Path-Specific Counterfactual Fairness Assessment

Description

Path-specific counterfactual fairness assessment tests whether a protected attribute, such as sex or race, influences a decision along routes that the assessor judges unfair. The assessor first draws a causal graph, a diagram of which variables influence which, and marks each route from the protected attribute to the decision as acceptable or unacceptable. In a classic admissions example, women applied more often to departments with low admission rates, so the effect of sex through department choice was treated as acceptable, while any direct effect of sex was not. The technique then estimates how the decision would change if the protected attribute were switched, counting only the effect that travels along the unacceptable routes. A large estimate points to unfair influence along those routes, and the estimate is only as sound as the graph and the choice of routes.

Example Use Cases

Fairness

A company auditing its CV-screening model draws a causal graph from gender to the shortlisting decision. Routes through qualifications are marked acceptable, while the direct route and the route through career-break length are marked unacceptable. The estimated effect along the unacceptable routes tells the audit team whether the model disadvantages women beyond what job-relevant qualifications explain.

A research team auditing ventilation decisions in intensive care uses recorded patient data to estimate how much of the effect of patient race on starting invasive ventilation passes through pulse-oximeter readings, which tend to overstate oxygen levels in dark-skinned patients, and how much through other routes.

Limitations

Which routes count as unacceptable is a value judgement the assessor must make and defend, and different reasonable choices give different verdicts on the same model. The technique does not settle that question. It only measures the effect along the routes it is given.
The causal graph comes from domain knowledge and cannot be fully checked against observational data. If the graph has a wrong arrow or omits a relevant variable, the assessment can certify an unfair model as compliant, and nothing inside the method flags the error.
Path-specific effects are often not identifiable from observational data alone. When unobserved confounders sit between the protected attribute and the mediating variables or the outcome, the effect along a specific route cannot be uniquely recovered without further parametric assumptions, so the result becomes a bound rather than a point estimate.
Estimating the effect along a route means fitting a model for each variable on it. Outside linear or low-dimensional settings the calculation is hard and relies on approximations, and the answer is only as good as the fitted model's match to the data.
An average effect can hide opposing individual effects. A population-level result does not establish the paper's individual counterfactual criterion for every person.

Resources

Research Papers

Path-Specific Counterfactual Fairness

Silvia Chiappa•Jul 17, 2019

We consider the problem of learning fair decision systems from data in which a sensitive attribute might affect the decision along both fair and unfair pathways. We introduce a counterfactual approach to disregard effects along unfair pathways that does not incur in the same loss of individual-specific information as previous approaches. Our method corrects observations adversely affected by the sensitive attribute, and uses these to form a decision. We leverage recent developments in deep learning and approximate inference to develop a VAE-type method that is widely applicable to complex nonlinear models.

A Causal Framework for Discovering and Removing Direct and Indirect Discrimination

Lu Zhang, Yongkai Wu, and Xintao Wu•Aug 1, 2017

In this paper, we investigate the problem of discovering both direct and indirect discrimination from the historical data, and removing the discriminatory effects before the data is used for predictive analysis (e.g., building classifiers). The main drawback of existing methods is that they cannot distinguish the part of influence that is really caused by discrimination from all correlated influences. In our approach, we make use of the causal network to capture the causal structure of the data. Then we model direct and indirect discrimination as the path-specific effects, which accurately identify the two types of discrimination as the causal effects transmitted along different paths in the network. Based on that, we propose an effective algorithm for discovering direct and indirect discrimination, as well as an algorithm for precisely removing both types of discrimination while retaining good data utility. Experiments using the real dataset show the effectiveness of our approaches.

Causal Fairness Analysis: A Causal Toolkit for Fair Machine Learning

Drago Plečko and Elias Bareinboim•Jan 31, 2024

Decision-making systems based on AI and machine learning have been used throughout a wide range of real-world scenarios, including healthcare, law enforcement, education, and finance. It is no longer far-fetched to envision a future where autonomous systems will drive entire business decisions and, more broadly, support large-scale decision-making infrastructure to solve society’s most challenging problems. Issues of unfairness and discrimination are pervasive when decisions are being made by humans, and remain (or are potentially amplified) when decisions are made using machines with little transparency, accountability, and fairness. In this monograph, we introduce a framework for causal fairness analysis with the intent of filling in this gap, i.e., understanding, modeling, and possibly solving issues of fairness in decision-making settings. The main insight of our approach will be to link the quantification of the disparities present in the observed data with the underlying, often unobserved, collection of causal mechanisms that generate the disparity in the first place, a challenge we call the Fundamental Problem of Causal Fairness Analysis (FPCFA). In order to solve the FPCFA, we study the problem of decomposing variations and empirical measures of fairness that attribute such variations to structural mechanisms and different units of the population. Our effort culminates in the Fairness Map, the first systematic attempt to organize and explain the relationship between various criteria found in the literature. Finally, we study which causal assumptions are minimally needed for performing causal fairness analysis and propose the Fairness Cookbook, which allows one to assess the existence of disparate impact and disparate treatment.

Software Packages

faircause: Causal Fairness Analysis (dplecko/CFA)

Sep 22, 2026

Causal Fairness Analysis

Tutorials

Causal Fairness Analysis vignette: Census bias detection

Documentations

Causal Fairness Analysis — faircause vignettes

Related Techniques

Name	Description	Assurance Goals
Counterfactual Fairness Assessment	Counterfactual Fairness Assessment tests whether a model would have made the same predictions for a person had their protected attribute, such as race or sex, been different, with everything that attribute influences changed accordingly (Kusner et al., 2017). The assessor first draws a causal graph, a diagram of which variables cause which, covering the protected attribute, the model's inputs and unobserved background factors, and fits it to data. For each person, the fitted graph is used to infer their background factors, switch the attribute to another value and recompute every input it influences, which gives a counterfactual version of that person. The model is run on both versions, and the gap between the two predictions, person by person or as a distance between the two sets of predictions, is the evidence for or against counterfactual fairness under the assumed graph. Every causal effect of the attribute counts as unfair, including effects passing through inputs such as test scores, which separates this criterion from versions that allow some pathways.	Fairness
Causal Discrimination Testing	Causal discrimination testing checks whether a system gives different outputs to otherwise identical inputs because of a protected attribute, by changing only that attribute, or a name, term or other proxy that signals it, and measuring how often or how much the output changes. Galhotra, Brun and Meliou (2017), who called the approach fairness testing, generate inputs from a description of the valid input fields, make copies of each that differ only in the protected attribute (for example, recorded sex), run them all through the system as a black box, and report the causal discrimination score: the share of inputs whose decision changes, sampled until it is known within a chosen margin of error and confidence level. For text, the same test substitutes identity terms or names in otherwise unchanged sentences, such as one religion or ethnicity for another, and reports the average change in the model's score across the substitutions, which Garg et al. (2019) call the counterfactual token fairness gap. Because every other part of the input is held fixed, the test needs no model of how the attribute influences other variables: it measures whether the system itself responds to the attribute, or to the chosen proxy. The output is a score per attribute and, if the tester logs them, the input pairs whose outputs differed.	Fairness
Demographic Parity Assessment	Demographic Parity Assessment checks whether a model gives a favourable outcome, such as a loan approval or an interview invitation, to each demographic group at the same rate. It compares those rates without adjusting for how many people in each group would qualify. Statistical Parity Difference is the gap between group rates (a signed difference between two named groups in AIF360, the highest rate minus the lowest across all groups in Fairlearn), and the Disparate Impact ratio divides the lowest rate by the highest. A ratio below 0.8 is a common warning sign, following the four-fifths rule used by US employment regulators. A gap shows that outcomes differ between groups but not why, and equal rates do not show that individuals were treated fairly.	Fairness