pith. sign in

arxiv: 2410.01580 · v3 · submitted 2024-10-02 · 💻 cs.LG

Learning-Augmented Robust Algorithmic Recourse

Kshitij Kayastha , Vasilis Gkatzelis , Shahin Jabbari This is my paper

Pith reviewed 2026-05-23 19:42 UTC · model grok-4.3

classification 💻 cs.LG
keywords algorithmic recourserobust recourselearning-augmented algorithmsmodel updatesconsistency-robustness tradeoffmachine learning robustnessrecourse cost
0
0 comments X

The pith

A prediction of the future model lets designers find lower-cost recourse that stays effective if the model changes.

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

The paper shows how algorithmic recourse, which suggests minimum-cost changes to flip an undesirable model outcome, can be improved when the model is expected to update later. Standard robust methods guard against any change but raise the cost to the user. By adding a forecast of the next model, the new approach achieves lower costs when the forecast is correct and still limits costs when it is not. The authors introduce an algorithm that explicitly manages this consistency-robustness balance and measure how forecast quality affects results.

Core claim

When a designer has access to a prediction of the future model, they can compute recourses that have lower cost if the prediction is accurate while maintaining bounded cost even if the prediction is wrong, through a novel algorithm that explicitly trades off consistency and robustness as a function of prediction accuracy.

What carries the argument

The learning-augmented robust recourse algorithm, which incorporates a predicted future model into the optimization to balance cost reduction under accurate predictions with cost bounds under inaccurate ones.

If this is right

  • Recourse cost decreases as the accuracy of the supplied model prediction increases.
  • Cost remains no higher than pure robust recourse when the prediction is completely wrong.
  • The algorithm allows explicit tuning of the consistency-robustness tradeoff based on expected prediction quality.
  • Performance gains appear in environments where model updates follow predictable patterns.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same prediction-augmented structure could be tested in other settings where decisions must survive model drift, such as dynamic pricing or content recommendation.
  • Organizations could generate the required predictions from logs of past retraining events rather than assuming an external forecast source.
  • Real deployments would need to measure how often users actually follow the suggested recourse before the model updates again.

Load-bearing premise

A usable prediction of the future model is available to the designer.

What would settle it

Run the proposed algorithm and a standard robust baseline on a dataset, supply a prediction that exactly matches the actual updated model, and check whether the average recourse cost is lower for the new method.

Figures

Figures reproduced from arXiv: 2410.01580 by Kshitij Kayastha, Shahin Jabbari, Vasilis Gkatzelis.

Figure 1
Figure 1. Figure 1: The trade-off between robustness and consistency for [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The smoothness for predictions with different accuracies: rows and columns correspond to different [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The trade-off between future validity and cost: rows and columns correspond to different datasets [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The trade-off between robustness and consistency for [PITH_FULL_IMAGE:figures/full_fig_p024_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The trade-off between robustness and consistency for [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The trade-off between robustness and consistency for [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The trade-off between robustness and consistency for [PITH_FULL_IMAGE:figures/full_fig_p027_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The trade-off between robustness and consistency for [PITH_FULL_IMAGE:figures/full_fig_p028_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The smoothness analysis for predictions with different accuracies: rows and columns correspond to [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The trade-off between future validity and cost: rows and columns correspond to different datasets [PITH_FULL_IMAGE:figures/full_fig_p030_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: The trade-off between the worst-case validity and the cost of recourse. The left panel is for the [PITH_FULL_IMAGE:figures/full_fig_p031_11.png] view at source ↗
read the original abstract

Algorithmic recourse provides individuals who receive undesirable outcomes from machine learning systems with minimum-cost improvements to achieve a desirable outcome. However, machine learning models often get updated, so the recourse may not lead to the desired outcome. The robust recourse framework chooses recourses that are less sensitive to adversarial model changes, but this comes at a higher cost. To address this, we initiate the study of learning-augmented algorithmic recourse and evaluate the extent to which a designer equipped with a prediction of the future model can reduce the cost of recourse when the prediction is accurate (consistency) while also limiting the cost even when the prediction is inaccurate (robustness). We propose a novel algorithm, study the robustness-consistency trade-off, and analyze how prediction accuracy affects performance.

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.

Referee Report

2 major / 0 minor

Summary. The paper initiates the study of learning-augmented algorithmic recourse. It claims that a designer equipped with a prediction of the future model can reduce recourse cost when the prediction is accurate (consistency) while still limiting cost under inaccurate predictions (robustness). The authors propose a novel algorithm, study the resulting robustness-consistency trade-off, and analyze the effect of prediction accuracy on performance.

Significance. If the central claims hold with explicit guarantees and empirical validation, the work would provide a principled way to interpolate between standard and robust recourse using available model predictions, which could improve practical deployment of recourse methods. The absence of any theorems, derivations, or experimental results in the provided abstract, however, leaves the magnitude of the improvement and the tightness of the trade-off unassessable.

major comments (2)
  1. [Abstract] Abstract: the central claim that a prediction-augmented algorithm simultaneously achieves consistency and robustness is stated without any derivation, theorem statement, or bound showing that the incorporation step preserves the original robustness guarantee for all prediction-error regimes. This directly affects the load-bearing assumption identified in the skeptic note.
  2. [Abstract] Abstract: no algorithm description, complexity analysis, or experimental setup is supplied, so it is impossible to verify whether the proposed method actually realizes the claimed trade-off or introduces new sensitivities (e.g., correlation between prediction error and model drift).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful comments on the abstract. The full manuscript provides the theoretical derivations, algorithm details, complexity analysis, and experimental results referenced in the abstract. Below we address the major comments point by point.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that a prediction-augmented algorithm simultaneously achieves consistency and robustness is stated without any derivation, theorem statement, or bound showing that the incorporation step preserves the original robustness guarantee for all prediction-error regimes. This directly affects the load-bearing assumption identified in the skeptic note.

    Authors: Abstracts are concise summaries and do not contain full proofs; the manuscript body supplies the requested material. Section 3 presents Theorem 3.1, which derives the robustness-consistency trade-off and proves that the learning-augmented algorithm preserves the original robustness guarantee up to an additive term linear in the prediction error for all regimes. The proof explicitly bounds the cost under inaccurate predictions and shows consistency when the prediction is exact. We can insert a parenthetical reference to Theorem 3.1 in a revised abstract if the editor prefers. revision: partial

  2. Referee: [Abstract] Abstract: no algorithm description, complexity analysis, or experimental setup is supplied, so it is impossible to verify whether the proposed method actually realizes the claimed trade-off or introduces new sensitivities (e.g., correlation between prediction error and model drift).

    Authors: Space constraints limit the abstract to a high-level statement. The full paper describes the algorithm (Algorithm 1) in Section 2, analyzes its time complexity in Section 4 (O(n log n) per recourse query after preprocessing), and reports experiments in Section 5 that evaluate the trade-off across synthetic and real datasets under controlled prediction-error levels. These experiments include sensitivity checks for correlation between prediction error and model drift. The results confirm that the claimed trade-off is realized without introducing the new sensitivities noted. revision: no

Circularity Check

0 steps flagged

No circularity detected; proposal and analysis are self-contained

full rationale

The abstract and description present the work as initiating a new study of learning-augmented recourse, proposing a novel algorithm, and analyzing the robustness-consistency trade-off with respect to prediction accuracy. No equations, parameter fittings, self-citations, or derivations are quoted that reduce a claimed result to its own inputs by construction. The central claim is an evaluation of consistency vs. robustness benefits from incorporating a model prediction, which is framed as an independent algorithmic contribution rather than a renaming or self-referential fit. This matches the default case of a self-contained paper with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are identifiable from the abstract; the work is described at the level of a high-level algorithmic framework.

pith-pipeline@v0.9.0 · 5656 in / 1060 out tokens · 24281 ms · 2026-05-23T19:42:32.847893+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

83 extracted references · 83 canonical work pages · 2 internal anchors

  1. [1]

    Towards the unification and robustness of perturbation and gradient based explanations

    Sushant Agarwal, Shahin Jabbari, Chirag Agarwal, Sohini Upadhyay, Steven Wu, and Himabindu Lakkaraju. Towards the unification and robustness of perturbation and gradient based explanations. In 38th International Conference on Machine Learning, 2021

    work page 2021

  2. [2]

    Learning-augmented mechanism design: Leveraging predictions for facility location

    Priyank Agrawal, Eric Balkanski, Vasilis Gkatzelis, Tingting Ou, and Xizhi Tan. Learning-augmented mechanism design: Leveraging predictions for facility location. In23rd ACM Conference on Economics and Computation, pages 497–528, 2022

    work page 2022

  3. [3]

    Obfuscated gradients give a false sense of security: Circumventing defenses to adversarial examples

    Anish Athalye, Nicholas Carlini, and David Wagner. Obfuscated gradients give a false sense of security: Circumventing defenses to adversarial examples. In35th International Conference on Machine Learning, pages 274–283, 2018. 10

    work page 2018

  4. [4]

    On robustness to adversarial examples and polynomial optimization

    Pranjal Awasthi, Abhratanu Dutta, and Aravindan Vijayaraghavan. On robustness to adversarial examples and polynomial optimization. InAdvances in Neural Information Processing Systems 32, pages 13737–13747, 2019

    work page 2019

  5. [5]

    fairmlbook.org, 2019

    Solon Barocas, Moritz Hardt, and Arvind Narayanan.Fairness and Machine Learning: Limitations and Opportunities. fairmlbook.org, 2019

    work page 2019

  6. [6]

    The hidden assumptions behind counterfactual ex- planations and principal reasons

    Solon Barocas, Andrew Selbst, and Manish Raghavan. The hidden assumptions behind counterfactual ex- planations and principal reasons. In3rd ACM Conference on Fairness, Accountability, and Transparency, pages 80–89, 2020

    work page 2020

  7. [7]

    RelatIF: Identifying explanatory training samples via relative influence

    Elnaz Barshan, Marc-Etienne Brunet, and Gintare Dziugaite. RelatIF: Identifying explanatory training samples via relative influence. In23rd International Conference on Artificial Intelligence and Statistics, pages 1899–1909, 2020

    work page 1909

  8. [8]

    Information discrepancy in strategic learning

    Yahav Bechavod, Chara Podimata, Zhiwei Steven Wu, and Juba Ziani. Information discrepancy in strategic learning. In38th International Conference on Machine Learning, pages 1691–1715, 2022

    work page 2022

  9. [9]

    Fairness in algorithmic recourse through the lens of substantive equality of opportunity.CoRR, abs/2401.16088, 2024

    Andrew Bell, João Fonseca, Carlo Abrate, Francesco Bonchi, and Julia Stoyanovich. Fairness in algorithmic recourse through the lens of substantive equality of opportunity.CoRR, abs/2401.16088, 2024

    work page arXiv 2024

  10. [10]

    Ben-Tal, L

    A. Ben-Tal, L. El Ghaoui, and A. Nemirovski.Robust Optimization. Princeton University Press, 2009

    work page 2009

  11. [11]

    Fairness in criminal justice risk assessments: The state of the art.Sociological Methods & Research, 50(1):3–44, 2021

    Richard Berk, Hoda Heidari, Shahin Jabbari, Michael Kearns, and Aaron Roth. Fairness in criminal justice risk assessments: The state of the art.Sociological Methods & Research, 50(1):3–44, 2021

    work page 2021

  12. [12]

    Counterfactual metarules for local and global recourse

    Tom Bewley, Salim Amoukou, Saumitra Mishra, Daniele Magazzeni, and Manuela Veloso. Counterfactual metarules for local and global recourse. In41st International Conference on Machine Learning, 2024

    work page 2024

  13. [13]

    Consistent counterfactuals for deep models

    Emily Black, Zifan Wang, and Matt Fredrikson. Consistent counterfactuals for deep models. In10th International Conference on Learning Representations, 2022

    work page 2022

  14. [14]

    Cambridge University Press, 2014

    Stephen Boyd and Lieven Vandenberghe.Convex Optimization. Cambridge University Press, 2014

    work page 2014

  15. [15]

    Generating counterfactual explanations under temporal constraints

    Andrei Buliga, Chiara Di Francescomarino, Chiara Ghidini, Marco Montali, and Massimiliano Ronzani. Generating counterfactual explanations under temporal constraints. In39th Annual AAAI Conference on Artificial Intelligence, pages 15622–15631, 2025

    work page 2025

  16. [16]

    Cambridge University Press, 2006

    Nicolò Cesa-Bianchi and Gábor Lugosi.Prediction, learning, and games. Cambridge University Press, 2006

    work page 2006

  17. [17]

    Feature responsiveness scores: Model-agnostic explanations for recourse

    Seung Hyun Cheon, Anneke Wernerfelt, Sorelle Friedler, and Berk Ustun. Feature responsiveness scores: Model-agnostic explanations for recourse. In13th International Conference on Learning Representations, 2025

    work page 2025

  18. [18]

    Springer, 1967

    John Danskin.The Theory of Max-Min and its Application to Weapons Allocation Problems. Springer, 1967

    work page 1967

  19. [19]

    Retiring Adult: New datasets for fair machine learning

    Frances Ding, Moritz Hardt, John Miller, and Ludwig Schmidt. Retiring Adult: New datasets for fair machine learning. InAdvances in Neural Information Processing Systems 34, pages 6478–6490, 2021

    work page 2021

  20. [20]

    On the adversarial robustness of causal algorithmic recourse

    Ricardo Dominguez-Olmedo, Amir-Hossein Karimi, and Bernhard Schölkopf. On the adversarial robustness of causal algorithmic recourse. In39th International Conference on Machine Learning, pages 5324–5342, 2022. 11

    work page 2022

  21. [21]

    Robust counterfactual explanations for tree-based ensembles

    Sanghamitra Dutta, Jason Long, Saumitra Mishra, Cecilia Tilli, and Daniele Magazzeni. Robust counterfactual explanations for tree-based ensembles. In39th International Conference on Machine Learning, volume 162, pages 5742–5756, 2022

    work page 2022

  22. [22]

    Collective counterfactual explanations via optimal transport.CoRR, abs/2402.04579, 2024

    Ahmad-Reza Ehyaei, Ali Shirali, and Samira Samadi. Collective counterfactual explanations via optimal transport.CoRR, abs/2402.04579, 2024

    work page arXiv 2024

  23. [23]

    The risks of recourse in binary classification

    Hidde Fokkema, Damien Garreau, and Tim van Erven. The risks of recourse in binary classification. In 27th International Conference on Artificial Intelligence and Statistics, pages 550–558, 2024

    work page 2024

  24. [24]

    Setting the right expectations: Algorithmic recourse over time

    João Fonseca, Andrew Bell, Carlo Abrate, Francesco Bonchi, and Julia Stoyanovich. Setting the right expectations: Algorithmic recourse over time. In3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization, pages 29:1–29:11, 2023

    work page 2023

  25. [25]

    On the impact of algorithmic recourse on social segregation

    Ruijiang Gao and Himabindu Lakkaraju. On the impact of algorithmic recourse on social segregation. In40th International Conference on Machine Learning, pages 10727–10743, 2023

    work page 2023

  26. [26]

    From search to sampling: Generative models for robust algorithmic recourse

    Prateek Garg, Lokesh Nagalapatti, and Sunita Sarawagi. From search to sampling: Generative models for robust algorithmic recourse. In13th International Conference on Learning Representations, 2025

    work page 2025

  27. [27]

    South German Credit (UPDATE)

    Ulrike Grömping. South German Credit (UPDATE). UCI Machine Learning Repository, 2020. DOI: https://doi.org/10.24432/C5QG88

    work page doi:10.24432/c5qg88 2020

  28. [28]

    Equalimprovability: A new fairness notion considering the long-term impact

    OzgurGuldogan, YuchenZeng, Jy-yongSohn, RamtinPedarsani, andKangwookLee. Equalimprovability: A new fairness notion considering the long-term impact. In11th International Conference on Learning Representations, 2023

    work page 2023

  29. [29]

    RoCourseNet: Robust training of a prediction aware recourse model

    Hangzhi Guo, Feiran Jia, Jinghui Chen, Anna Squicciarini, and Amulya Yadav. RoCourseNet: Robust training of a prediction aware recourse model. In32nd ACM International Conference on Information and Knowledge Management, pages 619–628. ACM, 2023

    work page 2023

  30. [30]

    Equalizing recourse across groups.CoRR, abs/1909.03166, 2019

    Vivek Gupta, Pegah Nokhiz, Chitradeep Roy, and Suresh Venkatasubramanian. Equalizing recourse across groups.CoRR, abs/1909.03166, 2019

    work page arXiv 1909

  31. [31]

    Robust counterfactual explanations for neural networks with probabilistic guarantees

    Faisal Hamman, Erfaun Noorani, Saumitra Mishra, Daniele Magazzeni, and Sanghamitra Dutta. Robust counterfactual explanations for neural networks with probabilistic guarantees. In40th International Conference on Machine Learning, pages 12351–12367, 2023

    work page 2023

  32. [32]

    Equality of opportunity in supervised learning

    Moritz Hardt, Eric Price, and Nathan Srebro. Equality of opportunity in supervised learning. In30th Annual Conference on Neural Information Processing Systems, pages 3315–3323, 2016

    work page 2016

  33. [33]

    On the long-term impact of algorithmic decision policies: Effort unfairness and feature segregation through social learning

    Hoda Heidari, Vedant Nanda, and Krishna Gummadi. On the long-term impact of algorithmic decision policies: Effort unfairness and feature segregation through social learning. In36th International Conference on Machine Learning, pages 2692–2701, 2019

    work page 2019

  34. [34]

    Statlog (German Credit Data)

    Hans Hofmann. Statlog (German Credit Data). UCI Machine Learning Repository, 1994. DOI: https://doi.org/10.24432/C5NC77

    work page doi:10.24432/c5nc77 1994

  35. [35]

    Recourse under model multiplicity via argumentative ensembling

    Junqi Jiang, Francesco Leofante, Antonio Rago, and Francesca Toni. Recourse under model multiplicity via argumentative ensembling. In23rd International Conference on Autonomous Agents and Multiagent Systems, pages 954–963, 2024

    work page 2024

  36. [36]

    Robust counterfactual explanations in machine learning: A survey

    Junqi Jiang, Francesco Leofante, Antonio Rago, and Francesca Toni. Robust counterfactual explanations in machine learning: A survey. In33rd International Joint Conference on Artificial Intelligence, pages 8086–8094, 2024. 12

    work page 2024

  37. [37]

    Towards Realistic Individual Recourse and Actionable Explanations in Black-Box Decision Making Systems

    Shalmali Joshi, Oluwasanmi Koyejo, Warut Vijitbenjaronk, Been Kim, and Joydeep Ghosh. Towards realistic individual recourse and actionable explanations in black-box decision making systems.CoRR, abs/1907.09615, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  38. [38]

    Learning decision trees and forests with algorithmic recourse

    Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, and Yuichi Ike. Learning decision trees and forests with algorithmic recourse. In41st International Conference on Machine Learning, 2024

    work page 2024

  39. [39]

    Model-agnostic counterfactual explanations for consequential decisions

    Amir-Hossein Karimi, Gilles Barthe, Borja Balle, and Isabel Valera. Model-agnostic counterfactual explanations for consequential decisions. In23rd International Conference on Artificial Intelligence and Statistics, pages 895–905, 2020

    work page 2020

  40. [40]

    Algorithmic recourse under imperfect causal knowledge: a probabilistic approach

    Amir-Hossein Karimi, Bodo Julius von Kügelgen, Bernhard Schölkopf, and Isabel Valera. Algorithmic recourse under imperfect causal knowledge: a probabilistic approach. InAdvances in Neural Information Processing Systems 33, 2020

    work page 2020

  41. [41]

    Learning-augmented private algorithms for multiple quantile release

    Mikhail Khodak, Kareem Amin, Travis Dick, and Sergei Vassilvitskii. Learning-augmented private algorithms for multiple quantile release. In40th International Conference on Machine Learning, volume 202, pages 16344–16376, 2023

    work page 2023

  42. [42]

    Understanding black-box predictions via influence functions

    Pang Wei Koh and Percy Liang. Understanding black-box predictions via influence functions. In34th International Conference on Machine Learning, pages 1885–1894, 2017

    work page 2017

  43. [43]

    The case for learned index structures

    Tim Kraska, Alex Beutel, Ed Chi, Jeffrey Dean, and Neoklis Polyzotis. The case for learned index structures. InInternational Conference on Management of Data, pages 489–504, 2018

    work page 2018

  44. [44]

    Interpretable decision sets: A joint framework for description and prediction

    Himabindu Lakkaraju, Stephen Bach, and Jure Leskovec. Interpretable decision sets: A joint framework for description and prediction. In22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 1675–1684, 2016

    work page 2016

  45. [45]

    Promoting counterfactual robustness through diversity

    Francesco Leofante and Nico Potyka. Promoting counterfactual robustness through diversity. In38th AAAI Conference on Artificial Intelligence, pages 21322–21330, 2024

    work page 2024

  46. [46]

    Implicit bias of gradient descent based adversarial training on separable data

    Yan Li, Ethan Fang, Huan Xu, and Tuo Zhao. Implicit bias of gradient descent based adversarial training on separable data. In8th International Conference on Learning Representations, 2020

    work page 2020

  47. [47]

    Algorithms with predictions, 2024

    Alexander Lindermayr and Nicole Megow. Algorithms with predictions, 2024. URL https:// algorithms-with-predictions.github.io/

    work page 2024

  48. [48]

    Interpretable counterfactual explanations guided by prototypes

    Arnaud Van Looveren and Janis Klaise. Interpretable counterfactual explanations guided by prototypes. InMachine Learning and Knowledge Discovery in Databases, pages 650–665, 2021

    work page 2021

  49. [49]

    A unified approach to interpreting model predictions

    Scott Lundberg and Su-In Lee. A unified approach to interpreting model predictions. InAdvances in Neural Information Processing Systems 30, pages 4765–4774, 2017

    work page 2017

  50. [50]

    Competitive caching with machine learned advice

    Thodoris Lykouris and Sergei Vassilvitskii. Competitive caching with machine learned advice. In35th International Conference on Machine Learning, pages 3302–3311, 2018

    work page 2018

  51. [51]

    Towards deep learning models resistant to adversarial attacks

    Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. Towards deep learning models resistant to adversarial attacks. In6th International Conference on Learning Representations, 2018

    work page 2018

  52. [52]

    should this loan be approved or denied?

    Amy Mickel Min Li and Stanley Taylor. “should this loan be approved or denied?”: A large dataset with class assignment guidelines.Journal of Statistics Education, 26, 2018

    work page 2018

  53. [53]

    Algorithms with predictions

    Michael Mitzenmacher and Sergei Vassilvitskii. Algorithms with predictions. In Tim Roughgarden, editor,Beyond the Worst-Case Analysis of Algorithms, pages 646–662. Cambridge University Press, 2020. 13

    work page 2020

  54. [54]

    Robust explanations for private support vector machines.CoRR, abs/2102.03785, 2021

    Rami Mochaourab, Sugandh Sinha, Stanley Greenstein, and Panagiotis Papapetrou. Robust explanations for private support vector machines.CoRR, abs/2102.03785, 2021

    work page arXiv 2021

  55. [55]

    Distributionally robust recourse action

    Duy Nguyen, Ngoc Bui, and Viet Anh Nguyen. Distributionally robust recourse action. In11th International Conference on Learning Representations, 2023

    work page 2023

  56. [56]

    Robust bayesian recourse

    Tuan-Duy Nguyen, Ngoc Bui, Duy Nguyen, Man-Chung Sue, and Viet Anh Nguyen. Robust bayesian recourse. In38th Conference on Uncertainty in Artificial Intelligence, pages 1498–1508, 2022

    work page 2022

  57. [57]

    Learning model-agnostic counterfactual explanations for tabular data

    Martin Pawelczyk, Klaus Broelemann, and Gjergji Kasneci. Learning model-agnostic counterfactual explanations for tabular data. In29th The ACM Web Conference, pages 3126–3132, 2020

    work page 2020

  58. [58]

    Exploring counterfactual explanations through the lens of adversarial examples: A theoretical and empirical analysis

    Martin Pawelczyk, Chirag Agarwal, Shalmali Joshi, Sohini Upadhyay, and Himabindu Lakkaraju. Exploring counterfactual explanations through the lens of adversarial examples: A theoretical and empirical analysis. In25th International Conference on Artificial Intelligence and Statistics, pages 4574–4594, 2022

    work page 2022

  59. [59]

    Probabilistically robust recourse: Navigating the trade-offs between costs and robustness in algorithmic recourse

    Martin Pawelczyk, Teresa Datta, Johannes van den Heuvel, Gjergji Kasneci, and Himabindu Lakkaraju. Probabilistically robust recourse: Navigating the trade-offs between costs and robustness in algorithmic recourse. In11th International Conference on Learning Representations, 2023

    work page 2023

  60. [60]

    On the trade-off between actionable explanations and the right to be forgotten

    Martin Pawelczyk, Tobias Leemann, Asia Biega, and Gjergji Kasneci. On the trade-off between actionable explanations and the right to be forgotten. In11th International Conference on Learning Representations, 2023

    work page 2023

  61. [61]

    Performative prediction

    Juan Perdomo, Tijana Zrnic, Celestine Mendler-Dünner, and Moritz Hardt. Performative prediction. In 37th International Conference on Machine Learning, pages 7599–7609, 2020

    work page 2020

  62. [62]

    Improving online algorithms via ML predictions

    Manish Purohit, Zoya Svitkina, and Ravi Kumar. Improving online algorithms via ML predictions. In Advances in Neural Information Processing Systems 31, pages 9684–9693, 2018

    work page 2018

  63. [63]

    Beyond individualized recourse: Interpretable and interactive summaries of actionable recourses

    Kaivalya Rawal and Himabindu Lakkaraju. Beyond individualized recourse: Interpretable and interactive summaries of actionable recourses. InAdvances in Neural Information Processing Systems 33, 2020

    work page 2020

  64. [64]

    Why should I trust you?

    Marco Túlio Ribeiro, Sameer Singh, and Carlos Guestrin. “Why should I trust you?": Explaining the predictions of any classifier. In22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 1135–1144, 2016

    work page 2016

  65. [65]

    Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead.Nat

    Cynthia Rudin. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead.Nat. Mach. Intell., 1(5):206–215, 2019

    work page 2019

  66. [66]

    Grad-CAM: Visual explanations from deep networks via gradient-based localization

    Ramprasaath Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-CAM: Visual explanations from deep networks via gradient-based localization. In 16th IEEE International Conference on Computer Vision, pages 618–626, 2017

    work page 2017

  67. [67]

    Counterfactual explanations can be manipulated

    Dylan Slack, Anna Hilgard, Himabindu Lakkaraju, and Sameer Singh. Counterfactual explanations can be manipulated. InAdvances in Neural Information Processing Systems 34, pages 62–75, 2021

    work page 2021

  68. [68]

    SmoothGrad: removing noise by adding noise

    Daniel Smilkov, Nikhil Thorat, Been Kim, Fernanda Viégas, and Martin Wattenberg. Smoothgrad: removing noise by adding noise.CoRR, abs/1706.03825, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  69. [69]

    Axiomatic attribution for deep networks

    Mukund Sundararajan, Ankur Taly, and Qiqi Yan. Axiomatic attribution for deep networks. In34th International Conference on Machine Learning, pages 3319–3328, 2017

    work page 2017

  70. [70]

    Towards robust and reliable algorithmic recourse

    Sohini Upadhyay, Shalmali Joshi, and Himabindu Lakkaraju. Towards robust and reliable algorithmic recourse. InAdvances in Neural Information Processing Systems 34, pages 16926–16937, 2021. 14

    work page 2021

  71. [71]

    Actionable recourse in linear classification

    Berk Ustun, Alexander Spangher, and Yang Liu. Actionable recourse in linear classification. In3rd ACM Conference on Fairness, Accountability, and Transparency, pages 10–19, 2019

    work page 2019

  72. [72]

    The philosophical basis of algorithmic recourse

    Suresh Venkatasubramanian and Mark Alfano. The philosophical basis of algorithmic recourse. In3rd ACM Conference on Fairness, Accountability, and Transparency, pages 284–293, 2020

    work page 2020

  73. [73]

    & Hines, K

    Sahil Verma, John Dickerson, and Keegan Hines. Counterfactual explanations for machine learning: A review.CoRR, abs/2010.10596, 2020

    work page arXiv 2010

  74. [74]

    Efficient training of robust decision trees against adversarial examples

    Daniël Vos and Sicco Verwer. Efficient training of robust decision trees against adversarial examples. In 38th International Conference on Machine Learning, volume 139, pages 10586–10595, 2021

    work page 2021

  75. [75]

    Robust optimal classification trees against adversarial examples

    Daniël Vos and Sicco Verwer. Robust optimal classification trees against adversarial examples. In36th AAAI Conference on Artificial Intelligence, pages 8520–8528, 2022

    work page 2022

  76. [76]

    Counterfactual explanations without opening the black box: automated decisions and the GDPR.Harvard Journal of Law and Technology, 31(2):841–887, 2018

    Sandra Wachter, Brent Mittelstadt, and Chris Russell. Counterfactual explanations without opening the black box: automated decisions and the GDPR.Harvard Journal of Law and Technology, 31(2):841–887, 2018

    work page 2018

  77. [77]

    A bayesian framework for learning rule sets for interpretable classification.J

    Tong Wang, Cynthia Rudin, Finale Doshi-Velez, Yimin Liu, Erica Klampfl, and Perry MacNeille. A bayesian framework for learning rule sets for interpretable classification.J. Mach. Learn. Res., 18: 70:1–70:37, 2017

    work page 2017

  78. [78]

    On the convergence and robustness of adversarial training

    Yisen Wang, Xingjun Ma, James Bailey, Jinfeng Yi, Bowen Zhou, and Quanquan Gu. On the convergence and robustness of adversarial training. In36th International Conference on Machine Learning, pages 6586–6595, 2019

    work page 2019

  79. [79]

    Provable defenses against adversarial examples via the convex outer adversarial polytope

    Eric Wong and Zico Kolter. Provable defenses against adversarial examples via the convex outer adversarial polytope. In35th International Conference on Machine Learning, pages 5283–5292, 2018

    work page 2018

  80. [80]

    Mechanism design with predictions

    Chenyang Xu and Pinyan Lu. Mechanism design with predictions. In31st International Joint Conference on Artificial Intelligence, pages 571–577, 2022

    work page 2022

Showing first 80 references.