COMPAS recidivism predictions show racial bias against African Americans, study finds.
problem Racial bias in recidivism prediction algorithms.
method Causal analysis using FACT, a fairness measure grounded in causal inference.
result COMPAS shows racial bias against African American defendants, robust to unmeasured confounding.
When might human input help (or not) when assessing risk in fairness domains? Dressel and Farid (2018) asked Mechanical Turk workers to evaluate a subset of defendants in the ProPublica COMPAS data for risk of recidivism, and concluded that COMPAS predictions were no more accurate or fair than predictions made by human…
Develops tools to audit ML models for bias and unfairness.
problem Auditing ML models for individual bias and unfairness.
method Formalizes the task as an optimization problem and develops inferential tools for the optimal value.
result Demonstrates the utility of tools in revealing biases in COMPAS recidivism prediction instrument.
Black-box risk scoring models permeate our lives, yet are typically proprietary or opaque. We propose Distill-and-Compare, a model distillation and comparison approach to audit such models. To gain insight into black-box models, we treat them as teachers, training transparent student models to mimic the risk scores ass…
Advocates focusing on utility functions to avoid unfair outcomes.
problem Unfair outcomes from fairness criteria in optimizing policies.
method Defines value of information fairness and proposes modifying utility functions.
result Value of information fairness leads to better answers than existing fairness notions.
Recidivism prediction scores are used across the USA to determine sentencing and supervision for hundreds of thousands of inmates. One such generator of recidivism prediction scores is Northpointe's Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) score, used in states like California and F…
We offer a graphical interpretation of unfairness in a dataset as the presence of an unfair causal path in the causal Bayesian network representing the data-generation mechanism. We use this viewpoint to revisit the recent debate surrounding the COMPAS pretrial risk assessment tool and, more generally, to point out tha…
The paper reconciles two conflicting fairness criteria in algorithmic risk scores.
problem How to reconcile calibration and equal error rates in algorithmic risk scores.
method Derive necessary and sufficient conditions for existence of calibrated scores achieving equal error rates, then present an algorithm to find the most accurate score subject to both criteria.
result The method can eliminate error disparities while maintaining calibration and improve profit in credit lending.
Interpretable ML models predict recidivism as well as non-interpretable methods and are more fair.
problem Improving recidivism prediction models for fairness and interpretability.
method Trained interpretable ML models on two recidivism datasets, compared to existing methods, and analyzed fairness.
result Interpretable ML models can predict recidivism as well as non-interpretable methods and are more fair.
EL framework certifies and flags bias in ML models without distributional assumptions.
problem Systematic performance disparities across sensitive subpopulations in ML models.
method Empirical likelihood-based approach for non-parametric fairness auditing.
result EL framework outperforms bootstrap methods in certification and subpopulation discovery.
We propose definitions of fairness in machine learning and artificial intelligence systems that are informed by the framework of intersectionality, a critical lens arising from the Humanities literature which analyzes how interlocking systems of power and oppression affect individuals along overlapping dimensions inclu…
Methods for interpreting machine learning black-box models increase the outcomes' transparency and in turn generates insight into the reliability and fairness of the algorithms. However, the interpretations themselves could contain significant uncertainty that undermines the trust in the outcomes and raises concern abo…
We present the design and implementation of a custom discrete optimization technique for building rule lists over a categorical feature space. Our algorithm produces rule lists with optimal training performance, according to the regularized empirical risk, with a certificate of optimality. By leveraging algorithmic bou…
Framework improves fairness in machine learning models using adversarial techniques.
problem Unfair decision processes in machine learning models.
method Two-step framework involving fairness optimization and evasion attacks.
result Framework significantly improves fairness metrics on tested datasets.
Intersectionality is a framework that analyzes how interlocking systems of power and oppression affect individuals along overlapping dimensions including race, gender, sexual orientation, class, and disability. Intersectionality theory therefore implies it is important that fairness in artificial intelligence systems b…
We revisit the notion of individual fairness proposed by Dwork et al. A central challenge in operationalizing their approach is the difficulty in eliciting a human specification of a similarity metric. In this paper, we propose an operationalization of individual fairness that does not rely on a human specification of …
The paper tackles fairness in forecasting and learning linear dynamical systems.
problem Under-representation bias in training data for multiple subgroups.
method Introducing subgroup-fair and instant-fair learning of LDS from multiple trajectories of varying lengths, using hierarchies of convexifications of non-commutative polynomial optimisation problems.
result Empirical results show both the beneficial impact of fairness considerations on statistical performance and encouraging effects of exploiting sparsity on run time.
We present a novel subset scan method to detect if a probabilistic binary classifier has statistically significant bias -- over or under predicting the risk -- for some subgroup, and identify the characteristics of this subgroup. This form of model checking and goodness-of-fit test provides a way to interpretably detec…
A new method assesses algorithmic fairness using game theory.
problem Evaluating algorithmic fairness without proprietary data.
method Cohort Shapley value, a game-theoretic approach.
result Identifies individual impact of protected attributes.
Study federates measurement of demographic disparities from quantile sketches.
problem Misalignment of fairness goals with siloed data collection and privacy regulations.
method Federated auditing of demographic parity through score distributions, using Wasserstein--Frechet variance and quantile summaries.
result Proposes a one-shot, communication-efficient protocol to estimate global disparity and its decomposition.
Maximal correlation framework improves fairness in machine learning algorithms.
problem Ensuring fairness in machine learning algorithms.
method Introducing maximal correlation framework for fairness constraints and deriving regularizers.
result The approach provides smooth performance-fairness tradeoff curves and competitive performance.
This paper examines the stability of learned explanations for black-box predictions via model distillation with decision trees. One approach to intelligibility in machine learning is to use an understandable `student' model to mimic the output of an accurate `teacher'. Here, we consider the use of regression trees as a…
In the context of machine learning, disparate impact refers to a form of systematic discrimination whereby the output distribution of a model depends on the value of a sensitive attribute (e.g., race or gender). In this paper, we propose an information-theoretic framework to analyze the disparate impact of a binary cla…
We consider settings in which the right notion of fairness is not captured by simple mathematical definitions (such as equality of error rates across groups), but might be more complex and nuanced and thus require elicitation from individual or collective stakeholders. We introduce a framework in which pairs of individ…
Neural Additive Models combine neural nets with interpretable models for high stakes tasks.
problem Inability to understand how neural networks make decisions.
method Combines neural nets with generalized additive models to create Neural Additive Models (NAMs).
result NAMs are more accurate than intelligible models and as accurate as state-of-the-art generalized additive models.
Paper shows LIME and SHAP explanations can be fooled by adversarial attacks.
problem Reliability of post hoc explanation methods like LIME and SHAP.
method Adversarial attack technique to hide classifier biases.
result Popular explanation techniques can be fooled into generating biased explanations.
New classifiers ensure fairness by adjusting a base classifier's operating characteristics.
problem Ensuring fairness in binary classification with multiple group constraints.
method Intervening directly on a base classifier's operating characteristics using group-wise ROC convex hulls and post-processing.
result Methods satisfy multiple fairness constraints (DP, EO, PP) with minimal interventions and near-oracle accuracy.
LDA-XGB1 balances fairness and accuracy in lending models.
problem Fair lending practices and model interpretability in binary classification.
method Biobjective optimization using binning and information value, leveraging XGBoost.
result Achieves effective balance between accuracy, fairness, and interpretability.
Study optimizes fairness in predictive models by balancing utility and separation.
problem Balancing fairness and utility in predictive models.
method Information-theoretic approach using conditional mutual information (CMI).
result Reduces separation violations while maintaining or improving utility.
Paper exposes vulnerabilities in interpreting machine learning models using adversarial attacks on PD plots.
problem Vulnerability of permutation-based interpretation methods, particularly PD plots, to adversarial attacks.
method Adversarial framework to manipulate black-box models and produce deceptive PD plots.
result It is possible to hide discriminatory behaviors in machine learning models through interpretation tools like PD plots.
New method to understand incentives from complex models.
problem Understanding how complex models incentivize actions.
method Formulated as a Markov Decision Process (MDP) and solved using MDP tools.
result Identifies optimal actions to maximize model output.
A popular methodology for building binary decision-making classifiers in the presence of imperfect information is to first construct a non-binary "scoring" classifier that is calibrated over all protected groups, and then to post-process this score to obtain a binary decision. We study the feasibility of achieving vari…
This master's thesis discusses an important issue regarding how algorithmic decision making (ADM) is used in crime forecasting. In America forecasting tools are widely used by judiciary systems for making decisions about risk offenders based on criminal justice for risk offenders. By making use of such tools, the judic…
Proposes a new method for fairness in machine learning with multiple protected attributes.
problem Ensuring fairness in machine learning models with continuous and multiple protected attributes.
method Distance covariance regularisation framework to mitigate association between model predictions and protected attributes.
result Demonstrates effectiveness in mitigating fairness gerrymandering in regression tasks.
Big data sets must be carefully partitioned into statistically similar data subsets that can be used as representative samples for big data analysis tasks. In this paper, we propose the random sample partition (RSP) data model to represent a big data set as a set of non-overlapping data subsets, called RSP data blocks,…
A new method for handling imbalanced big data using ensembles and smart data.
problem Imbalanced data distribution in big data scenarios.
method Smart Data driven Decision Trees Ensemble (SD_DeTE) methodology.
result SD_DeTE outperforms Random Forest in handling imbalanced binary classification problems in big data.
Prevents sensitive data generation in diffusion models using labeled and unlabeled data.
problem Generating sensitive data in diffusion models using unlabeled data.
method Positive-Unlabeled Diffusion Models, approximating ELBO with labeled and unlabeled data.
result Prevents the generation of sensitive data without compromising image quality.
Study reveals Data Shapley's inconsistent performance in data selection tasks.
problem Inconsistency of Data Shapley's performance in data selection across different settings.
method Hypothesis testing framework and identification of utility functions.
result Data Shapley's performance is no better than random selection without specific constraints.
PRRO generates synthetic tabular data that improves SL performance and class distribution.
problem Low SL utility of synthetic data due to class imbalance and overlooked data relationships.
method Data pruning and column reordering to optimize SL utility.
result Synthetic data generated with PRRO enhances predictive performance and class distribution.
Defines data science as a natural ecosystem with challenges and missions.
problem Challenges and missions in data science due to 5D complexities and data life cycle phases.
method Systemic and data-centric view of data science as a fusion of data universe and its challenges, formalizing a general-purpose architecture.
result Essential data science as a natural ecosystem integrating specific disciplines and high-impact applications.
Synthetic data enhances analytics but requires careful volume management.
problem Accuracy of statistical methods on synthetic data vs. raw data.
method Synthetic Data Generation for Analytics framework using tabular diffusion models.
result Error rate decreases with more synthetic data but may stabilize or increase.
This paper introduces C-DSL to improve data mining outcomes by considering context.
problem Data collection ambiguities, data imbalance, hidden biases, lack of domain info, and data incompleteness.
method Developed Context-Driven Data Science Lifecycle (C-DSL) to address data quality issues.
result Tangible improvements to data mining outcomes were achieved through C-DSL.
Proposes using probabilistic models for privacy-preserving synthetic data.
problem Designing high-quality synthetic data for privacy preservation.
method Formulate the problem through probabilistic modelling, choosing a model for the data.
result Statistical discoveries can be reliably reproduced from synthetic data.
New test ensures quality of shared data in machine learning.
problem Ensuring quality of external data in machine learning tasks.
method Distribution-free two-sample testing procedures grounded in conformal outlier detection.
result Identifies valuable external data agents for model personalization.
Paper creates fair synthetic data ensuring equal predictions across sensitive attributes.
problem Ensuring fair predictions across sensitive attributes in synthetic data.
method Equalizing target probability distributions across sensitive attributes in synthetic data generation.
result Synthetic data provides strong fair predictions, equal across all thresholds.
DPA preserves data distribution in reduced dimensions.
problem Loss of data distribution in dimension reduction.
method DPA combines encoder and decoder to match data distribution.
result DPA successfully reconstructs data distribution.
Efficient synthetic data generation improves model performance on tabular data.
problem Improving model robustness and performance with scarce or low-quality data.
method Hardness characterization to identify high-value training points, generating synthetic data only from these points.
result Synthetic data generated from hardest points outperforms non-targeted methods on tabular datasets.
For most problems in science and engineering we can obtain data sets that describe the observed system from various perspectives and record the behavior of its individual components. Heterogeneous data sets can be collectively mined by data fusion. Fusion can focus on a specific target relation and exploit directly ass…