Algorithmic decision-making systems are success-fully being adopted in a wide range of domains for diverse tasks. While the potential benefits of algorithmic decision-making are many, the importance of trusting these systems has only recently attracted attention. There has been a recent resurgence of interest in explainable artificial intelligence (XAI) that aims to reduce the opacity of a model by explaining its behavior, its predictions or both, thus allowing humans to scrutinize and trust the model. A host of technical advances have been made and several explanation methods have been proposed in recent years that address the problem of model explainability. In this tutorial, we will present these novel explanation approaches, characterize their strengths and limitations, and enumerate opportunities for data management research in the context of XAI.

Algorithmic decision-making systems are successfully being adopted in a wide range of domains for diverse tasks. While the potential benefits of algorithmic decision-making are many, the importance of trusting these systems has only recently attracted attention. There is growing concern that these systems are complex, opaque and non-intuitive, and hence are difficult to trust. There has been a recent resurgence of interest in explainable artificial intelligence (XAI) that aims to reduce the opacity of a model by explaining its behavior, its predictions or both, thus allowing humans to scrutinize and trust the model. A host of technical advances have been made and several explanation methods have been proposed in recent years that address the problem of model explainability and transparency. In this tutorial, we will present these novel explanation approaches, characterize their strengths and limitations, position existing work with respect to the database (DB) community, and enumerate opportunities for data management research in the context of XAI.

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 by utilizing techniques from the machine learning (ML) community to approximate causal responsibility, and using 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.

Machine learning (ML) models, while increasingly being used to make life-altering decisions, are known to reinforce systemic bias and discrimination. Consequently, practitioners and model developers need tools to facilitate debugging for bias in ML models. We introduce Gopher, a system that generates compact, interpretable and causal explanations for ML model bias. Gopher identifies the top-k coherent subsets of the training data that are root causes for model bias by quantifying the extent to which removing or updating a subset can resolve the bias. We describe the architecture of Gopher and will walk the audience through real-world use cases to highlight how Gopher generates explanations that enable data scientists to understand how subsets of the training data contribute to the bias of a machine learning (ML) model. Gopher is available as open-source software; The code and the demonstration video are available at https://gopher-sys.github.io/.

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 four 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.

Explainable artificial intelligence (XAI) aims to reduce the opacity of AI-based decision-making systems, allowing humans to scrutinize and trust them. Unlike prior work that attributes the responsibility for an algorithm's decisions to its inputs as a purely associational concept, we propose a principled causality-based approach for explaining black-box decision-making systems. We present the demonstration of Lewis, a system that generates explanations for black-box algorithms at the global, contextual, and local levels, and provides actionable recourse for individuals negatively affected by an algorithm's decision. Lewis makes no assumptions about the internals of the algorithm except for the availability of its input-output data. The explanations generated by Lewis are based on 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 describe the system layout of Lewis wherein an end-user specifies the underlying causal model and Lewis generates explanations for particular use-cases, compares them with explanations generated by state-of-the-art approaches in XAI, and provides actionable recourse when applicable. Lewis has been developed as open-source software; the code and the demonstration video are available at lewis-system.github.io.

Data fusion addresses the problem of consolidating data from disparate information providers into a single unified interface. The different data sources often provide conflicting information for the same data item. Recently, several automated data fusion models have been proposed to resolve conflicts and identify correct data. Although quite effective, these data fusion models do not achieve a close-to-perfect accuracy. We present the demonstration of a system that leverages users as first-class citizens to confirm data conflicts and rapidly improve the effectiveness of fusion. This demonstration is built on solutions proposed in our previous work [1]. To utilize the user judiciously, our system presents claims in an order that is the most beneficial to effectiveness of fusion across data items. We describe ranking algorithms that are built on concepts from information theory and decision theory, and do not need access to ground truth. We describe the user input framework and demonstrate how conflict resolution can be expedited with minimal feedback from the user. We show that: (a) the framework can be easily adopted to existing data fusion models without any internal changes to the models, and (b) the framework can integrate both perfect and imperfect feedback from users.

In domains such as the Web, sensor networks and social media, sources often provide conflicting information for the same data item. Several data fusion techniques have been proposed recently to resolve conflicts and identify correct data. The performance of these fusion systems, while quite accurate, is far from perfect. In this paper, we propose to leverage user feedback for validating data conflicts and rapidly improving the performance of fusion. To present the most beneficial data items for the user to validate, we take advantage of the level of consensus among sources, and the output of fusion to generate an effective ordering of items. We first evaluate data items individually, and then define a novel decision-theoretic framework based on the concept of value of perfect information (VPI) to order items by their ability to boost the performance of fusion. We further derive approximate formulae to scale up the decision-theoretic framework to large-scale data. We empirically evaluate our algorithms on three real-world datasets with different characteristics, and show that the accuracy of fusion can be significantly improved even while requesting feedback on a few data items. We also show that the performance of the proposed methods depends on the characteristics of data, and assess the trade-off between the amount of feedback acquired, and the effectiveness and efficiency of the methods.