arXiv subjects

Babak Salimi

Publications and source records attributed to Babak Salimi.

At least 37 records · Page 2Linked to original sources

arXiv2022

Explaining Image Classifiers Using Contrastive Counterfactuals in Generative Latent Spaces

Despite their high accuracies, modern complex image classifiers cannot be trusted for sensitive tasks due to their unknown decision-making process and potential biases. Counterfactual explanations are very effective in providing transparency for these black-box algorithms. Nevertheless, generating counterfactuals that can have a consistent impact on classifier outputs and yet expose interpretable feature changes is a very challenging task. We introduce a novel method to generate causal and yet interpretable counterfactual explanations for image classifiers using pretrained generative models without any re-training or conditioning. The generative models in this technique are not bound to be trained on the same data as the target classifier. We use this framework to obtain contrastive and causal sufficiency and necessity scores as global explanations for black-box classifiers. On the task of face attribute classification, we show how different attributes influence the classifier output by providing both causal and contrastive feature attributions, and the corresponding counterfactual images.

cs.CV↗

arXiv2022

Through the Data Management Lens: Experimental Analysis and Evaluation of Fair Classification

Classification, a heavily-studied data-driven machine learning task, drives an increasing number of prediction systems involving critical human decisions such as loan approval and criminal risk assessment. However, classifiers often demonstrate discriminatory behavior, especially when presented with biased data. Consequently, fairness in classification has emerged as a high-priority research area. Data management research is showing an increasing presence and interest in topics related to data and algorithmic fairness, including the topic of fair classification. The interdisciplinary efforts in fair classification, with machine learning research having the largest presence, have resulted in a large number of fairness notions and a wide range of approaches that have not been systematically evaluated and compared. In this paper, we contribute a broad analysis of 13 fair classification approaches and additional variants, over their correctness, fairness, efficiency, scalability, robustness to data errors, sensitivity to underlying ML model, data efficiency, and stability using a variety of metrics and real-world datasets. Our analysis highlights novel insights on the impact of different metrics and high-level approach characteristics on different aspects of performance. We also discuss general principles for choosing approaches suitable for different practical settings, and identify areas where data-management-centric solutions are likely to have the most impact.

cs.LG↗

arXiv2022

Interpretable Data-Based Explanations for Fairness Debugging

A wide variety of fairness metrics and eXplainable Artificial Intelligence (XAI) approaches have been proposed in the literature to identify bias in machine learning models that are used in critical real-life contexts. However, merely reporting on a model's bias, or generating explanations using existing XAI techniques is insufficient to locate and eventually mitigate sources of bias. We introduce Gopher, a system that produces compact, interpretable and causal explanations for bias or unexpected model behavior by identifying coherent subsets of the training data that are root-causes for this behavior. Specifically, we introduce the concept of causal responsibility that quantifies the extent to which intervening on training data by removing or updating subsets of it can resolve the bias. Building on this concept, we develop an efficient approach for generating the top-k patterns that explain model bias that utilizes techniques from the machine learning (ML) community to approximate causal responsibility and uses pruning rules to manage the large search space for patterns. Our experimental evaluation demonstrates the effectiveness of Gopher in generating interpretable explanations for identifying and debugging sources of bias.

cs.LG↗

arXiv2022

HypeR: Hypothetical Reasoning With What-If and How-To Queries Using a Probabilistic Causal Approach

What-if (provisioning for an update to a database) and how-to (how to modify the database to achieve a goal) analyses provide insights to users who wish to examine hypothetical scenarios without making actual changes to a database and thereby help plan strategies in their fields. Typically, such analyses are done by testing the effect of an update in the existing database on a specific view created by a query of interest. In real-world scenarios, however, an update to a particular part of the database may affect tuples and attributes in a completely different part due to implicit semantic dependencies. To allow for hypothetical reasoning while accommodating such dependencies, we develop HypeR, a framework that supports what-if and how-to queries accounting for probabilistic dependencies among attributes captured by a probabilistic causal model. We extend the SQL syntax to include the necessary operators for expressing these hypothetical queries, define their semantics, devise efficient algorithms and optimizations to compute their results using concepts from causality and probabilistic databases, and evaluate the effectiveness of our approach experimentally.

cs.DB↗

arXiv2021

Heterogeneous Treatment Effects in Social Networks

We study treatment effect modifiers for causal analysis in a social network, where neighbors' characteristics or network structure may affect the outcome of a unit, and the goal is to identify sub-populations with varying treatment effects using such network properties. We propose a novel framework for this purpose that facilitates data-driven decision making by testing hypotheses about complex effect modifiers in terms of network features or network patterns (e.g., characteristics of neighbors of a unit or belonging to a triangle), and by identifying sub-populations for which a treatment is likely to be effective or harmful. We describe a hypothesis testing approach that accounts for a unit's covariates, their neighbors' covariates, and patterns in the social network, and devise an algorithm incorporating ideas from causal inference, hypothesis testing, and graph theory to verify a hypothesized effect modifier. In addition, we develop a novel algorithm for the discovery of network patterns that are potential effect modifiers. We perform extensive experimental evaluations with a real development economics dataset about the treatment effect of belonging to a financial support network called self-help groups on risk tolerance, and also with a synthetic dataset with known ground truths simulating a vaccine efficacy trial, to evaluate our framework and algorithms.

cs.SI↗

arXiv2021

Explaining Black-Box Algorithms Using Probabilistic Contrastive Counterfactuals

There has been a recent resurgence of interest in explainable artificial intelligence (XAI) that aims to reduce the opaqueness of AI-based decision-making systems, allowing humans to scrutinize and trust them. Prior work in this context has focused on the attribution of responsibility for an algorithm's decisions to its inputs wherein responsibility is typically approached as a purely associational concept. In this paper, we propose a principled causality-based approach for explaining black-box decision-making systems that addresses limitations of existing methods in XAI. At the core of our framework lies probabilistic contrastive counterfactuals, a concept that can be traced back to philosophical, cognitive, and social foundations of theories on how humans generate and select explanations. We show how such counterfactuals can quantify the direct and indirect influences of a variable on decisions made by an algorithm, and provide actionable recourse for individuals negatively affected by the algorithm's decision. Unlike prior work, our system, LEWIS: (1)can compute provably effective explanations and recourse at local, global and contextual levels (2)is designed to work with users with varying levels of background knowledge of the underlying causal model and (3)makes no assumptions about the internals of an algorithmic system except for the availability of its input-output data. We empirically evaluate LEWIS on three real-world datasets and show that it generates human-understandable explanations that improve upon state-of-the-art approaches in XAI, including the popular LIME and SHAP. Experiments on synthetic data further demonstrate the correctness of LEWIS's explanations and the scalability of its recourse algorithm.

cs.AI↗

arXiv2020

Causal Relational Learning

Causal inference is at the heart of empirical research in natural and social sciences and is critical for scientific discovery and informed decision making. The gold standard in causal inference is performing randomized controlled trials; unfortunately these are not always feasible due to ethical, legal, or cost constraints. As an alternative, methodologies for causal inference from observational data have been developed in statistical studies and social sciences. However, existing methods critically rely on restrictive assumptions such as the study population consisting of homogeneous elements that can be represented in a single flat table, where each row is referred to as a unit. In contrast, in many real-world settings, the study domain naturally consists of heterogeneous elements with complex relational structure, where the data is naturally represented in multiple related tables. In this paper, we present a formal framework for causal inference from such relational data. We propose a declarative language called CaRL for capturing causal background knowledge and assumptions and specifying causal queries using simple Datalog-like rules.CaRL provides a foundation for inferring causality and reasoning about the effect of complex interventions in relational domains. We present an extensive experimental evaluation on real relational data to illustrate the applicability of CaRL in social sciences and healthcare.

cs.DB↗

arXiv2019

Mining Approximate Acyclic Schemes from Relations

Acyclic schemes have numerous applications in databases and in machine learning, such as improved design, more efficient storage, and increased performance for queries and machine learning algorithms. Multivalued dependencies (MVDs) are the building blocks of acyclic schemes. The discovery from data of both MVDs and acyclic schemes is more challenging than other forms of data dependencies, such as Functional Dependencies, because these dependencies do not hold on subsets of data, and because they are very sensitive to noise in the data; for example a single wrong or missing tuple may invalidate the schema. In this paper we present Maimon, a system for discovering approximate acyclic schemes and MVDs from data. We give a principled definition of approximation, by using notions from information theory, then describe the two components of Maimon: mining for approximate MVDs, then reconstructing acyclic schemes from approximate MVDs. We conduct an experimental evaluation of Maimon on 20 real-world datasets, and show that it can scale up to 1M rows, and up to 30 columns.

cs.DB↗

arXiv2019

Capuchin: Causal Database Repair for Algorithmic Fairness

Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflect discrimination, suggesting a database repair problem. Existing treatments of fairness rely on statistical correlations that can be fooled by statistical anomalies, such as Simpson's paradox. Proposals for causality-based definitions of fairness can correctly model some of these situations, but they require specification of the underlying causal models. In this paper, we formalize the situation as a database repair problem, proving sufficient conditions for fair classifiers in terms of admissible variables as opposed to a complete causal model. We show that these conditions correctly capture subtle fairness violations. We then use these conditions as the basis for database repair algorithms that provide provable fairness guarantees about classifiers trained on their training labels. We evaluate our algorithms on real data, demonstrating improvement over the state of the art on multiple fairness metrics proposed in the literature while retaining high utility.

cs.DB↗

arXiv2019

Data Management for Causal Algorithmic Fairness

Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflects discrimination, suggesting a data management problem. In this paper, we first make a distinction between associational and causal definitions of fairness in the literature and argue that the concept of fairness requires causal reasoning. We then review existing works and identify future opportunities for applying data management techniques to causal algorithmic fairness.

cs.DB↗

arXiv2019

MobilityMirror: Bias-Adjusted Transportation Datasets

We describe customized synthetic datasets for publishing mobility data. Private companies are providing new transportation modalities, and their data is of high value for integrative transportation research, policy enforcement, and public accountability. However, these companies are disincentivized from sharing data not only to protect the privacy of individuals (drivers and/or passengers), but also to protect their own competitive advantage. Moreover, demographic biases arising from how the services are delivered may be amplified if released data is used in other contexts. We describe a model and algorithm for releasing origin-destination histograms that removes selected biases in the data using causality-based methods. We compute the origin-destination histogram of the original dataset then adjust the counts to remove undesirable causal relationships that can lead to discrimination or violate contractual obligations with data owners. We evaluate the utility of the algorithm on real data from a dockless bike share program in Seattle and taxi data in New York, and show that these adjusted transportation datasets can retain utility while removing bias in the underlying data.

cs.DB↗

arXiv2018

Bias in OLAP Queries: Detection, Explanation, and Removal

On line analytical processing (OLAP) is an essential element of decision-support systems. OLAP tools provide insights and understanding needed for improved decision making. However, the answers to OLAP queries can be biased and lead to perplexing and incorrect insights. In this paper, we propose HypDB, a system to detect, explain, and to resolve bias in decision-support queries. We give a simple definition of a \emph{biased query}, which performs a set of independence tests on the data to detect bias. We propose a novel technique that gives explanations for bias, thus assisting an analyst in understanding what goes on. Additionally, we develop an automated method for rewriting a biased query into an unbiased query, which shows what the analyst intended to examine. In a thorough evaluation on several real datasets we show both the quality and the performance of our techniques, including the completely automatic discovery of the revolutionary insights from a famous 1973 discrimination case.

cs.DB↗

arXiv2017

A Framework for Inferring Causality from Multi-Relational Observational Data using Conditional Independence

The study of causality or causal inference - how much a given treatment causally affects a given outcome in a population - goes way beyond correlation or association analysis of variables, and is critical in making sound data driven decisions and policies in a multitude of applications. The gold standard in causal inference is performing "controlled experiments", which often is not possible due to logistical or ethical reasons. As an alternative, inferring causality on "observational data" based on the "Neyman-Rubin potential outcome model" has been extensively used in statistics, economics, and social sciences over several decades. In this paper, we present a formal framework for sound causal analysis on observational datasets that are given as multiple relations and where the population under study is obtained by joining these base relations. We study a crucial condition for inferring causality from observational data, called the "strong ignorability assumption" (the treatment and outcome variables should be independent in the joined relation given the observed covariates), using known conditional independences that hold in the base relations. We also discuss how the structure of the conditional independences in base relations given as graphical models help infer new conditional independences in the joined relation. The proposed framework combines concepts from databases, statistics, and graphical models, and aims to initiate new research directions spanning these fields to facilitate powerful data-driven decisions in today's big data world.

cs.DB↗

arXiv2017

Causes for Query Answers from Databases: Datalog Abduction, View-Updates, and Integrity Constraints

Causality has been recently introduced in databases, to model, characterize, and possibly compute causes for query answers. Connections between QA-causality and consistency-based diagnosis and database repairs (wrt. integrity constraint violations) have already been established. In this work we establish precise connections between QA-causality and both abductive diagnosis and the view-update problem in databases, allowing us to obtain new algorithmic and complexity results for QA-causality. We also obtain new results on the complexity of view-conditioned causality, and investigate the notion of QA-causality in the presence of integrity constraints, obtaining complexity results from a connection with view-conditioned causality. The abduction connection under integrity constraints allows us to obtain algorithmic tools for QA-causality.

cs.DB↗

arXiv2016

From Causes for Database Queries to Repairs and Model-Based Diagnosis and Back

In this work we establish and investigate connections between causes for query answers in databases, database repairs wrt. denial constraints, and consistency-based diagnosis. The first two are relatively new research areas in databases, and the third one is an established subject in knowledge representation. We show how to obtain database repairs from causes, and the other way around. Causality problems are formulated as diagnosis problems, and the diagnoses provide causes and their responsibilities. The vast body of research on database repairs can be applied to the newer problems of computing actual causes for query answers and their responsibilities. These connections, which are interesting per se, allow us, after a transition -inspired by consistency-based diagnosis- to computational problems on hitting sets and vertex covers in hypergraphs, to obtain several new algorithmic and complexity results for database causality.

cs.DB↗

arXiv2016

ZaliQL: A SQL-Based Framework for Drawing Causal Inference from Big Data

Causal inference from observational data is a subject of active research and development in statistics and computer science. Many toolkits have been developed for this purpose that depends on statistical software. However, these toolkits do not scale to large datasets. In this paper we describe a suite of techniques for expressing causal inference tasks from observational data in SQL. This suite supports the state-of-the-art methods for causal inference and run at scale within a database engine. In addition, we introduce several optimization techniques that significantly speedup causal inference, both in the online and offline setting. We evaluate the quality and performance of our techniques by experiments of real datasets.

cs.DB↗

arXiv2016

Quantifying Causal Effects on Query Answering in Databases

The notion of actual causation, as formalized by Halpern and Pearl, has been recently applied to relational databases, to characterize and compute actual causes for possibly unexpected answers to monotone queries. Causes take the form of database tuples, and can be ranked according to their causal responsibility, a numerical measure of their relevance as a cause to the query answer. In this work we revisit this notion, introducing and making a case for an alternative measure of causal contribution, that of causal effect. The measure generalizes actual causes, and can be applied beyond monotone queries. We show that causal effect provides intuitive and intended results.

cs.DB↗

arXiv2016

Causes for Query Answers from Databases, Datalog Abduction and View-Updates: The Presence of Integrity Constraints

Causality has been recently introduced in databases, to model, characterize and possibly compute causes for query results (answers). Connections between queryanswer causality, consistency-based diagnosis, database repairs (wrt. integrity constraint violations), abductive diagnosis and the view-update problem have been established. In this work we further investigate connections between query-answer causality and abductive diagnosis and the view-update problem. In this context, we also define and investigate the notion of query-answer causality in the presence of integrity constraints.

cs.DB↗