A Neurosymbolic Approach to Counterfactual Fairness

Xenia Heilmann, Chiara Manganini, Mattia Cerrato, Vaishak Belle
Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, PMLR 284:1004-1025, 2025.

Abstract

Integrating fairness into machine learning models has been an important consideration for the last decade. Here, neurosymbolic models offer a valuable opportunity, as they allow the specification of symbolic, logical constraints that are often guaranteed to be satisfied. However, research on neurosymbolic applications to algorithmic fairness is still in an early stage. With our work, we bridge this gap by integrating counterfactual fairness into the neurosymbolic framework of Logic Tensor Networks (LTN). We use LTN to express accuracy and counterfactual fairness constraints in first-order logic and employ them to achieve desirable levels of both performance and fairness at training time. Our approach is agnostic to the underlying causal model and data generation technique; as such, it may be easily integrated into existing pipelines that generate and extract counterfactual examples. We show, through concrete examples on three real-world datasets, that logical reasoning about counterfactual fairness has some important advantages, among which its intrinsic interpretability, and its flexibility in handling subgroup fairness. Compared to three recent methodologies in counterfactual fairness, our experiments show that a neurosymbolic, LTN-based approach attains better levels of counterfactual fairness.

Cite this Paper

BibTeX
@InProceedings{pmlr-v284-heilmann25a, title = {A Neurosymbolic Approach to Counterfactual Fairness}, author = {Heilmann, Xenia and Manganini, Chiara and Cerrato, Mattia and Belle, Vaishak}, booktitle = {Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning}, pages = {1004--1025}, year = {2025}, editor = {H. Gilpin, Leilani and Giunchiglia, Eleonora and Hitzler, Pascal and van Krieken, Emile}, volume = {284}, series = {Proceedings of Machine Learning Research}, month = {08--10 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v284/main/assets/heilmann25a/heilmann25a.pdf}, url = {https://proceedings.mlr.press/v284/heilmann25a.html}, abstract = {Integrating fairness into machine learning models has been an important consideration for the last decade. Here, neurosymbolic models offer a valuable opportunity, as they allow the specification of symbolic, logical constraints that are often guaranteed to be satisfied. However, research on neurosymbolic applications to algorithmic fairness is still in an early stage. With our work, we bridge this gap by integrating counterfactual fairness into the neurosymbolic framework of Logic Tensor Networks (LTN). We use LTN to express accuracy and counterfactual fairness constraints in first-order logic and employ them to achieve desirable levels of both performance and fairness at training time. Our approach is agnostic to the underlying causal model and data generation technique; as such, it may be easily integrated into existing pipelines that generate and extract counterfactual examples. We show, through concrete examples on three real-world datasets, that logical reasoning about counterfactual fairness has some important advantages, among which its intrinsic interpretability, and its flexibility in handling subgroup fairness. Compared to three recent methodologies in counterfactual fairness, our experiments show that a neurosymbolic, LTN-based approach attains better levels of counterfactual fairness.} }
Endnote
%0 Conference Paper %T A Neurosymbolic Approach to Counterfactual Fairness %A Xenia Heilmann %A Chiara Manganini %A Mattia Cerrato %A Vaishak Belle %B Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning %C Proceedings of Machine Learning Research %D 2025 %E Leilani H. Gilpin %E Eleonora Giunchiglia %E Pascal Hitzler %E Emile van Krieken %F pmlr-v284-heilmann25a %I PMLR %P 1004--1025 %U https://proceedings.mlr.press/v284/heilmann25a.html %V 284 %X Integrating fairness into machine learning models has been an important consideration for the last decade. Here, neurosymbolic models offer a valuable opportunity, as they allow the specification of symbolic, logical constraints that are often guaranteed to be satisfied. However, research on neurosymbolic applications to algorithmic fairness is still in an early stage. With our work, we bridge this gap by integrating counterfactual fairness into the neurosymbolic framework of Logic Tensor Networks (LTN). We use LTN to express accuracy and counterfactual fairness constraints in first-order logic and employ them to achieve desirable levels of both performance and fairness at training time. Our approach is agnostic to the underlying causal model and data generation technique; as such, it may be easily integrated into existing pipelines that generate and extract counterfactual examples. We show, through concrete examples on three real-world datasets, that logical reasoning about counterfactual fairness has some important advantages, among which its intrinsic interpretability, and its flexibility in handling subgroup fairness. Compared to three recent methodologies in counterfactual fairness, our experiments show that a neurosymbolic, LTN-based approach attains better levels of counterfactual fairness.
APA
Heilmann, X., Manganini, C., Cerrato, M. & Belle, V.. (2025). A Neurosymbolic Approach to Counterfactual Fairness. Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, in Proceedings of Machine Learning Research 284:1004-1025 Available from https://proceedings.mlr.press/v284/heilmann25a.html.

Related Material