1,849 Hits in 6.2 sec

On the Fairness of Causal Algorithmic Recourse [article]

Julius von Kügelgen, Amir-Hossein Karimi, Umang Bhatt, Isabel Valera, Adrian Weller, Bernhard Schölkopf
2022 arXiv   pre-print
Algorithmic fairness is typically studied from the perspective of predictions.  ...  We study theoretically and empirically how to enforce fair causal recourse by altering the classifier and perform a case study on the Adult dataset.  ...  Acknowledgements We are grateful to Chris Russell for insightful feedback on connections to existing fairness notions within machine learning and philosophy.  ... 
arXiv:2010.06529v5 fatcat:mrd5yjey55gvzknxuwvtmgc4ru

A survey of algorithmic recourse:contrastive explanations and consequential recommendations

Amir-Hossein Karimi, Gilles Barthe, Bernhard Schölkopf, Isabel Valera
2022 ACM Computing Surveys  
In this work, we focus on algorithmic recourse , which is concerned with providing explanations and recommendations to individuals who are unfavorably treated by automated decision-making systems.  ...  We first perform an extensive literature review, and align the efforts of many authors by presenting unified definitions , formulations , and solutions to recourse.  ...  AHK is also appreciative of Julius von Kügelgen and Umang Bhatt for fruitful discussions on recourse and fairness, Muhammad Waleed Gondal and the anonymous reviewers for constructive feedback throughout  ... 
doi:10.1145/3527848 fatcat:fo6dntzsvnbfdhwz7bmsgcgb54

A survey of algorithmic recourse: definitions, formulations, solutions, and prospects [article]

Amir-Hossein Karimi, Gilles Barthe, Bernhard Schölkopf, Isabel Valera
2020 arXiv   pre-print
In this work, we focus on algorithmic recourse, which is concerned with providing explanations and recommendations to individuals who are unfavourably treated by automated decision-making systems.  ...  We first perform an extensive literature review, and align the efforts of many authors by presenting unified definitions, formulations, and solutions to recourse.  ...  This is the concern of algorithmic recourse. Contributions: Our review brings together the plethora of recent works on algorithmic recourse.  ... 
arXiv:2010.04050v1 fatcat:zsrqbickgvd4bbhnle4nlk7buq

On the Adversarial Robustness of Causal Algorithmic Recourse [article]

Ricardo Dominguez-Olmedo, Amir-Hossein Karimi, Bernhard Schölkopf
2021 arXiv   pre-print
In order to shift part of the burden of robustness from the decision-subject to the decision-maker, we propose a model regularizer that encourages the additional cost of seeking robust recourse to be low  ...  Recourse recommendations should ideally be robust to reasonably small uncertainty in the features of the individual seeking recourse.  ...  On the fairness of causal algorithmic recourse. ICML 2021 Workshop on Algorithmic Recourse, 2021. [40] Sandra Wachter, Brent Mittelstadt, and Chris Russell.  ... 
arXiv:2112.11313v1 fatcat:ko4gnhopijduhihzmli5iha2mi

Algorithmic Recourse in Partially and Fully Confounded Settings Through Bounding Counterfactual Effects [article]

Julius von Kügelgen, Nikita Agarwal, Jakob Zeitler, Afsaneh Mastouri, Bernhard Schölkopf
2021 arXiv   pre-print
If the lower bound is above a certain threshold, i.e., on the other side of the decision boundary, recourse is guaranteed in expectation.  ...  The proposed approach only requires specification of the causal graph and confounding structure and bounds the expected counterfactual effect of recourse actions.  ...  This work was supported by the German Federal Ministry of Education and Research (BMBF): Tübingen AI Center, FKZ: 01IS18039B; and by the Machine Learning Cluster of Excellence, EXC number 2064/1 -Project  ... 
arXiv:2106.11849v1 fatcat:fcbkijcym5ci3jeebveuq4qlmm

A Causal Perspective on Meaningful and Robust Algorithmic Recourse [article]

Gunnar König, Timo Freiesleben, Moritz Grosse-Wentrup
2021 arXiv   pre-print
Additionally, we introduce a relaxation of MAR called effective algorithmic recourse (EAR), which, under certain assumptions, yields meaningful recourse by only allowing interventions on causes of the  ...  Extending the work of Karimi et al. (2021), we propose meaningful algorithmic recourse (MAR) that only recommends actions that improve both prediction and target.  ...  We take a causal perspective on the issue at hand and argue that robustness and meaningfulness are related problems ICML (International Conference on Machine Learning) Workshop on Algorithmic Recourse.  ... 
arXiv:2107.07853v1 fatcat:dic2ilqtuvh37igrmfppkqmb7y

Explaining Black-Box Algorithms Using Probabilistic Contrastive Counterfactuals [article]

Sainyam Galhotra, Romila Pradhan, Babak Salimi
2021 arXiv   pre-print
Experiments on synthetic data further demonstrate the correctness of LEWIS's explanations and the scalability of its recourse algorithm.  ...  knowledge of the underlying causal model and (3)makes no assumptions about the internals of an algorithmic system except for the availability of its input-output data.  ...  Algorithmic fairness. The critical role of causality and background knowledge is recognized and acknowledged in the algorithmic fairness literature [45, 43, 65, 76, 25, 79, 81, 78] .  ... 
arXiv:2103.11972v2 fatcat:vh6f4f2kvfezrf6i26a2cflqpa

Algorithmic Recourse: from Counterfactual Explanations to Interventions [article]

Amir-Hossein Karimi, Bernhard Schölkopf, Isabel Valera
2020 arXiv   pre-print
In this work, we rely on causal reasoning to caution against the use of counterfactual explanations as a recommendable set of actions for recourse.  ...  Existing works have primarily focused on designing algorithms to obtain counterfactual explanations for a wide range of settings.  ...  Acknowledgments and Disclosure of Funding The authors would like to thank Adrián Javaloy Bornás and Julius von Kügelgen for their valuable feedback on drafts of the manuscript.  ... 
arXiv:2002.06278v4 fatcat:eanijkx2kza5zbysp5bnidn7zm

Multi-Agent Algorithmic Recourse [article]

Andrew O'Brien, Edward Kim
2021 arXiv   pre-print
Past work has largely focused on the effect algorithmic recourse has on a single agent.  ...  In this work, we show that when the assumption of a single agent environment is relaxed, current approaches to algorithmic recourse fail to guarantee certain ethically desirable properties.  ...  fair way.  ... 
arXiv:2110.00673v1 fatcat:ib6ojpe7afa7zaqraqjvchttte

Counterfactual Explanations Can Be Manipulated [article]

Dylan Slack and Sophie Hilgard and Himabindu Lakkaraju and Sameer Singh
2021 arXiv   pre-print
We perform experiments on loan and violent crime prediction data sets where certain subgroups achieve up to 20x lower cost recourse under the perturbation.  ...  We describe how these models can unfairly provide low-cost recourse for specific subgroups in the data while appearing fair to auditors.  ...  The views expressed are those of the authors and do not reflect the official policy or position of the funding agencies.  ... 
arXiv:2106.02666v2 fatcat:twqf4amos5bo5p2npmft4z2pvi

Algorithmic recourse under imperfect causal knowledge: a probabilistic approach [article]

Amir-Hossein Karimi, Julius von Kügelgen, Bernhard Schölkopf, Isabel Valera
2020 arXiv   pre-print
Recent work has discussed the limitations of counterfactual explanations to recommend actions for algorithmic recourse, and argued for the need of taking causal relationships between features into consideration  ...  The second removes any assumptions on the structural equations by instead computing the average effect of recourse actions on individuals similar to the person who seeks recourse, leading to a novel subpopulation-based  ...  Moreover, a special thanks to Adrià Garriga-Alonso for insightful input on some of the GP-derivations and to Adrián Javaloy Bornás for invaluable help with the CVAE-training.  ... 
arXiv:2006.06831v3 fatcat:cxklr77ghbhzxhgjonsgpngy24

Amortized Generation of Sequential Algorithmic Recourses for Black-box Models [article]

Sahil Verma, Keegan Hines, John P. Dickerson
2021 arXiv   pre-print
Algorithmic Recourses (ARs) provide "what if" feedback of the form "if an input datapoint were x' instead of x, then an ML-based system's output would be y' instead of y."  ...  We evaluate our approach using three real-world datasets and show successful generation of sequential ARs that respect other recourse desiderata.  ...  Results Table 3 shows the performance of FASTAR and all the baselines on the recourse desiderata.  ... 
arXiv:2106.03962v2 fatcat:535mjkbur5arragdefsh5so2ve

Causal Learning for Socially Responsible AI [article]

Lu Cheng, Ahmadreza Mosallanezhad, Paras Sheth, Huan Liu
2022 arXiv   pre-print
One of these approaches is causal learning (CL). We survey state-of-the-art methods of CL for SRAI.  ...  We begin by examining the seven CL tools to enhance the social responsibility of AI, then review how existing works have succeeded using these tools to tackle issues in developing SRAI such as fairness  ...  Acknowledgements This material is based upon work supported by, or in part by, the U.S. Army Research Laboratory and the U.S.  ... 
arXiv:2104.12278v2 fatcat:hk42masibjai7pnv5y2t7gj7ja

Algorithmic Recourse in the Wild: Understanding the Impact of Data and Model Shifts [article]

Kaivalya Rawal, Ece Kamar, Himabindu Lakkaraju
2021 arXiv   pre-print
Our theoretical results establish a lower bound on the probability of recourse invalidation due to model shifts, and show the existence of a tradeoff between this invalidation probability and typical notions  ...  of "cost" minimized by modern recourse generation algorithms.  ...  Conclusion In this paper, we analyse the impact of distribution shifts on recourses generated by state-of-the-art algorithms.  ... 
arXiv:2012.11788v3 fatcat:ilfoacj7ufhppb3mkgdmn6defe

Learning Models for Actionable Recourse [article]

Alexis Ross, Himabindu Lakkaraju, Osbert Bastani
2022 arXiv   pre-print
We demonstrate the efficacy of our approach via extensive experiments on real data.  ...  ., an applicant denied a loan) with recourse -- i.e., a description of how they can change their features to obtain a positive outcome.  ...  The views expressed are those of the authors and do not reflect the official policy or position of the funding agencies.  ... 
arXiv:2011.06146v3 fatcat:xe5mjazbczfwnkouuzhmjf5l5e
« Previous Showing results 1 — 15 out of 1,849 results