economic models of how to distribute limited resources such as aid or interventions. They consider two main approaches: one that tries to target specific individuals based on predictions, and another that allocates at the level of groups or units without needing individual data. By adding privacy constraints, the methods protect sensitive information but can reduce how accurately they identify the people who would benefit most. The paper supplies mathematical bounds that show how much privacy costs in terms of lost efficiency and reduced targeting precision, under different assumptions about what data is available and whether the underlying distribution is known.

Core claim

Through this analysis, we provide clean, interpretable bounds characterizing the tradeoffs between privacy, efficiency, and targeting precision in allocation.

Load-bearing premise

That existing private optimization techniques can be directly synthesized with non-private economic allocation models to yield meaningful, interpretable bounds without additional unstated assumptions on data or distributions.

read the original abstract

Algorithmic predictions are increasingly used to inform the allocation of scarce resources. The promise of these methods is that, through machine learning, they can better identify the people who would benefit most from interventions. Recently, however, several works have called this assumption into question by demonstrating the existence of settings where simple, unit-level allocation strategies can meet or even exceed the performance of those based on individual-level targeting. Separately, other works have objected to individual-level targeting on privacy grounds, leading to an unusual situation where a single solution, unit-level targeting, is recommended for reasons of both privacy and utility. Motivated by the desire to fully understand the interplay of privacy and targeting levels, we initiate the study of aid allocation systems that satisfy differential privacy, synthesizing existing works on private optimization with the economic models of aid allocation used in the non-private literature. To this end, we investigate private variants of both individual and unit-level allocation strategies in both stochastic and distribution-free settings under a range of constraints on data availability. Through this analysis, we provide clean, interpretable bounds characterizing the tradeoffs between privacy, efficiency, and targeting precision in allocation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Circularity Check

0 steps flagged

No significant circularity; synthesis of independent prior works

full rationale

The paper's abstract and described approach synthesize existing literature on differential privacy in optimization with non-private economic allocation models to derive bounds on privacy-efficiency tradeoffs. No equations or steps in the provided text reduce a claimed prediction or bound to a fitted parameter, self-definition, or unverified self-citation chain. Central claims appear to rest on combining externally established techniques rather than re-deriving them from the paper's own inputs. Minor self-citations (if present in the full text) are not load-bearing for the core synthesis, consistent with normal scholarly practice. This yields a low circularity score with no specific reduction exhibited.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only abstract available; no explicit free parameters, axioms, or invented entities are stated. The synthesis assumes standard differential privacy definitions and existing economic allocation models can be combined without further specification.

pith-pipeline@v0.9.0 · 5485 in / 940 out tokens · 23473 ms · 2026-05-10T08:38:19.157063+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

73 extracted references · 27 canonical work pages · 1 internal anchor

[1]

The challenges of targeting social pro- tection programs, March 2022

Abdul Latif Jameel Poverty Action Lab (J-PAL). The challenges of targeting social pro- tection programs, March 2022. Blog post. URL:https://www.povertyactionlab.org/ blog/3-10-22/challenges-targeting-social-protection-programs

2022
[2]

Gaussian inequalities.Random Fields and Geom- etry, pages 49–64, 2007

Robert J Adler and Jonathan E Taylor. Gaussian inequalities.Random Fields and Geom- etry, pages 49–64, 2007

2007
[3]

PhD thesis, University of California, Berkeley, 2024

Emily Aiken.Targeting Social Protection Programs with Machine Learning and Digital Data. PhD thesis, University of California, Berkeley, 2024

2024
[4]

Machine learning and phone data can improve targeting of humanitarian aid.Nature, 603(7903):864–870, 2022

Emily Aiken, Suzanne Bellue, Dean Karlan, Chris Udry, and Joshua E Blumenstock. Machine learning and phone data can improve targeting of humanitarian aid.Nature, 603(7903):864–870, 2022

2022
[5]

Program targeting with machine learning and mobile phone data: Evidence from an anti-poverty intervention in afghanistan.Journal of Development Economics, 161:103016, 2023

Emily L Aiken, Guadalupe Bedoya, Joshua E Blumenstock, and Aidan Coville. Program targeting with machine learning and mobile phone data: Evidence from an anti-poverty intervention in afghanistan.Journal of Development Economics, 161:103016, 2023

2023
[6]

Best arm identification in multi-armed bandits

Jean-Yves Audibert and S´ ebastien Bubeck. Best arm identification in multi-armed bandits. InCOLT-23th Conference on learning theory-2010, pages 13–p, 2010

2010
[7]

The price of selection in differential privacy

Mitali Bafna and Jonathan Ullman. The price of selection in differential privacy. InCon- ference on Learning Theory, pages 151–168. PMLR, 2017

2017
[8]

Proper losses, moduli of convexity, and surrogate regret bounds

Han Bao. Proper losses, moduli of convexity, and surrogate regret bounds. InThe Thirty Sixth Annual Conference on Learning Theory, pages 525–547. PMLR, 2023

2023
[9]

In- terventions over predictions: Reframing the ethical debate for actuarial risk assessment

Chelsea Barabas, Madars Virza, Karthik Dinakar, Joichi Ito, and Jonathan Zittrain. In- terventions over predictions: Reframing the ethical debate for actuarial risk assessment. In Conference on fairness, accountability and transparency, pages 62–76. PMLR, 2018

2018
[10]

Behavior revealed in mobile phone usage predicts credit repayment.The World Bank Economic Review, 34(3):618–634, 2020

Daniel Bj¨ orkegren and Darrell Grissen. Behavior revealed in mobile phone usage predicts credit repayment.The World Bank Economic Review, 34(3):618–634, 2020

2020
[11]

Big data privacy in emerging market fintech and financial services: A research agenda.arXiv preprint arXiv:2310.04970, 2023

Joshua E Blumenstock and Nitin Kohli. Big data privacy in emerging market fintech and financial services: A research agenda.arXiv preprint arXiv:2310.04970, 2023

work page arXiv 2023
[12]

On track for success: The use of early warning indicator and intervention systems to build a grad nation

Mary Bruce, John M Bridgeland, Joanna Hornig Fox, and Robert Balfanz. On track for success: The use of early warning indicator and intervention systems to build a grad nation. Civic Enterprises, 2011

2011
[13]

Stackelberg games for adversarial prediction prob- lems

Michael Br¨ uckner and Tobias Scheffer. Stackelberg games for adversarial prediction prob- lems. InProceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 547–555, 2011

2011
[14]

Loss functions for binary class probability estimation and classification: Structure and applications.Working draft, November, 3:13, 2005

Andreas Buja, Werner Stuetzle, and Yi Shen. Loss functions for binary class probability estimation and classification: Structure and applications.Working draft, November, 3:13, 2005

2005
[15]

Concentrated differential privacy: Simplifications, exten- sions, and lower bounds

Mark Bun and Thomas Steinke. Concentrated differential privacy: Simplifications, exten- sions, and lower bounds. InTheory of cryptography conference, pages 635–658. Springer, 2016. 17

2016
[16]

Good allocations from bad estimates.arXiv preprint arXiv:2601.05597, 2026

S´ ılvia Casacuberta and Moritz Hardt. Good allocations from bad estimates.arXiv preprint arXiv:2601.05597, 2026

work page arXiv 2026
[17]

Noisy Dual Mirror Descent: A Near Optimal Algorithm for Jointly-DP Convex Resource Allocation

Du Chen and Geoffrey A Chua. Noisy Dual Mirror Descent: A Near Optimal Algorithm for Jointly-DP Convex Resource Allocation
[18]

B. R. Data. How Biometric Devices Are Putting Afghans in Danger. URL:https:// interaktiv.br.de/biometrie-afghanistan/en/
[19]

Differentially private matchings.arXiv preprint arXiv:2501.00926, 2025

Michael Dinitz, George Z Li, Quanquan C Liu, and Felix Zhou. Differentially private matchings.arXiv preprint arXiv:2501.00926, 2025

work page arXiv 2025
[20]

Calibrating noise to sensitivity in private data analysis

Cynthia Dwork, Frank McSherry, Kobbi Nissim, and Adam Smith. Calibrating noise to sensitivity in private data analysis. InTheory of cryptography conference, pages 265–284. Springer, 2006

2006
[21]

Private stochastic convex optimization: optimal rates in linear time

Vitaly Feldman, Tomer Koren, and Kunal Talwar. Private stochastic convex optimization: optimal rates in linear time. InProceedings of the 52nd Annual ACM SIGACT Symposium on Theory of Computing, pages 439–449, 2020

2020
[22]

Empiri- cally understanding the value of prediction in allocation, 2026

Unai Fischer-Abaigar, Emily Aiken, Christoph Kern, and Juan Carlos Perdomo. Empiri- cally understanding the value of prediction in allocation, 2026. URL:https://arxiv.org/ abs/2602.08786,arXiv:2602.08786

work page internal anchor Pith review arXiv 2026
[23]

The Value of Prediction in Identifying the Worst-Off, January 2025

Unai Fischer-Abaigar, Christoph Kern, and Juan Carlos Perdomo. The Value of Prediction in Identifying the Worst-Off, January 2025. arXiv:2501.19334 [cs]. URL:http://arxiv. org/abs/2501.19334,doi:10.48550/arXiv.2501.19334

work page doi:10.48550/arxiv.2501.19334 2025
[24]

A joint exponential mechanism for differentially private top-k

Jennifer Gillenwater, Matthew Joseph, Andres Munoz, and Monica Ribero Diaz. A joint exponential mechanism for differentially private top-k. InInternational Conference on Machine Learning, pages 7570–7582. PMLR, 2022

2022
[25]

Performative Power, November 2022

Moritz Hardt, Meena Jagadeesan, and Celestine Mendler-D¨ unner. Performative Power, November 2022. arXiv:2203.17232 [cs]. URL:http://arxiv.org/abs/2203.17232

work page arXiv 2022
[26]

Algorithmic Collective Action in Machine Learning

Moritz Hardt, Eric Mazumdar, Celestine Mendler-D¨ unner, and Tijana Zrnic. Algorithmic Collective Action in Machine Learning
[27]

Performative Prediction: Past and Future

Moritz Hardt and Celestine Mendler-D¨ unner. Performative Prediction: Past and Future, October 2023. arXiv:2310.16608 [cs]. URL:http://arxiv.org/abs/2310.16608

work page arXiv 2023
[28]

Almost Tight Error Bounds on Differentially Private Continual Counting

Monika Henzinger, Jalaj Upadhyay, and Sarvagya Upadhyay. Almost Tight Error Bounds on Differentially Private Continual Counting. 2023.doi:10.1137/1.9781611977554. ch183

work page doi:10.1137/1.9781611977554 2023
[29]

Private Matchings and Allocations

Justin Hsu, Zhiyi Huang, Aaron Roth, Tim Roughgarden, and Zhiwei Steven Wu. Private Matchings and Allocations.SIAM Journal on Computing, 45(6):1953–1984, January 2016. URL:http://epubs.siam.org/doi/10.1137/15100271X,doi:10.1137/15100271X

work page doi:10.1137/15100271x 1953
[30]

Jointly private convex programming

Justin Hsu, Zhiyi Huang, Aaron Roth, and Zhiwei Steven Wu. Jointly private convex programming. InProceedings of the twenty-seventh annual ACM-SIAM symposium on Discrete algorithms, pages 580–599. SIAM, 2016

2016
[31]

A Frugal Model for Accurate Early Student Failure Prediction

Gagaoua Ikram, Armelle Brun, and Anne Boyer. A Frugal Model for Accurate Early Student Failure Prediction. 18
[32]

Allocation Multiplicity: Evaluating the Promises of the Rashomon Set, March 2025

Shomik Jain, Margaret Wang, Kathleen Creel, and Ashia Wilson. Allocation Multiplicity: Evaluating the Promises of the Rashomon Set, March 2025. arXiv:2503.16621 [cs]. URL: http://arxiv.org/abs/2503.16621,doi:10.48550/arXiv.2503.16621

work page doi:10.48550/arxiv.2503.16621 2025
[33]

What is the Bureaucratic Counterfactual? Cat- egorical versus Algorithmic Prioritization in U.S

Rebecca Ann Johnson and Simone Zhang. What is the Bureaucratic Counterfactual? Cat- egorical versus Algorithmic Prioritization in U.S. Social Policy. In2022 ACM Confer- ence on Fairness, Accountability, and Transparency, pages 1671–1682, Seoul Republic of Korea, June 2022. ACM. URL:https://dl.acm.org/doi/10.1145/3531146.3533223, doi:10.1145/3531146.3533223

work page doi:10.1145/3531146.3533223 2022
[34]

Optimal best arm identification under differential privacy

Marc Jourdan and Achraf Azize. Optimal best arm identification under differential privacy. arXiv preprint arXiv:2510.17348, 2025

work page arXiv 2025
[35]

Expanding perspectives on data privacy: Insights from rural togo.Proceedings of the ACM on Human-Computer Interaction, 9(2):1–29, 2025

Zoe Kahn, Meyebinesso Farida Carelle Pere, Emily Aiken, Nitin Kohli, and Joshua E Blu- menstock. Expanding perspectives on data privacy: Insights from rural togo.Proceedings of the ACM on Human-Computer Interaction, 9(2):1–29, 2025

2025
[36]

Adaptive treatment assignment in experiments for policy choice.Econometrica, 89(1):113–132, 2021

Maximilian Kasy and Anja Sautmann. Adaptive treatment assignment in experiments for policy choice.Econometrica, 89(1):113–132, 2021

2021
[37]

Pai, Ryan Rogers, Aaron Roth, and Jonathan Ullman

Michael Kearns, Mallesh M. Pai, Ryan Rogers, Aaron Roth, and Jonathan Ullman. Robust mediators in large games, 2015. URL:https://arxiv.org/abs/1512.02698,arXiv:1512. 02698

work page arXiv 2015
[38]

Kim and Moritz Hardt

Michael P. Kim and Moritz Hardt. Is Your Model Predicting the Past? InEquity and Access in Algorithms, Mechanisms, and Optimization, pages 1–8, Boston MA USA, October
[39]

In Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization

ACM. URL:https://dl.acm.org/doi/10.1145/3617694.3623225,doi:10.1145/ 3617694.3623225

work page doi:10.1145/3617694.3623225
[40]

Human decisions and machine predictions.The Quarterly Journal of Economics, 133(1):237–293, 2018

Jon Kleinberg, Himabindu Lakkaraju, Jure Leskovec, Jens Ludwig, and Sendhil Mul- lainathan. Human decisions and machine predictions.The Quarterly Journal of Economics, 133(1):237–293, 2018

2018
[41]

Privacy guarantees for personal mobility data in humanitarian response.Scientific Reports, 14(1):28565, 2024

Nitin Kohli, Emily Aiken, and Joshua E Blumenstock. Privacy guarantees for personal mobility data in humanitarian response.Scientific Reports, 14(1):28565, 2024

2024
[42]

Enabling Humanitarian Applications with Targeted Differential Privacy, August 2024

Nitin Kohli and Joshua Blumenstock. Enabling Humanitarian Applications with Targeted Differential Privacy, August 2024. arXiv:2408.13424 [cs]. URL:http://arxiv.org/abs/ 2408.13424

work page arXiv 2024
[43]

T., and Wang, Y

Alex Kulesza, Ananda Theertha Suresh, and Yuyan Wang. Mean estimation in the add- remove model of differential privacy.arXiv preprint arXiv:2312.06658, 2023

work page arXiv 2023
[44]

Metalearners for estimat- ing heterogeneous treatment effects using machine learning.Proceedings of the national academy of sciences, 116(10):4156–4165, 2019

S¨ oren R K¨ unzel, Jasjeet S Sekhon, Peter J Bickel, and Bin Yu. Metalearners for estimat- ing heterogeneous treatment effects using machine learning.Proceedings of the national academy of sciences, 116(10):4156–4165, 2019

2019
[45]

Reimagining the machine learning life cycle to improve educational outcomes of students.Proceedings of the National Academy of Sciences, 120(9):e2204781120, 2023

Lydia T Liu, Serena Wang, Tolani Britton, and Rediet Abebe. Reimagining the machine learning life cycle to improve educational outcomes of students.Proceedings of the National Academy of Sciences, 120(9):e2204781120, 2023

2023
[46]

Nelson, Vicki Yang, Kathryn Edin, and Matthew J

Ian Lundberg, Rachel Brown-Weinstock, Susan Clampet-Lundquist, Sarah Pachman, Tim- othy J. Nelson, Vicki Yang, Kathryn Edin, and Matthew J. Salganik. The origins of unpredictability in life trajectory prediction tasks, October 2023. arXiv:2310.12871 [stat]. URL:http://arxiv.org/abs/2310.12871. 19

work page arXiv 2023
[47]

A Low-Cost Privacy-Preserving Digital Wallet for Humanitarian Aid Distribution, October 2024

Eva Luvison, Sylvain Chatel, Justinas Sukaitis, Vincent Graf Narbel, Carmela Troncoso, and Wouter Lueks. A Low-Cost Privacy-Preserving Digital Wallet for Humanitarian Aid Distribution, October 2024. arXiv:2410.15942 [cs]. URL:http://arxiv.org/abs/2410. 15942

work page arXiv 2024
[48]

On the limiting distributions of estimates based on samples from finite universes.The Annals of Mathematical Statistics, pages 535–545, 1948

William G Madow. On the limiting distributions of estimates based on samples from finite universes.The Annals of Mathematical Statistics, pages 535–545, 1948

1948
[49]

Categorizing Variants of Goodhart’s Law, February

David Manheim and Scott Garrabrant. Categorizing Variants of Goodhart’s Law, February
[50]

Categorizing Variants of Goodhart's Law

arXiv:1803.04585 [cs]. URL:http://arxiv.org/abs/1803.04585

work page Pith review arXiv
[51]

Who pays the rent? implications of spatial inequality for prediction-based allocation policies

Tasfia Mashiat, Patrick J Fowler, and Sanmay Das. Who pays the rent? implications of spatial inequality for prediction-based allocation policies. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, volume 8, pages 1686–1697, 2025

2025
[52]

Privacy integrated queries: an extensible platform for privacy- preserving data analysis

Frank D McSherry. Privacy integrated queries: an extensible platform for privacy- preserving data analysis. InProceedings of the 2009 ACM SIGMOD International Confer- ence on Management of data, pages 19–30, 2009

2009
[53]

MIT press, 2004

Herv´ e Moulin.Fair division and collective welfare. MIT press, 2004

2004
[54]

On the relationship between binary clas- sification, bipartite ranking, and binary class probability estimation.Advances in neural information processing systems, 26, 2013

Harikrishna Narasimhan and Shivani Agarwal. On the relationship between binary clas- sification, bipartite ranking, and binary class probability estimation.Advances in neural information processing systems, 26, 2013

2013
[55]

Differentially private estimation of heterogeneous causal effects

Fengshi Niu, Harsha Nori, Brian Quistorff, Rich Caruana, Donald Ngwe, and Aadharsh Kannan. Differentially private estimation of heterogeneous causal effects. InConference on Causal Learning and Reasoning, pages 618–633. PMLR, 2022

2022
[56]

Perdomo, Tolani Britton, Moritz Hardt, and Rediet Abebe

Juan C. Perdomo, Tolani Britton, Moritz Hardt, and Rediet Abebe. Difficult Lessons on Social Prediction from Wisconsin Public Schools, September 2023. arXiv:2304.06205 [cs]. URL:http://arxiv.org/abs/2304.06205,doi:10.48550/arXiv.2304.06205

work page doi:10.48550/arxiv.2304.06205 2023
[57]

The Relative Value of Prediction in Algorithmic Decision Making

Juan Carlos Perdomo. The Relative Value of Prediction in Algorithmic Decision Making
[58]

Correlated noise mechanisms for differentially private learning, 2025

Krishna Pillutla, Jalaj Upadhyay, Christopher A Choquette-Choo, Krishnamurthy Dvi- jotham, Arun Ganesh, Monika Henzinger, Jonathan Katz, Ryan McKenna, H Brendan McMahan, Keith Rush, et al. Correlated noise mechanisms for differentially private learn- ing.arXiv preprint arXiv:2506.08201, 2025

work page arXiv 2025
[59]

When do citizens resist the use of ai algorithms in public policy? theory and evidence

Shir Raviv. When do citizens resist the use of ai algorithms in public policy? theory and evidence. Available at SSRN:https://ssrn.com/abstract=4328400orhttp:// dx.doi.org/10.2139/ssrn.4328400, January 2023. URL:https://ssrn.com/abstract= 4328400

work page doi:10.2139/ssrn.4328400 2023
[60]

Salganik, Ian Lundberg, Alexander T

Matthew J. Salganik, Ian Lundberg, Alexander T. Kindel, Caitlin E. Ahearn, Khaled Al-Ghoneim, Abdullah Almaatouq, Drew M. Altschul, Jennie E. Brand, Nicole Bohme Carnegie, Ryan James Compton, Debanjan Datta, Thomas Davidson, Anna Filippova, Connor Gilroy, Brian J. Goode, Eaman Jahani, Ridhi Kashyap, Antje Kirchner, Stephen McKay, Allison C. Morgan, Alex P...

work page doi:10.1073/pnas.1915006117 2020
[61]

Probability inequalities for the sum in sampling without replacement

Robert J Serfling. Probability inequalities for the sum in sampling without replacement. The Annals of Statistics, pages 39–48, 1974

1974
[62]

Allocation Requires Prediction Only if In- equality Is Low, June 2024

Ali Shirali, Rediet Abebe, and Moritz Hardt. Allocation Requires Prediction Only if In- equality Is Low, June 2024. arXiv:2406.13882 [cs, econ]. URL:http://arxiv.org/abs/ 2406.13882

work page arXiv 2024
[63]

2003.The computation factory: de Prony’s project for making tables in the 1790s

Dan Sperber, David Premack, Ann J. Premack, Dan Sperber, and Ann J. Premack, editors. The looping effects of human kinds. Symposia of the Fyssen foundation. Clarendon, Oxford, 1995.doi:10.1093/acprof:oso/9780198524021.001.0001

work page doi:10.1093/acprof:oso/9780198524021.001.0001 1995
[64]

Tight lower bounds for differentially private selec- tion

Thomas Steinke and Jonathan Ullman. Tight lower bounds for differentially private selec- tion. In2017 IEEE 58th Annual Symposium on Foundations of Computer Science (FOCS), pages 552–563. IEEE, 2017

2017
[65]

Estimation and inference of heterogeneous treatment effects using random forests.Journal of the American Statistical Association, 113(523):1228–1242, 2018

Stefan Wager and Susan Athey. Estimation and inference of heterogeneous treatment effects using random forests.Journal of the American Statistical Association, 113(523):1228–1242, 2018

2018
[66]

Against predictive optimization: On the legitimacy of decision-making algorithms that optimize predictive accuracy.ACM Journal on Responsible Computing, 1(1):1–45, 2024

Angelina Wang, Sayash Kapoor, Solon Barocas, and Arvind Narayanan. Against predictive optimization: On the legitimacy of decision-making algorithms that optimize predictive accuracy.ACM Journal on Responsible Computing, 1(1):1–45, 2024

2024
[67]

Not Yet Another Digital ID: Privacy-Preserving Humanitarian Aid Distribution

Boya Wang, Wouter Lueks, Justinas Sukaitis, Vincent Graf Narbel, and Carmela Troncoso. Not Yet Another Digital ID: Privacy-Preserving Humanitarian Aid Distribution. In2023 IEEE Symposium on Security and Privacy (SP), pages 645–663, May 2023. ISSN: 2375-

2023
[68]

URL:https://ieeexplore.ieee.org/document/10179306/?arnumber=10179306, doi:10.1109/SP46215.2023.10179306

work page doi:10.1109/sp46215.2023.10179306 2023
[69]

Learning treatment effects while treating those in need

Bryan Wilder and Pim Welle. Learning treatment effects while treating those in need. In Proceedings of the 26th ACM Conference on Economics and Computation, pages 448–473, 2025

2025
[70]

Revisiting targeting in social assistance: A new look at old challenges

World Bank. Revisiting targeting in social assistance: A new look at old challenges. Policy research report, World Bank, 2022. URL:https://documents1.worldbank.org/curated/ en/099300006162240907/pdf/P1648760df4c480270bd0b02288943f413e.pdf. 21

work page arXiv 2022
[71]

ILA within units

Andrew Chi-Chin Yao. Probabilistic computations: Toward a unified measure of complex- ity. In18th Annual Symposium on Foundations of Computer Science (sfcs 1977), pages 222–227. IEEE, 1977. Appendices A Differences between Our Model and that of Shirali et al. [60] In their definition of ULA, the authors of [60] make the additional assumption that unit-lev...

1977
[72]

Letk ′ =k−nλ≥0 denote the actual number of individuals we treat

To start, we have that: σ∗ := max j σ2 j ≤ M2 2ψn2 + M(N−n/M) 4nN = M2 2ψn2 + P−n 4nN (134) and by Lemma 18, we know that: σ2 := 1 M MX j=1 σ2 j ≤ M2 2ψn2 + (P−n)(¯ρ M(1−¯ρM)−3¯ρ2 M G2) nN (135) 34 Finally, Jensen’s inequality then gives us that: ¯σ:= 1 M MX j=1 σj ≤σ(136) Applying Lemma 6, our combined error from privacy and statistical estimation become...
[73]

For ease of exposition, we sidestep this point by assuming that the population of each unit is lower-bounded byN, and that this fact is publicly known

Secondly, because the sensitivity ofρ j scales with|U j|−1, which is now private, getting the additive noise scale as tight as possible requires some subtlety [42]. For ease of exposition, we sidestep this point by assuming that the population of each unit is lower-bounded byN, and that this fact is publicly known. With those adjustments made, our regret ...