Privacy Constrained Fairness Estimation for Decision Trees

Florian van der Steen; Fr\'e Vink; Heysem Kaya

arxiv: 2312.08413 · v1 · submitted 2023-12-13 · 💻 cs.LG · cs.CR· cs.CY

Privacy Constrained Fairness Estimation for Decision Trees

Florian van der Steen , Fr\'e Vink , Heysem Kaya This is my paper

Pith reviewed 2026-05-24 04:58 UTC · model grok-4.3

classification 💻 cs.LG cs.CRcs.CY

keywords differential privacyfairnessdecision treesstatistical parityprivacy preservinginterpretable modelsPAFER

0 comments

The pith

A trusted third party adds Laplacian noise to sensitive attributes so statistical parity of decision trees can be estimated with low error while preserving privacy.

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

The paper examines the tension between measuring fairness in AI, protecting sensitive personal data, and keeping models interpretable. It proposes the PAFER method to estimate statistical parity for decision trees under differential privacy constraints. A legal entity holds the sensitive attributes and adds noise before the calculation occurs. Experiments indicate the Laplacian mechanism keeps estimation error low and delivers stronger privacy protection. The same method shows better results on decision trees that humans find simpler to read.

Core claim

PAFER estimates statistical parity in decision trees by routing sensitive attribute data through a trusted third-party legal entity that applies differential privacy noise, specifically the Laplacian mechanism, before any fairness computation. This produces low-error parity estimates while providing high-certainty privacy for individuals. Both experiments and theory show the estimates are more accurate when the underlying decision trees are the ones humans generally find easier to interpret.

What carries the argument

Privacy-Aware Fairness Estimation of Rules (PAFER), which applies the Laplacian differential privacy mechanism to sensitive attributes held by a third party before computing statistical parity on decision tree rules.

Load-bearing premise

A trusted third-party legal entity exists that securely holds the sensitive attribute data and applies the differential privacy noise before fairness estimation occurs.

What would settle it

A benchmark dataset experiment in which the Laplacian mechanism produces statistical parity error rates substantially above the levels reported in the paper or allows re-identification of individuals despite the added noise.

read the original abstract

The protection of sensitive data becomes more vital, as data increases in value and potency. Furthermore, the pressure increases from regulators and society on model developers to make their Artificial Intelligence (AI) models non-discriminatory. To boot, there is a need for interpretable, transparent AI models for high-stakes tasks. In general, measuring the fairness of any AI model requires the sensitive attributes of the individuals in the dataset, thus raising privacy concerns. In this work, the trade-offs between fairness, privacy and interpretability are further explored. We specifically examine the Statistical Parity (SP) of Decision Trees (DTs) with Differential Privacy (DP), that are each popular methods in their respective subfield. We propose a novel method, dubbed Privacy-Aware Fairness Estimation of Rules (PAFER), that can estimate SP in a DP-aware manner for DTs. DP, making use of a third-party legal entity that securely holds this sensitive data, guarantees privacy by adding noise to the sensitive data. We experimentally compare several DP mechanisms. We show that using the Laplacian mechanism, the method is able to estimate SP with low error while guaranteeing the privacy of the individuals in the dataset with high certainty. We further show experimentally and theoretically that the method performs better for DTs that humans generally find easier to interpret.

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.

Desk Editor's Note private letter to a colleague

PAFER estimates statistical parity on decision trees by routing sensitive attributes through a third-party legal entity that adds Laplacian noise, but the privacy claim collapses without that intermediary.

read the letter

The main takeaway is that this paper gives a concrete way to audit statistical parity on decision trees while injecting differential privacy noise, but only if a trusted third party holds the sensitive labels and applies the mechanism first. They call the approach PAFER and test it on decision trees specifically, comparing Laplacian, Gaussian, and other mechanisms in experiments. The Laplacian version keeps estimation error low while meeting the privacy bound, and both the experiments and some theory suggest the error stays smaller on trees that people find easier to read anyway. That last part is a modest but useful observation for anyone who wants interpretable models in regulated settings. The work is new in its narrow focus on trees plus the explicit third-party DP wrapper, though it builds directly on existing DP mechanisms the authors cite. The experiments appear to support the low-error claim under the stated conditions, and the interpretability angle is backed by both runs and a theoretical sketch. The soft spot is obvious and load-bearing: without the trusted legal entity that securely holds the sensitive data and adds the noise, the privacy guarantee does not apply and the reported results cannot be reproduced in the same way. The abstract does not spell out tight error bounds after noise addition or how tree depth interacts with the mechanism, so those details will need checking in the full text. The paper is aimed at people who already work on fairness audits in privacy-constrained domains such as finance or healthcare and who also care about keeping models interpretable. It is worth sending to peer review because it tackles a real compliance tension with a working method and some supporting runs, even though reviewers will rightly press on the third-party assumption and the exact error analysis.

Referee Report

2 major / 1 minor

Summary. The paper proposes PAFER, a method for estimating statistical parity (SP) of decision trees under differential privacy. It relies on a trusted third-party legal entity to hold sensitive attribute data and apply the Laplacian mechanism to add noise before fairness estimation. Experiments compare DP mechanisms and claim that the Laplacian approach yields low SP estimation error while providing high privacy certainty; the method is also claimed to perform better (experimentally and theoretically) on DTs that humans find easier to interpret.

Significance. If the error bounds and privacy guarantees can be realized without an idealized trusted entity and if the interpretability claim is rigorously supported, the work would address a relevant intersection of privacy, fairness, and model interpretability. The explicit use of an external DP mechanism rather than self-referential fitting is a methodological strength, but the overall contribution is limited by the load-bearing trust assumption.

major comments (2)

[Abstract] Abstract: the privacy guarantee that PAFER 'guarantees the privacy of the individuals in the dataset with high certainty' is predicated on the existence of a 'third-party legal entity that securely holds this sensitive data' and applies the Laplacian mechanism. No trust model, threat analysis, or alternative (e.g., local DP or secure multi-party computation) is provided, rendering the central privacy claim inapplicable in the absence of this entity.
[Abstract] Abstract: the claim that the method 'performs better for DTs that humans generally find easier to interpret' is presented as both experimental and theoretical support, yet no definition or metric for 'easier to interpret' is given, nor is the theoretical argument sketched. This weakens the cross-claim linking interpretability to estimation quality.

minor comments (1)

The abstract states that 'several DP mechanisms' are compared experimentally but only names the Laplacian result; a table or section listing all mechanisms and their error/privacy metrics would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed comments on our manuscript. We address each major comment below, proposing revisions to strengthen the presentation of our assumptions and claims while preserving the core contribution of PAFER.

read point-by-point responses

Referee: [Abstract] Abstract: the privacy guarantee that PAFER 'guarantees the privacy of the individuals in the dataset with high certainty' is predicated on the existence of a 'third-party legal entity that securely holds this sensitive data' and applies the Laplacian mechanism. No trust model, threat analysis, or alternative (e.g., local DP or secure multi-party computation) is provided, rendering the central privacy claim inapplicable in the absence of this entity.

Authors: The PAFER approach is formulated under an explicit central DP model that assumes a trusted third-party legal entity holds the sensitive attributes and applies the Laplacian mechanism. This assumption is stated in the problem setup and abstract. We agree that the manuscript would benefit from greater clarity on the trust model. In revision, we will add a subsection detailing the trust assumptions, a brief threat analysis consistent with the central model, and a note on why local DP or MPC alternatives fall outside the current scope (as they would alter the mechanism and error analysis). This clarification does not alter the technical results but addresses applicability. revision: yes
Referee: [Abstract] Abstract: the claim that the method 'performs better for DTs that humans generally find easier to interpret' is presented as both experimental and theoretical support, yet no definition or metric for 'easier to interpret' is given, nor is the theoretical argument sketched. This weakens the cross-claim linking interpretability to estimation quality.

Authors: The manuscript links better performance to simpler decision trees via experiments on trees of varying depth and rule count, with a theoretical argument that lower complexity reduces the relative impact of Laplacian noise on the SP estimate. We agree that the abstract claim requires an explicit definition and sketched argument. In revision, we will define 'easier to interpret' in terms of tree depth and number of leaves (standard proxies in the interpretability literature), include the theoretical sketch in the main text, and tie the experimental results directly to this metric. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on standard DP mechanisms

full rationale

The paper introduces PAFER as a method to estimate statistical parity for decision trees under differential privacy by adding Laplacian noise via a trusted third-party entity. This setup draws on established DP properties and experimental validation rather than any self-definitional loop, fitted parameter renamed as prediction, or self-citation chain that reduces the central claim to its own inputs. The privacy guarantee and error bounds follow directly from the external Laplacian mechanism without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the method appears to rest on standard differential privacy primitives and the existence of a trusted third party.

pith-pipeline@v0.9.0 · 5767 in / 1132 out tokens · 20993 ms · 2026-05-24T04:58:20.006982+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages

[1]

Procedia Computer Science 162, 503–513 (2019) https://doi.org/10.1016/j.procs.2019.12.017

Zhu, L., Qiu, D., Ergu, D., Ying, C., Liu, K.: A study on predicting loan d efault based on the random forest algorithm. Procedia Computer Science 162, 503–513 (2019) https://doi.org/10.1016/j.procs.2019.12.017

work page doi:10.1016/j.procs.2019.12.017 2019
[2]

Science 187(4175), 398–404 (1975) https://doi.org/10.1126/science.187.4175.398

Bickel, P.J., Hammel, E.A., O’Connell, J.W.: Sex bias in graduate admission s: Data from berkeley: Measuring bias is harder than is usually assumed , and the evidence is sometimes contrary to expectation. Science 187(4175), 398–404 (1975) https://doi.org/10.1126/science.187.4175.398

work page doi:10.1126/science.187.4175.398 1975
[3]

Big Data 5(2), 153–163 (2017) https://doi.org/10.1089/big.2016.0047

Chouldechova, A.: Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments. Big Data 5(2), 153–163 (2017) https://doi.org/10.1089/big.2016.0047 . Accessed 2023-01-13

work page doi:10.1089/big.2016.0047 2017
[4]

Section: Digits (2015)

Barr, A.: Google Mistakenly Tags Black People as ‘Goril- las,’ Showing Limits of Algorithms. Section: Digits (2015). https://www.wsj.com/articles/BL-DGB-42522 Accessed 2023-02-04

work page 2015
[5]

https://www.amnesty.org/en/documents/eur35/4686/2021/en/ Accessed 2023-02-04 37

Xenophobic machines: Discrimination through unregulated use of algorithms in the Dutch childcare beneﬁts scandal (2021). https://www.amnesty.org/en/documents/eur35/4686/2021/en/ Accessed 2023-02-04 37

work page 2021
[6]

Oﬃcial Journal of the European Union L 119 , 1–88 (14-04-2016)

Regulation (eu) 2016/679 of the european parliament and of the council of 27 april 2016 on the protection of natural persons with regard to th e processing of personal data and on the free movement of such data, and repea ling directive 95/46/ec (general data protection regulation). Oﬃcial Journal of the European Union L 119 , 1–88 (14-04-2016)

work page 2016
[7]

Oﬃcial Journal of the Europ ean Union COM/2021/206 ﬁnal , 1–107 (21-04-2021)

Proposal for a regulation of the european parliament and of the council laying down harmonised rules on artiﬁcial intelligence (artiﬁcial intelligence a ct) and amending certain union legislative acts. Oﬃcial Journal of the Europ ean Union COM/2021/206 ﬁnal , 1–107 (21-04-2021)

work page 2021
[8]

https://algoritmes.overheid.nl/ Accessed 2023-02-04

Het Algoritmeregister van de Nederlandse overheid (2022). https://algoritmes.overheid.nl/ Accessed 2023-02-04

work page 2022
[9]

Oﬃcial Journal of the European Union L 60/34 , 34–85 (04-02-2014)

Directive 2014/17/eu of the european parliament and of the cou ncil of 4 february 2014 on credit agreements for consumers relating to residential im movable prop- erty and amending directives 2008/48/ec and 2013/36/eu and reg ulation (eu). Oﬃcial Journal of the European Union L 60/34 , 34–85 (04-02-2014)

work page 2014
[10]

The Guardian (2018)

Cadwalladr, C., Graham-Harrison, E.: Revealed: 50 million Faceboo k proﬁles har- vested for Cambridge Analytica in major data breach. The Guardian (2018). Chap. News. Accessed 2023-02-04

work page 2018
[11]

https://www.upguard.com/breaches/facebook-user-data-lea k Accessed 2023-02-04

Losing Face: Two More Cases of Third-Party Facebook App Dat a Exposure |UpGuard (2019). https://www.upguard.com/breaches/facebook-user-data-lea k Accessed 2023-02-04

work page 2019
[12]

Artiﬁcial Intelligence and Law 22(2), 175–209 (2014) 38 https://doi.org/10.1007/s10506-013-9152-0

Berendt, B., Preibusch, S.: Better decision support through e xploratory discrimination-aware data mining: foundations and empirical evi- dence. Artiﬁcial Intelligence and Law 22(2), 175–209 (2014) 38 https://doi.org/10.1007/s10506-013-9152-0

work page doi:10.1007/s10506-013-9152-0 2014
[13]

https://www.propublica.org/article/machine-bias-risk-assessme nts-in-criminal-sentencing Accessed 2023-01-06

Mattu, S., Larson, J., Kirchner, L., Angwin, J.: Machine Bias (201 6). https://www.propublica.org/article/machine-bias-risk-assessme nts-in-criminal-sentencing Accessed 2023-01-06

work page 2023
[14]

In: Proceed ings of the Inter- national Workshop on Software Fairness, pp

Verma, S., Rubin, J.: Fairness deﬁnitions explained. In: Proceed ings of the Inter- national Workshop on Software Fairness, pp. 1–7. ACM, Gothenb urg Sweden (2018). https://doi.org/10.1145/3194770.3194776

work page doi:10.1145/3194770.3194776 2018
[15]

In: Proceedings of the 3rd Innovations in Theoretica l Computer Science Conference On - ITCS ’12, pp

Dwork, C., Hardt, M., Pitassi, T., Reingold, O., Zemel, R.: Fairness t hrough awareness. In: Proceedings of the 3rd Innovations in Theoretica l Computer Science Conference On - ITCS ’12, pp. 214–226. ACM Press, Cambr idge, Massachusetts (2012). https://doi.org/10.1145/2090236.2090255

work page doi:10.1145/2090236.2090255 2012
[16]

FEDERAL REGIST ER 44(43) (01-03-1979)

Questions and answers to clarify and provide a common interpre tation of the uni- form guidelines on employee selection procedures. FEDERAL REGIST ER 44(43) (01-03-1979)

work page 1979
[17]

In: Advances in Neural Information Processing Sy stems, vol

Hardt, M., Price, E., Price, E., Srebro, N.: Equality of Opportunit y in Super- vised Learning. In: Advances in Neural Information Processing Sy stems, vol. 29, pp. 3315–3323. Curran Associates, Inc., Red Hook, New York (2016). https://proceedings.neurips.cc/paper/2016/hash/9d26823 67c3935defcb1f9e247a97c0d- Abstract.html Accessed 2023-01-10

work page 2016
[18]

Machine Le arning 106, 1039–1082 (2017) https://doi.org/10.1007/s10994-017-5633-9

Bertsimas, D., Dunn, J.: Optimal classiﬁcation trees. Machine Le arning 106, 1039–1082 (2017) https://doi.org/10.1007/s10994-017-5633-9

work page doi:10.1007/s10994-017-5633-9 2017
[19]

In: Machin e Learning: ECML-94: European Conference on Machine Learning Catania, Ita ly, April 6–8, 1994 Proceedings 7, pp

Oliver, J.J., Hand, D.: Averaging over decision stumps. In: Machin e Learning: ECML-94: European Conference on Machine Learning Catania, Ita ly, April 6–8, 1994 Proceedings 7, pp. 231–241 (1994). https://doi.org/10.1007/3-540-57868-4 39 61 . Springer

work page doi:10.1007/3-540-57868-4 1994
[20]

Molnar, C.: Interpretable Machine Learning, 2nd edn. (2022). https://christophm.github.io/interpretable-ml-book

work page 2022
[21]

Borisov, T

Borisov, V., Leemann, T., Seßler, K., Haug, J., Pawelczyk, M., Kas - neci, G.: Deep neural networks and tabular data: A survey. IEEE Transactions on Neural Networks and Learning Systems (2022) https://doi.org/10.1109/TNNLS.2022.3229161

work page doi:10.1109/tnnls.2022.3229161 2022
[22]

Differential Privacy,

Dwork, C.: Diﬀerential privacy. In: 33rd International Colloqu ium, ICALP 2006 on Automata, Languages and Programming, pp. 1–12. Spring er, Berlin, Heidelberg (2006). https://doi.org/10.1007/11787006 1

work page doi:10.1007/11787006 2006
[23]

2014.The Algorithmic Foundations of Differential Privacy

Dwork, C., Roth, A.: The Algorithmic Foundations of Diﬀerential P rivacy. Foun- dations and Trends in Theoretical Computer Science 9(3-4), 211–407 (2013) https://doi.org/10.1561/0400000042 . Accessed 2023-01-27

work page doi:10.1561/0400000042 2013
[24]

In: 48th Annual IEEE Symposium on Foundations of Computer Science (FOCS ’07), pp

McSherry, F., Talwar, K.: Mechanism design via diﬀerential privac y. In: 48th Annual IEEE Symposium on Foundations of Computer Science (FOCS ’07), pp. 94–103 (2007). https://doi.org/10.1109/FOCS.2007.66 . IEEE

work page doi:10.1109/focs.2007.66 2007
[25]

In: 2010 IEEE International Conference on Data Mining , pp

Kamiran, F., Calders, T., Pechenizkiy, M.: Discrimination Aware Dec ision Tree Learning. In: 2010 IEEE International Conference on Data Mining , pp. 869–874. IEEE, Sydney, Australia (2010). https://doi.org/10.1109/ICDM.2010.50

work page doi:10.1109/icdm.2010.50 2010
[26]

Springer, Berlin, Heidelber g (1999)

Ausiello, G., Crescenzi, P., Gambosi, G., Kann, V., Marchetti-Spac camela, A., Protasi, M.: Complexity and Approximation: Combinatorial Optimizatio n Prob- lems and Their Approximability Properties. Springer, Berlin, Heidelber g (1999). https://doi.org/10.1007/978-3-642-58412-1 40

work page doi:10.1007/978-3-642-58412-1 1999
[27]

In: Koyejo, S., Mohamed, S., Agarwal, A., B elgrave, D., Cho, K., Oh, A

Linden, J., Weerdt, M., Demirovi´ c, E.: Fair and optimal decision trees: A dynamic programming approach. In: Koyejo, S., Mohamed, S., Agarwal, A., B elgrave, D., Cho, K., Oh, A. (eds.) Advances in Neural Information Processin g Systems, vol. 35, pp. 38899–38911. Curran Associates, Inc., Red Hook, Ne w York (2022). https://proceedings.neurips.cc/paper ﬁle...

work page 2022
[28]

Jo, N., Aghaei, S., Benson, J., G´ omez, A., Vayanos, P.: Learning Opti- mal Fair Classiﬁcation Trees. arXiv. arXiv:2201.09932 [cs, math] (20 22). http://arxiv.org/abs/2201.09932 Accessed 2022-12-08

work page arXiv 2022
[29]

In: 2015 IEEE 28th International Symposium on Computer-Based Medical Systems, pp

Mohammed, N., Barouti, S., Alhadidi, D., Chen, R.: Secure and Priva te Manage- ment of Healthcare Databases for Data Mining. In: 2015 IEEE 28th International Symposium on Computer-Based Medical Systems, pp. 191–196. IE EE, Sao Carlos, Brazil (2015). https://doi.org/10.1109/CBMS.2015.54

work page doi:10.1109/cbms.2015.54 2015
[30]

ACM Computing Surveys 52(4), 1–33 (2020) https://doi.org/10.1145/3337064

Fletcher, S., Islam, M.Z.: Decision Tree Classiﬁcation with Diﬀeren- tial Privacy: A Survey. ACM Computing Surveys 52(4), 1–33 (2020) https://doi.org/10.1145/3337064

work page doi:10.1145/3337064 2020
[31]

In: Advances in Neural Information Processing System s, vol

Lahoti, P., Beutel, A., Chen, J., Lee, K., Prost, F., Thain, N., Wang , X., Chi, E.: Fairness without Demographics through Adversarially Reweig hted Learning. In: Advances in Neural Information Processing System s, vol. 33, pp. 728–740. Curran Associates, Inc., Red Hook, New York (2 020). https://proceedings.neurips.cc/paper/2020/hash/07fc15c 9d169ee48573...

work page 2020
[32]

In: Proceedings of the Fifteenth ACM International Conference on Web Search And Data Mining, pp

Zhao, T., Dai, E., Shu, K., Wang, S.: Towards Fair Classiﬁers Withou t Sensitive Attributes: Exploring Biases in Related Features. In: Proceedings of the Fifteenth ACM International Conference on Web Search And Data Mining, pp. 1433–1442. 41 ACM, Virtual Event AZ USA (2022). https://doi.org/10.1145/3488560.3498493

work page doi:10.1145/3488560.3498493 2022
[33]

In: Proceedings of the 20 23 ACM Con- ference on Fairness, Accountability, and Transparency, pp

Hamman, F., Chen, J., Dutta, S.: Can querying for bias leak prote cted attributes? achieving privacy with smooth sensitivity. In: Proceedings of the 20 23 ACM Con- ference on Fairness, Accountability, and Transparency, pp. 135 8–1368 (2023). https://doi.org/10.1145/3593013.3594086

work page doi:10.1145/3593013.3594086 2023
[34]

An Introduction To Compressive Sampling,

Cand` es, E.J., Wakin, M.B.: An introduction to compressive sam- pling. IEEE signal processing magazine 25(2), 21–30 (2008) https://doi.org/10.1109/MSP.2007.914731

work page doi:10.1109/msp.2007.914731 2008
[35]

In: 2011 IEEE Control and System Graduate Research Colloquium, pp

Navada, A., Ansari, A.N., Patil, S., Sonkamble, B.A.: Overview of use of decision tree algorithms in machine learning. In: 2011 IEEE Control and System Graduate Research Colloquium, pp. 37–42 (20 11). https://doi.org/10.1109/ICSGRC.2011.5991826

work page doi:10.1109/icsgrc.2011.5991826 2011
[36]

Journal of Machine Learning Research 18(70), 1–37 (2017)

Wang, T., Rudin, C., Doshi-Velez, F., Liu, Y., Klampﬂ, E., MacNeille, P.: A Bayesian Framework for Learning Rule Sets for Interpretable Clas siﬁcation. Journal of Machine Learning Research 18(70), 1–37 (2017)

work page 2017
[37]

EPRS in-depth analysis, 1–8 (2020)

Szczepa´ nski, M.: Is data the new oil? competition issues in the dig ital economy. EPRS in-depth analysis, 1–8 (2020)

work page 2020
[38]

https://archive.ics.uci.edu/ml/datasets/adult Accessed 2023-02-05

Kohavi, R., Becker, B.: UCI Machine Learning Repository: Adult D ata Set (2016). https://archive.ics.uci.edu/ml/datasets/adult Accessed 2023-02-05

work page 2016
[39]

https://archive.ics.uci.edu/ml/datasets/statlog+(german+credit+data) Accessed 2023-02-05

Hofmann, H.: Statlog (German Credit Data) Data Set (2013). https://archive.ics.uci.edu/ml/datasets/statlog+(german+credit+data) Accessed 2023-02-05

work page 2013
[40]

Liu, T., Huynh, N., and van der Schaar, M

Miller, G.A.: The magical number seven, plus or minus two: Some limits on our 42 capacity for processing information. Psychological Review 63(2), 81–97 (1956) https://doi.org/10.1037/h0043158 . Accessed 2023-07-11

work page doi:10.1037/h0043158 1956
[41]

Information F usion 58, 82–115 (2020) https://doi.org/10.1016/j.inﬀus.2019.12.012

Barredo Arrieta, A., D ´ ıaz-Rodr ´ ıguez, N., Del Ser, J., Bennet ot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., Chatila , R., Herrera, F.: Explainable Artiﬁcial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information F usion 58, 82–115 (2020) https://doi.org/10.1016...

work page doi:10.1016/j.in 2020
[42]

In: Proceedings of the Thirty-ninth Annual AC M Symposium on Theory of Computing, pp

Nissim, K., Raskhodnikova, S., Smith, A.: Smooth sensitivity and sa mpling in pri- vate data analysis. In: Proceedings of the Thirty-ninth Annual AC M Symposium on Theory of Computing, pp. 75–84. ACM, San Diego California USA (2 007). https://doi.org/10.1145/1250790.1250803

work page doi:10.1145/1250790.1250803
[43]

Springer, Cham (2017)

Mendel, J.M.: Uncertain Rule-Based Fuzzy Systems: Introduc- tion and New Directions, 2nd Edition. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-51370-6

work page doi:10.1007/978-3-319-51370-6 2017
[44]

Journal of Machine Lea rning Research 18(234), 1–78 (2018)

Angelino, E., Larus-Stone, N., Alabi, D., Seltzer, M., Rudin, C.: Lea rning certiﬁ- ably optimal rule lists for categorical data. Journal of Machine Lea rning Research 18(234), 1–78 (2018)

work page 2018
[45]

Southeast Europe Journal of soft computing 5(1) (2016) https://doi.org/10.21533/scjournal.v5i1.102

Yazgana, P., Kusakci, A.O.: A literature survey on association ru le min- ing algorithms. Southeast Europe Journal of soft computing 5(1) (2016) https://doi.org/10.21533/scjournal.v5i1.102

work page doi:10.21533/scjournal.v5i1.102 2016
[46]

Aytekin, C.: Neural Networks are Decision Trees. arXiv. arXiv:2 210.05189 [cs] (2022). http://arxiv.org/abs/2210.05189 Accessed 2023-06-08 43 A B C K A ¬A B ¬B C ¬C J K becomes Fig. A1 A schematic display of the process by which a binary tree that has non-binary splits can be converted into a binary tree for a binary decision process . The dotted lines . ...

work page arXiv 2022

[1] [1]

Procedia Computer Science 162, 503–513 (2019) https://doi.org/10.1016/j.procs.2019.12.017

Zhu, L., Qiu, D., Ergu, D., Ying, C., Liu, K.: A study on predicting loan d efault based on the random forest algorithm. Procedia Computer Science 162, 503–513 (2019) https://doi.org/10.1016/j.procs.2019.12.017

work page doi:10.1016/j.procs.2019.12.017 2019

[2] [2]

Science 187(4175), 398–404 (1975) https://doi.org/10.1126/science.187.4175.398

Bickel, P.J., Hammel, E.A., O’Connell, J.W.: Sex bias in graduate admission s: Data from berkeley: Measuring bias is harder than is usually assumed , and the evidence is sometimes contrary to expectation. Science 187(4175), 398–404 (1975) https://doi.org/10.1126/science.187.4175.398

work page doi:10.1126/science.187.4175.398 1975

[3] [3]

Big Data 5(2), 153–163 (2017) https://doi.org/10.1089/big.2016.0047

Chouldechova, A.: Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments. Big Data 5(2), 153–163 (2017) https://doi.org/10.1089/big.2016.0047 . Accessed 2023-01-13

work page doi:10.1089/big.2016.0047 2017

[4] [4]

Section: Digits (2015)

Barr, A.: Google Mistakenly Tags Black People as ‘Goril- las,’ Showing Limits of Algorithms. Section: Digits (2015). https://www.wsj.com/articles/BL-DGB-42522 Accessed 2023-02-04

work page 2015

[5] [5]

https://www.amnesty.org/en/documents/eur35/4686/2021/en/ Accessed 2023-02-04 37

Xenophobic machines: Discrimination through unregulated use of algorithms in the Dutch childcare beneﬁts scandal (2021). https://www.amnesty.org/en/documents/eur35/4686/2021/en/ Accessed 2023-02-04 37

work page 2021

[6] [6]

Oﬃcial Journal of the European Union L 119 , 1–88 (14-04-2016)

Regulation (eu) 2016/679 of the european parliament and of the council of 27 april 2016 on the protection of natural persons with regard to th e processing of personal data and on the free movement of such data, and repea ling directive 95/46/ec (general data protection regulation). Oﬃcial Journal of the European Union L 119 , 1–88 (14-04-2016)

work page 2016

[7] [7]

Oﬃcial Journal of the Europ ean Union COM/2021/206 ﬁnal , 1–107 (21-04-2021)

Proposal for a regulation of the european parliament and of the council laying down harmonised rules on artiﬁcial intelligence (artiﬁcial intelligence a ct) and amending certain union legislative acts. Oﬃcial Journal of the Europ ean Union COM/2021/206 ﬁnal , 1–107 (21-04-2021)

work page 2021

[8] [8]

https://algoritmes.overheid.nl/ Accessed 2023-02-04

Het Algoritmeregister van de Nederlandse overheid (2022). https://algoritmes.overheid.nl/ Accessed 2023-02-04

work page 2022

[9] [9]

Oﬃcial Journal of the European Union L 60/34 , 34–85 (04-02-2014)

Directive 2014/17/eu of the european parliament and of the cou ncil of 4 february 2014 on credit agreements for consumers relating to residential im movable prop- erty and amending directives 2008/48/ec and 2013/36/eu and reg ulation (eu). Oﬃcial Journal of the European Union L 60/34 , 34–85 (04-02-2014)

work page 2014

[10] [10]

The Guardian (2018)

Cadwalladr, C., Graham-Harrison, E.: Revealed: 50 million Faceboo k proﬁles har- vested for Cambridge Analytica in major data breach. The Guardian (2018). Chap. News. Accessed 2023-02-04

work page 2018

[11] [11]

https://www.upguard.com/breaches/facebook-user-data-lea k Accessed 2023-02-04

Losing Face: Two More Cases of Third-Party Facebook App Dat a Exposure |UpGuard (2019). https://www.upguard.com/breaches/facebook-user-data-lea k Accessed 2023-02-04

work page 2019

[12] [12]

Artiﬁcial Intelligence and Law 22(2), 175–209 (2014) 38 https://doi.org/10.1007/s10506-013-9152-0

Berendt, B., Preibusch, S.: Better decision support through e xploratory discrimination-aware data mining: foundations and empirical evi- dence. Artiﬁcial Intelligence and Law 22(2), 175–209 (2014) 38 https://doi.org/10.1007/s10506-013-9152-0

work page doi:10.1007/s10506-013-9152-0 2014

[13] [13]

https://www.propublica.org/article/machine-bias-risk-assessme nts-in-criminal-sentencing Accessed 2023-01-06

Mattu, S., Larson, J., Kirchner, L., Angwin, J.: Machine Bias (201 6). https://www.propublica.org/article/machine-bias-risk-assessme nts-in-criminal-sentencing Accessed 2023-01-06

work page 2023

[14] [14]

In: Proceed ings of the Inter- national Workshop on Software Fairness, pp

Verma, S., Rubin, J.: Fairness deﬁnitions explained. In: Proceed ings of the Inter- national Workshop on Software Fairness, pp. 1–7. ACM, Gothenb urg Sweden (2018). https://doi.org/10.1145/3194770.3194776

work page doi:10.1145/3194770.3194776 2018

[15] [15]

In: Proceedings of the 3rd Innovations in Theoretica l Computer Science Conference On - ITCS ’12, pp

Dwork, C., Hardt, M., Pitassi, T., Reingold, O., Zemel, R.: Fairness t hrough awareness. In: Proceedings of the 3rd Innovations in Theoretica l Computer Science Conference On - ITCS ’12, pp. 214–226. ACM Press, Cambr idge, Massachusetts (2012). https://doi.org/10.1145/2090236.2090255

work page doi:10.1145/2090236.2090255 2012

[16] [16]

FEDERAL REGIST ER 44(43) (01-03-1979)

Questions and answers to clarify and provide a common interpre tation of the uni- form guidelines on employee selection procedures. FEDERAL REGIST ER 44(43) (01-03-1979)

work page 1979

[17] [17]

In: Advances in Neural Information Processing Sy stems, vol

Hardt, M., Price, E., Price, E., Srebro, N.: Equality of Opportunit y in Super- vised Learning. In: Advances in Neural Information Processing Sy stems, vol. 29, pp. 3315–3323. Curran Associates, Inc., Red Hook, New York (2016). https://proceedings.neurips.cc/paper/2016/hash/9d26823 67c3935defcb1f9e247a97c0d- Abstract.html Accessed 2023-01-10

work page 2016

[18] [18]

Machine Le arning 106, 1039–1082 (2017) https://doi.org/10.1007/s10994-017-5633-9

Bertsimas, D., Dunn, J.: Optimal classiﬁcation trees. Machine Le arning 106, 1039–1082 (2017) https://doi.org/10.1007/s10994-017-5633-9

work page doi:10.1007/s10994-017-5633-9 2017

[19] [19]

In: Machin e Learning: ECML-94: European Conference on Machine Learning Catania, Ita ly, April 6–8, 1994 Proceedings 7, pp

Oliver, J.J., Hand, D.: Averaging over decision stumps. In: Machin e Learning: ECML-94: European Conference on Machine Learning Catania, Ita ly, April 6–8, 1994 Proceedings 7, pp. 231–241 (1994). https://doi.org/10.1007/3-540-57868-4 39 61 . Springer

work page doi:10.1007/3-540-57868-4 1994

[20] [20]

Molnar, C.: Interpretable Machine Learning, 2nd edn. (2022). https://christophm.github.io/interpretable-ml-book

work page 2022

[21] [21]

Borisov, T

Borisov, V., Leemann, T., Seßler, K., Haug, J., Pawelczyk, M., Kas - neci, G.: Deep neural networks and tabular data: A survey. IEEE Transactions on Neural Networks and Learning Systems (2022) https://doi.org/10.1109/TNNLS.2022.3229161

work page doi:10.1109/tnnls.2022.3229161 2022

[22] [22]

Differential Privacy,

Dwork, C.: Diﬀerential privacy. In: 33rd International Colloqu ium, ICALP 2006 on Automata, Languages and Programming, pp. 1–12. Spring er, Berlin, Heidelberg (2006). https://doi.org/10.1007/11787006 1

work page doi:10.1007/11787006 2006

[23] [23]

2014.The Algorithmic Foundations of Differential Privacy

Dwork, C., Roth, A.: The Algorithmic Foundations of Diﬀerential P rivacy. Foun- dations and Trends in Theoretical Computer Science 9(3-4), 211–407 (2013) https://doi.org/10.1561/0400000042 . Accessed 2023-01-27

work page doi:10.1561/0400000042 2013

[24] [24]

In: 48th Annual IEEE Symposium on Foundations of Computer Science (FOCS ’07), pp

McSherry, F., Talwar, K.: Mechanism design via diﬀerential privac y. In: 48th Annual IEEE Symposium on Foundations of Computer Science (FOCS ’07), pp. 94–103 (2007). https://doi.org/10.1109/FOCS.2007.66 . IEEE

work page doi:10.1109/focs.2007.66 2007

[25] [25]

In: 2010 IEEE International Conference on Data Mining , pp

Kamiran, F., Calders, T., Pechenizkiy, M.: Discrimination Aware Dec ision Tree Learning. In: 2010 IEEE International Conference on Data Mining , pp. 869–874. IEEE, Sydney, Australia (2010). https://doi.org/10.1109/ICDM.2010.50

work page doi:10.1109/icdm.2010.50 2010

[26] [26]

Springer, Berlin, Heidelber g (1999)

Ausiello, G., Crescenzi, P., Gambosi, G., Kann, V., Marchetti-Spac camela, A., Protasi, M.: Complexity and Approximation: Combinatorial Optimizatio n Prob- lems and Their Approximability Properties. Springer, Berlin, Heidelber g (1999). https://doi.org/10.1007/978-3-642-58412-1 40

work page doi:10.1007/978-3-642-58412-1 1999

[27] [27]

In: Koyejo, S., Mohamed, S., Agarwal, A., B elgrave, D., Cho, K., Oh, A

Linden, J., Weerdt, M., Demirovi´ c, E.: Fair and optimal decision trees: A dynamic programming approach. In: Koyejo, S., Mohamed, S., Agarwal, A., B elgrave, D., Cho, K., Oh, A. (eds.) Advances in Neural Information Processin g Systems, vol. 35, pp. 38899–38911. Curran Associates, Inc., Red Hook, Ne w York (2022). https://proceedings.neurips.cc/paper ﬁle...

work page 2022

[28] [28]

Jo, N., Aghaei, S., Benson, J., G´ omez, A., Vayanos, P.: Learning Opti- mal Fair Classiﬁcation Trees. arXiv. arXiv:2201.09932 [cs, math] (20 22). http://arxiv.org/abs/2201.09932 Accessed 2022-12-08

work page arXiv 2022

[29] [29]

In: 2015 IEEE 28th International Symposium on Computer-Based Medical Systems, pp

Mohammed, N., Barouti, S., Alhadidi, D., Chen, R.: Secure and Priva te Manage- ment of Healthcare Databases for Data Mining. In: 2015 IEEE 28th International Symposium on Computer-Based Medical Systems, pp. 191–196. IE EE, Sao Carlos, Brazil (2015). https://doi.org/10.1109/CBMS.2015.54

work page doi:10.1109/cbms.2015.54 2015

[30] [30]

ACM Computing Surveys 52(4), 1–33 (2020) https://doi.org/10.1145/3337064

Fletcher, S., Islam, M.Z.: Decision Tree Classiﬁcation with Diﬀeren- tial Privacy: A Survey. ACM Computing Surveys 52(4), 1–33 (2020) https://doi.org/10.1145/3337064

work page doi:10.1145/3337064 2020

[31] [31]

In: Advances in Neural Information Processing System s, vol

Lahoti, P., Beutel, A., Chen, J., Lee, K., Prost, F., Thain, N., Wang , X., Chi, E.: Fairness without Demographics through Adversarially Reweig hted Learning. In: Advances in Neural Information Processing System s, vol. 33, pp. 728–740. Curran Associates, Inc., Red Hook, New York (2 020). https://proceedings.neurips.cc/paper/2020/hash/07fc15c 9d169ee48573...

work page 2020

[32] [32]

In: Proceedings of the Fifteenth ACM International Conference on Web Search And Data Mining, pp

Zhao, T., Dai, E., Shu, K., Wang, S.: Towards Fair Classiﬁers Withou t Sensitive Attributes: Exploring Biases in Related Features. In: Proceedings of the Fifteenth ACM International Conference on Web Search And Data Mining, pp. 1433–1442. 41 ACM, Virtual Event AZ USA (2022). https://doi.org/10.1145/3488560.3498493

work page doi:10.1145/3488560.3498493 2022

[33] [33]

In: Proceedings of the 20 23 ACM Con- ference on Fairness, Accountability, and Transparency, pp

Hamman, F., Chen, J., Dutta, S.: Can querying for bias leak prote cted attributes? achieving privacy with smooth sensitivity. In: Proceedings of the 20 23 ACM Con- ference on Fairness, Accountability, and Transparency, pp. 135 8–1368 (2023). https://doi.org/10.1145/3593013.3594086

work page doi:10.1145/3593013.3594086 2023

[34] [34]

An Introduction To Compressive Sampling,

Cand` es, E.J., Wakin, M.B.: An introduction to compressive sam- pling. IEEE signal processing magazine 25(2), 21–30 (2008) https://doi.org/10.1109/MSP.2007.914731

work page doi:10.1109/msp.2007.914731 2008

[35] [35]

In: 2011 IEEE Control and System Graduate Research Colloquium, pp

Navada, A., Ansari, A.N., Patil, S., Sonkamble, B.A.: Overview of use of decision tree algorithms in machine learning. In: 2011 IEEE Control and System Graduate Research Colloquium, pp. 37–42 (20 11). https://doi.org/10.1109/ICSGRC.2011.5991826

work page doi:10.1109/icsgrc.2011.5991826 2011

[36] [36]

Journal of Machine Learning Research 18(70), 1–37 (2017)

Wang, T., Rudin, C., Doshi-Velez, F., Liu, Y., Klampﬂ, E., MacNeille, P.: A Bayesian Framework for Learning Rule Sets for Interpretable Clas siﬁcation. Journal of Machine Learning Research 18(70), 1–37 (2017)

work page 2017

[37] [37]

EPRS in-depth analysis, 1–8 (2020)

Szczepa´ nski, M.: Is data the new oil? competition issues in the dig ital economy. EPRS in-depth analysis, 1–8 (2020)

work page 2020

[38] [38]

https://archive.ics.uci.edu/ml/datasets/adult Accessed 2023-02-05

Kohavi, R., Becker, B.: UCI Machine Learning Repository: Adult D ata Set (2016). https://archive.ics.uci.edu/ml/datasets/adult Accessed 2023-02-05

work page 2016

[39] [39]

https://archive.ics.uci.edu/ml/datasets/statlog+(german+credit+data) Accessed 2023-02-05

Hofmann, H.: Statlog (German Credit Data) Data Set (2013). https://archive.ics.uci.edu/ml/datasets/statlog+(german+credit+data) Accessed 2023-02-05

work page 2013

[40] [40]

Liu, T., Huynh, N., and van der Schaar, M

Miller, G.A.: The magical number seven, plus or minus two: Some limits on our 42 capacity for processing information. Psychological Review 63(2), 81–97 (1956) https://doi.org/10.1037/h0043158 . Accessed 2023-07-11

work page doi:10.1037/h0043158 1956

[41] [41]

Information F usion 58, 82–115 (2020) https://doi.org/10.1016/j.inﬀus.2019.12.012

Barredo Arrieta, A., D ´ ıaz-Rodr ´ ıguez, N., Del Ser, J., Bennet ot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., Chatila , R., Herrera, F.: Explainable Artiﬁcial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information F usion 58, 82–115 (2020) https://doi.org/10.1016...

work page doi:10.1016/j.in 2020

[42] [42]

In: Proceedings of the Thirty-ninth Annual AC M Symposium on Theory of Computing, pp

Nissim, K., Raskhodnikova, S., Smith, A.: Smooth sensitivity and sa mpling in pri- vate data analysis. In: Proceedings of the Thirty-ninth Annual AC M Symposium on Theory of Computing, pp. 75–84. ACM, San Diego California USA (2 007). https://doi.org/10.1145/1250790.1250803

work page doi:10.1145/1250790.1250803

[43] [43]

Springer, Cham (2017)

Mendel, J.M.: Uncertain Rule-Based Fuzzy Systems: Introduc- tion and New Directions, 2nd Edition. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-51370-6

work page doi:10.1007/978-3-319-51370-6 2017

[44] [44]

Journal of Machine Lea rning Research 18(234), 1–78 (2018)

Angelino, E., Larus-Stone, N., Alabi, D., Seltzer, M., Rudin, C.: Lea rning certiﬁ- ably optimal rule lists for categorical data. Journal of Machine Lea rning Research 18(234), 1–78 (2018)

work page 2018

[45] [45]

Southeast Europe Journal of soft computing 5(1) (2016) https://doi.org/10.21533/scjournal.v5i1.102

Yazgana, P., Kusakci, A.O.: A literature survey on association ru le min- ing algorithms. Southeast Europe Journal of soft computing 5(1) (2016) https://doi.org/10.21533/scjournal.v5i1.102

work page doi:10.21533/scjournal.v5i1.102 2016

[46] [46]

Aytekin, C.: Neural Networks are Decision Trees. arXiv. arXiv:2 210.05189 [cs] (2022). http://arxiv.org/abs/2210.05189 Accessed 2023-06-08 43 A B C K A ¬A B ¬B C ¬C J K becomes Fig. A1 A schematic display of the process by which a binary tree that has non-binary splits can be converted into a binary tree for a binary decision process . The dotted lines . ...

work page arXiv 2022