pith. sign in

arxiv: 2606.17684 · v1 · pith:XH5S6MULnew · submitted 2026-06-16 · 📊 stat.ML · cs.CY· cs.LG

Geometrical fairness in graph neural networks

Arturo P\'erez-Peralta , Sandra Ben\'itez-Pe\~na , Blas Kolic , Rosa E. Lillo This is my paper

Pith reviewed 2026-06-26 22:45 UTC · model grok-4.3

classification 📊 stat.ML cs.CYcs.LG
keywords graph neural networksfairnessgraph diffusionLaplacian operatorbias mitigationspectral filteringsubspace projection
0
0 comments X

The pith

Modifying the Laplacian with projections and filters reduces bias in graph diffusion while preserving task performance.

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

The paper introduces a fairness-aware version of graph diffusion by changing the underlying Laplacian operator. It applies subspace projections, spectral adjustments, and frequency-based filtering to lessen bias-related signals. A sympathetic reader would care because graph models often spread or magnify data biases through their message passing. The work uses the smoothing effect of diffusion to give a theoretical account of why the changes improve fairness. Tests on synthetic and real datasets show the changes deliver better fairness scores at small extra cost.

Core claim

By modifying the Laplacian operator through subspace projections, spectral adjustments, and frequency-based filtering, the diffusion process in graph-based models can selectively reduce bias components. This yields improved fairness metrics on both synthetic and real-world data while task performance stays competitive and added computation remains limited. The analysis rests on the intrinsic smoothing behavior of graph diffusion to establish the fairness properties.

What carries the argument

The modified Laplacian operator that incorporates complementary transformations to mitigate bias components.

If this is right

  • Fairness metrics improve on the tested synthetic and real-world datasets.
  • Task performance remains competitive with unmodified diffusion models.
  • Theoretical fairness properties follow from the smoothing behavior of the altered diffusion.
  • Extra computational cost stays limited.

Where Pith is reading between the lines

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

  • The same Laplacian modifications might transfer to other message-passing architectures that are not explicitly diffusion-based.
  • The approach could be tested on graphs whose bias arises from different sources than those examined in the paper.
  • Scaling experiments on much larger graphs would show whether the spectral steps remain practical.
  • The frequency-filtering step invites direct comparison with existing spectral graph filters used for denoising.

Load-bearing premise

The transformations can be applied to the Laplacian so they remove bias signals without erasing the graph structure needed for the main task.

What would settle it

If fairness metrics fail to improve or task accuracy drops sharply when the method is run on a dataset whose bias structure is independently known, the central claim would be refuted.

Figures

Figures reproduced from arXiv: 2606.17684 by Arturo P\'erez-Peralta, Blas Kolic, Rosa E. Lillo, Sandra Ben\'itez-Pe\~na.

Figure 1
Figure 1. Figure 1: Slow eigenvectors closely aligned with the sensitive attribute on the Laplacian encourage unfair solutions. Geomet￾ric fairness methods rescale the slow biased eigenvalues, thus increasing their frequency and encouraging energy dissipation in the direction aligned with the sensitive attribute. However, despite their merits, GNN architectures are not without their shortcomings, including their tendency to e… view at source ↗
Figure 2
Figure 2. Figure 2: Synthetic experiment. b controls for bias on the data, with b = 0 representing no bias in the graph and b = 1 configuring a high-bias setting in which it is difficult to make correct predictions without compromising statistical parity. As bias increases, our models drop performance while maintaining greater fairness, dominating the baseline in the utility-fairness trade-off in the high-bias regime. 0 6 12 … view at source ↗
Figure 3
Figure 3. Figure 3: First eigenvalues and correlations with the sensitive vector, ordered from smallest to largest eigenvalue. Smaller eigenvalues dissipate slowly, so low-frequency eigenvectors highly correlated with the sensitive attribute induce long-term bias. The quotient of the correlation divided by the eigenvalues provides a scale on bias dissipation, which our models consistently reduce. 3 4 5 6 7 8 9 10 SP (%) - [ ]… view at source ↗
Figure 4
Figure 4. Figure 4: Pareto efficient frontier of the methods. Our models are highlighted in blue, and the Pareto front in red. Our models are close to the frontier, with diagonal models mostly dominating other methods and the general parametrization close behind. Wang, 2021), FairSIN (Yang et al., 2024), and NIFTY (Agar￾wal et al., 2021) in terms of fairness-aware graph models. We also compare our models against standard tabu… view at source ↗
Figure 5
Figure 5. Figure 5: Ridge plots of the Dirichlet energies of the original NSD and the different interventions. Our models close the distance between the distributions of both sensitive groups. This plot shows that FSP is the method that modifies the underlying Laplacian the least [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Accuracy-Fairness trade-offs available when varying the γ parameter. All modifications start from the same point when γ = 0. As γ increases, fairness tends to increase and accuracy tends to decrease. However, the effect of modifications on the spectral gap can lead to spontaneous improvements in accuracy due to its relationship to oversquashing. Dirichlet Energy. In order to build further intuition, we app… view at source ↗
read the original abstract

Graph-based learning methods have become increasingly prominent due to their strong performance across diverse applications. Among these, recent frameworks grounded in diffusion processes provide a unifying perspective that extends traditional graph neural network formulations while addressing limitations of standard message-passing mechanisms. Despite these advances, concerns remain regarding the fairness of such models, as they may propagate or amplify biases present in the data. In this work, we introduce a fairness-aware adaptation of graph-based diffusion by modifying the underlying Laplacian operator. Our approach incorporates multiple complementary transformations, including subspace projections, spectral adjustments, and frequency-based filtering, to mitigate bias-related components. Leveraging the intrinsic smoothing properties of graph diffusion, we provide a principled analysis of the resulting behavior and establish theoretical insights into fairness properties. We evaluate the proposed framework on both synthetic and real-world datasets, demonstrating that it achieves competitive performance while improving fairness metrics with limited additional computational cost.

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 / 2 minor

Summary. The manuscript proposes a fairness-aware adaptation of graph diffusion processes for GNNs by modifying the Laplacian operator via subspace projections, spectral adjustments, and frequency-based filtering to mitigate bias-related components. It claims to leverage the smoothing properties of diffusion for theoretical insights into fairness and demonstrates competitive task performance with improved fairness metrics and limited extra computational cost on synthetic and real-world datasets.

Significance. If the separation of bias from task signal can be rigorously established, the work would offer a geometrically motivated framework for fairness in diffusion-based graph learning, extending standard GNN formulations with principled modifications to the Laplacian. The emphasis on limited computational overhead would be a practical strength if the empirical claims are substantiated with appropriate controls.

major comments (2)
  1. [Abstract] Abstract: The central claim that subspace projections, spectral adjustments, and frequency-based filtering can selectively mitigate bias components in the Laplacian without destroying task-relevant graph structure lacks any explicit condition (e.g., orthogonality between bias subspace and label function, or bounds on the perturbation of the diffusion kernel) guaranteeing that the modified operator remains a valid positive-semidefinite Laplacian whose stationary distribution and convergence rate preserve downstream performance. When bias and signal share spectral support, the transformations necessarily trade one for the other, undermining the fairness-performance tradeoff assertion.
  2. [Theoretical analysis] Theoretical analysis (implied in abstract claims): The 'principled analysis' of fairness properties via diffusion smoothing provides no derivation or theorem establishing how the transformations affect the spectrum while maintaining the required properties of the diffusion process; without such a result, the theoretical insights reduce to informal invocation of smoothing behavior rather than a load-bearing guarantee.
minor comments (2)
  1. [Abstract] The abstract does not specify the exact fairness definitions employed, the choice of baselines, or any statistical significance testing for the reported improvements.
  2. Notation for the modified Laplacian and the individual transformations (subspace projection, spectral adjustment, frequency filter) should be introduced with explicit equations early in the manuscript to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the major comments point by point below, indicating planned revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that subspace projections, spectral adjustments, and frequency-based filtering can selectively mitigate bias components in the Laplacian without destroying task-relevant graph structure lacks any explicit condition (e.g., orthogonality between bias subspace and label function, or bounds on the perturbation of the diffusion kernel) guaranteeing that the modified operator remains a valid positive-semidefinite Laplacian whose stationary distribution and convergence rate preserve downstream performance. When bias and signal share spectral support, the transformations necessarily trade one for the other, undermining the fairness-performance tradeoff assertion.

    Authors: We agree that the abstract would benefit from explicit conditions to support the central claim. In the revision we will add a concise statement noting the key assumption of approximate orthogonality between the bias subspace and the label function (with a cross-reference to the expanded theoretical discussion), along with a brief remark on the resulting bounds for the diffusion kernel. This addresses the concern about shared spectral support by clarifying that the method targets components where such separation holds to a sufficient degree. revision: yes

  2. Referee: [Theoretical analysis] Theoretical analysis (implied in abstract claims): The 'principled analysis' of fairness properties via diffusion smoothing provides no derivation or theorem establishing how the transformations affect the spectrum while maintaining the required properties of the diffusion process; without such a result, the theoretical insights reduce to informal invocation of smoothing behavior rather than a load-bearing guarantee.

    Authors: The referee is correct that the current theoretical section invokes diffusion smoothing without a formal derivation or theorem on the spectral effects. We will add a new proposition in the revised theoretical analysis that derives the impact of the subspace projection, spectral adjustment, and frequency filtering on the Laplacian spectrum, including a bound on the perturbation of the diffusion kernel under the orthogonality assumption. This will convert the insights into a load-bearing guarantee. revision: yes

Circularity Check

0 steps flagged

No circularity; claims rest on empirical results and diffusion properties

full rationale

The abstract and described approach present Laplacian modifications (subspace projections, spectral adjustments, frequency filtering) as a proposed method whose fairness effects are then analyzed via known smoothing behavior of diffusion. No derivation step reduces a claimed prediction or theoretical insight to a fitted parameter or self-citation by construction. Performance and fairness improvements are stated to be demonstrated on datasets, making the central claims externally falsifiable rather than tautological. No self-citation load-bearing, ansatz smuggling, or renaming of known results is visible.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review performed on abstract only; no concrete free parameters, axioms, or invented entities can be extracted. The approach implicitly assumes that bias components are identifiable and removable via linear algebraic operations on the Laplacian without further justification provided.

pith-pipeline@v0.9.1-grok · 5692 in / 1121 out tokens · 32066 ms · 2026-06-26T22:45:28.863279+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

90 extracted references · 2 canonical work pages

  1. [1]

    proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining , pages=

    Certifying and removing disparate impact , author=. proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining , pages=

  2. [2]

    Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Optuna: A next-generation hyperparameter optimization framework , author=. Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  3. [3]

    2020 IEEE international conference on big data (Big Data) , pages=

    Fairness metrics: A comparative analysis , author=. 2020 IEEE international conference on big data (Big Data) , pages=. 2020 , organization=

    2020

  4. [4]

    Artificial intelligence and statistics , pages=

    Fairness constraints: Mechanisms for fair classification , author=. Artificial intelligence and statistics , pages=. 2017 , organization=

    2017

  5. [5]

    2013 IEEE 13th international conference on data mining , pages=

    Controlling attribute effect in linear regression , author=. 2013 IEEE 13th international conference on data mining , pages=. 2013 , organization=

    2013

  6. [6]

    Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency , pages=

    Beyond Consistency: Nuanced Metrics for Individual Fairness , author=. Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency , pages=

    2025

  7. [7]

    arXiv preprint arXiv:2510.18473 , year=

    Benchmarking Fairness-aware Graph Neural Networks in Knowledge Graphs , author=. arXiv preprint arXiv:2510.18473 , year=

    arXiv

  8. [8]

    International scientific conference and international workshop present day trends of innovations , volume=

    Data analysis in public social networks , author=. International scientific conference and international workshop present day trends of innovations , volume=

  9. [9]

    2009 2nd international conference on computer, control and communication , pages=

    Classifying without discriminating , author=. 2009 2nd international conference on computer, control and communication , pages=. 2009 , organization=

    2009

  10. [10]

    Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society , pages=

    Mitigating unwanted biases with adversarial learning , author=. Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society , pages=

    2018

  11. [11]

    European Journal of Operational Research , volume=

    Fairness in credit scoring: Assessment, implementation and profit implications , author=. European Journal of Operational Research , volume=. 2022 , publisher=

    2022

  12. [12]

    37th Annual Conference on Legal Knowledge and Information Systems, JURIX 2024 , pages=

    A case-based-reasoning analysis of the COMPAS Dataset , author=. 37th Annual Conference on Legal Knowledge and Information Systems, JURIX 2024 , pages=. 2024 , organization=

    2024

  13. [13]

    2016 , note =

    Machine Bias , author=. 2016 , note =

    2016

  14. [14]

    2002 , publisher=

    Chebyshev polynomials , author=. 2002 , publisher=

    2002

  15. [15]

    2011 , publisher=

    Numerical methods for large eigenvalue problems: revised edition , author=. 2011 , publisher=

    2011

  16. [16]

    2005 , publisher=

    An introduction to difference equations , author=. 2005 , publisher=

    2005

  17. [17]

    Variations on Barb

    Farkas, B. Variations on Barb. The American Mathematical Monthly , volume=. 2016 , publisher=

    2016

  18. [18]

    2026 , eprint=

    On the use of graph models to achieve individual and group fairness , author=. 2026 , eprint=

    2026

  19. [19]

    2023 , isbn =

    Liu, Zheyuan and Zhang, Chunhui and Tian, Yijun and Zhang, Erchi and Huang, Chao and Ye, Yanfang and Zhang, Chuxu , title =. 2023 , isbn =. doi:10.1145/3543507.3583207 , booktitle =

    work page doi:10.1145/3543507.3583207 2023

  20. [20]

    Deniz and Shen, Yanning , booktitle=

    Kose, O. Deniz and Shen, Yanning , booktitle=. Fairness-aware Adaptive Network Link Prediction , year=

  21. [21]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Eqgnn: Equalized node opportunity in graphs , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  22. [22]

    ACM Transactions on Knowledge Discovery from Data , volume=

    Fairness-aware graph neural networks: A survey , author=. ACM Transactions on Knowledge Discovery from Data , volume=. 2024 , publisher=

    2024

  23. [23]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    Fairness in graph mining: A survey , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2023 , publisher=

    2023

  24. [24]

    arXiv preprint arXiv:2201.03681 , year=

    Fairedit: Preserving fairness in graph neural networks through greedy graph editing , author=. arXiv preprint arXiv:2201.03681 , year=

    arXiv

  25. [25]

    Uncertainty in artificial intelligence , pages=

    Towards a unified framework for fair and stable graph representation learning , author=. Uncertainty in artificial intelligence , pages=. 2021 , organization=

    2021

  26. [26]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Fairsin: Achieving fairness in graph neural networks through sensitive information neutralization , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  27. [27]

    Proceedings of the 14th ACM international conference on web search and data mining , pages=

    Say no to the discrimination: Learning fair graph neural networks with limited sensitive attribute information , author=. Proceedings of the 14th ACM international conference on web search and data mining , pages=

  28. [28]

    2023 57th Asilomar Conference on Signals, Systems, and Computers , pages=

    Fairness-aware graph filter design , author=. 2023 57th Asilomar Conference on Signals, Systems, and Computers , pages=. 2023 , organization=

    2023

  29. [29]

    2024 IEEE 13rd Sensor Array and Multichannel Signal Processing Workshop (SAM) , pages=

    Filtering as Rewiring for Bias Mitigation on Graphs , author=. 2024 IEEE 13rd Sensor Array and Multichannel Signal Processing Workshop (SAM) , pages=. 2024 , organization=

    2024

  30. [30]

    IEEE Transactions on Artificial Intelligence , volume=

    Fairdrop: Biased edge dropout for enhancing fairness in graph representation learning , author=. IEEE Transactions on Artificial Intelligence , volume=. 2021 , publisher=

    2021

  31. [31]

    International Conference on Learning Representations , year=

    FoSR: First-order spectral rewiring for addressing oversquashing in GNNs , author=. International Conference on Learning Representations , year=

  32. [32]

    Advances in neural information processing systems , volume=

    Beyond homophily in graph neural networks: Current limitations and effective designs , author=. Advances in neural information processing systems , volume=

  33. [33]

    Journal of Machine Learning Research , volume=

    Community detection and stochastic block models: recent developments , author=. Journal of Machine Learning Research , volume=

  34. [34]

    Proceedings of the web conference 2021 , pages=

    Pick and choose: a GNN-based imbalanced learning approach for fraud detection , author=. Proceedings of the web conference 2021 , pages=

    2021

  35. [35]

    The World Wide Web Conference , pages=

    Semi-supervised graph classification: A hierarchical graph perspective , author=. The World Wide Web Conference , pages=

  36. [36]

    Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Graph convolutional neural networks for web-scale recommender systems , author=. Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  37. [37]

    Advances in neural information processing systems , volume=

    Link prediction based on graph neural networks , author=. Advances in neural information processing systems , volume=

  38. [38]

    Social Network Analysis and Mining , volume=

    Enhancing social and collaborative learning using a stacked GNN-based community detection , author=. Social Network Analysis and Mining , volume=. 2024 , publisher=

    2024

  39. [39]

    Machine Learning: Science and Technology , volume=

    Graph neural networks in particle physics , author=. Machine Learning: Science and Technology , volume=. 2020 , publisher=

    2020

  40. [40]

    Drug discovery today , volume=

    Graph neural networks for automated de novo drug design , author=. Drug discovery today , volume=. 2021 , publisher=

    2021

  41. [41]

    International conference on machine learning , pages=

    Neural message passing for quantum chemistry , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  42. [42]

    SIAM Journal on Applied Mathematics , volume=

    Opinion dynamics on discourse sheaves , author=. SIAM Journal on Applied Mathematics , volume=. 2021 , publisher=

    2021

  43. [43]

    2014 , publisher=

    Sheaves, cosheaves and applications , author=. 2014 , publisher=

    2014

  44. [44]

    Advances in neural information processing systems , volume=

    Adaptive diffusion in graph neural networks , author=. Advances in neural information processing systems , volume=

  45. [45]

    International Conference on Learning Representations , year=

    How Powerful are Graph Neural Networks? , author=. International Conference on Learning Representations , year=

  46. [46]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Semi-supervised learning with graph learning-convolutional networks , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  47. [47]

    International Conference on Learning Representations , year=

    Graph Attention Networks , author=. International Conference on Learning Representations , year=

  48. [48]

    Advances in neural information processing systems , volume=

    Inductive representation learning on large graphs , author=. Advances in neural information processing systems , volume=

  49. [49]

    Neurocomputing , pages=

    Over-squashing in graph neural networks: A comprehensive survey , author=. Neurocomputing , pages=. 2025 , publisher=

    2025

  50. [50]

    International Conference on Machine Learning , pages=

    Revisiting over-smoothing and over-squashing using ollivier-ricci curvature , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  51. [51]

    Advances in Neural Information Processing Systems , volume=

    Not too little, not too much: a theoretical analysis of graph (over) smoothing , author=. Advances in Neural Information Processing Systems , volume=

  52. [52]

    IEEE Transactions on Neural Networks and Learning Systems , volume=

    Another perspective of over-smoothing: Alleviating semantic over-smoothing in deep GNNs , author=. IEEE Transactions on Neural Networks and Learning Systems , volume=. 2024 , publisher=

    2024

  53. [53]

    Advances in neural information processing systems , volume=

    Revisiting heterophily for graph neural networks , author=. Advances in neural information processing systems , volume=

  54. [54]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Graph neural networks with heterophily , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  55. [55]

    Proceedings of the 2025 SIAM International Conference on Data Mining (SDM) , pages=

    Unveiling the impact of local homophily on gnn fairness: In-depth analysis and new benchmarks , author=. Proceedings of the 2025 SIAM International Conference on Data Mining (SDM) , pages=. 2025 , organization=

    2025

  56. [56]

    arXiv preprint arXiv:2207.04376 , year=

    On graph neural network fairness in the presence of heterophilous neighborhoods , author=. arXiv preprint arXiv:2207.04376 , year=

    arXiv

  57. [57]

    Advances in Neural Information Processing Systems , volume=

    Neural sheaf diffusion: A topological perspective on heterophily and oversmoothing in gnns , author=. Advances in Neural Information Processing Systems , volume=

  58. [58]

    arXiv preprint arXiv:2512.00242 , year=

    Polynomial Neural Sheaf Diffusion: A Spectral Filtering Approach on Cellular Sheaves , author=. arXiv preprint arXiv:2512.00242 , year=

    Pith/arXiv arXiv

  59. [59]

    Big data's disparate impact , author=. Calif. L. Rev. , volume=. 2016 , publisher=

    2016

  60. [60]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , pages=

    A geometric solution to fair representations , author=. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , pages=

  61. [61]

    CEUR workshop proceedings , volume=

    A geometric framework for fairness , author=. CEUR workshop proceedings , volume=. 2023 , organization=

    2023

  62. [62]

    Advances in neural information processing systems , volume=

    Assessing social and intersectional biases in contextualized word representations , author=. Advances in neural information processing systems , volume=

  63. [63]

    2019 IEEE Conference on Visual Analytics Science and Technology (VAST) , pages=

    FairVis: Visual analytics for discovering intersectional bias in machine learning , author=. 2019 IEEE Conference on Visual Analytics Science and Technology (VAST) , pages=. 2019 , organization=

    2019

  64. [64]

    International Conference on Smart Data Intelligence , pages=

    Fairness in predicting recidivism score , author=. International Conference on Smart Data Intelligence , pages=. 2024 , organization=

    2024

  65. [65]

    Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Fairness in machine learning for healthcare , author=. Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  66. [66]

    Translation tutorial: 21 fairness definitions and their politics , author=. Proc. conf. fairness accountability transp., new york, usa , volume=

  67. [67]

    2023 , publisher=

    Fairness and machine learning: Limitations and opportunities , author=. 2023 , publisher=

    2023

  68. [68]

    Proceedings of the 2019 CHI conference on human factors in computing systems , pages=

    Improving fairness in machine learning systems: What do industry practitioners need? , author=. Proceedings of the 2019 CHI conference on human factors in computing systems , pages=

    2019

  69. [69]

    2009 IEEE international conference on data mining workshops , pages=

    Building classifiers with independency constraints , author=. 2009 IEEE international conference on data mining workshops , pages=. 2009 , organization=

    2009

  70. [70]

    Joint European conference on machine learning and knowledge discovery in databases , pages=

    Fairness-aware classifier with prejudice remover regularizer , author=. Joint European conference on machine learning and knowledge discovery in databases , pages=. 2012 , organization=

    2012

  71. [71]

    Proceedings of the conference on fairness, accountability, and transparency , pages=

    Classification with fairness constraints: A meta-algorithm with provable guarantees , author=. Proceedings of the conference on fairness, accountability, and transparency , pages=

  72. [72]

    2012 IEEE 12th international conference on data mining , pages=

    Decision theory for discrimination-aware classification , author=. 2012 IEEE 12th international conference on data mining , pages=. 2012 , organization=

    2012

  73. [73]

    Advances in neural information processing systems , volume=

    Equality of opportunity in supervised learning , author=. Advances in neural information processing systems , volume=

  74. [74]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    The regression analysis of binary sequences , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 1958 , publisher=

    1958

  75. [75]

    Advances in neural information processing systems , volume=

    Lightgbm: A highly efficient gradient boosting decision tree , author=. Advances in neural information processing systems , volume=

  76. [76]

    nature , volume=

    Learning representations by back-propagating errors , author=. nature , volume=. 1986 , publisher=

    1986

  77. [77]

    Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =

    Chen, Tianqi and Guestrin, Carlos , title =. 2016 , isbn =. doi:10.1145/2939672.2939785 , booktitle =

    work page doi:10.1145/2939672.2939785 2016

  78. [78]

    Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Inform: Individual fairness on graph mining , author=. Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  79. [79]

    Proceedings of the 3rd innovations in theoretical computer science conference , pages=

    Fairness through awareness , author=. Proceedings of the 3rd innovations in theoretical computer science conference , pages=

  80. [80]

    Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    A unified approach to quantifying algorithmic unfairness: Measuring individual &group unfairness via inequality indices , author=. Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

Showing first 80 references.