Research Papers

Fairness Evaluation in Text Classification: Machine Learning Practitioner Perspectives of Individual and Group Fairness

Zahra Ashktorab, Benjamin Hoover, Mayank Agarwal +4

Mitigating algorithmic bias is a critical task in the development and deployment of machine learning models. While several toolkits exist to aid machine learning practitioners in addressing fairness issues, little is known about the strategies practitioners employ to evaluate model fairness and what factors influence their assessment, particularly in the context of text classification. Two common approaches of evaluating the fairness of a model are group fairness and individual fairness. We run a study with Machine Learning practitioners (n=24) to understand the strategies used to evaluate models. Metrics presented to practitioners (group vs. individual fairness) impact which models they consider fair. Participants focused on risks associated with underpredicting/overpredicting and model sensitivity relative to identity token manipulations. We discover fairness assessment strategies involving personal experiences or how users form groups of identity tokens to test model fairness. We provide recommendations for interactive tools for evaluating fairness in text classification.

Deep artificial neural networks show high predictive performance in many fields, but they do not afford statistical inferences and their black-box operations are too complicated for humans to comprehend. Because positing that a relationship exists is often more important than prediction in scientific experiments and research models, machine learning is far less frequently used than inferential statistics. Additionally, statistics calls for improving the test of theory by showing the magnitude of the phenomena being studied. This article extends current XAI methods and develops a model agnostic hypothesis testing framework for machine learning. First, Fisher's variable permutation algorithm is tweaked to compute an effect size measure equivalent to Cohen's f2 for OLS regression models. Second, the Mann-Kendall test of monotonicity and the Theil-Sen estimator is applied to Apley's accumulated local effect plots to specify a variable's direction of influence and statistical significance. The usefulness of this approach is demonstrated on an artificial data set and a social survey with a Python sandbox implementation.

Fairness and bias are crucial concepts in artificial intelligence, yet they are relatively ignored in machine learning applications in clinical psychiatry. We computed fairness metrics and present bias mitigation strategies using a model trained on clinical mental health data. We collected structured data related to the admission, diagnosis, and treatment of patients in the psychiatry department of the University Medical Center Utrecht. We trained a machine learning model to predict future administrations of benzodiazepines on the basis of past data. We found that gender plays an unexpected role in the predictions-this constitutes bias. Using the AI Fairness 360 package, we implemented reweighing and discrimination-aware regularization as bias mitigation strategies, and we explored their implications for model performance. This is the first application of bias exploration and mitigation in a machine learning model trained on real clinical psychiatry data.

With the increasing pervasiveness of Artificial Intelligence (AI), many visual analytics tools have been proposed to examine fairness, but they mostly focus on data scientist users. Instead, tackling fairness must be inclusive and involve domain experts with specialized tools and workflows. Thus, domain-specific visualizations are needed for algorithmic fairness. Furthermore, while much work on AI fairness has focused on predictive decisions, less has been done for fair allocation and planning, which require human expertise and iterative design to integrate myriad constraints. We propose the Intelligible Fair Allocation (IF-Alloc) Framework that leverages explanations of causal attribution (Why), contrastive (Why Not) and counterfactual reasoning (What If, How To) to aid domain experts to assess and alleviate unfairness in allocation problems. We apply the framework to fair urban planning for designing cities that provide equal access to amenities and benefits for diverse resident types. Specifically, we propose an interactive visual tool, Intelligible Fair City Planner (IF-City), to help urban planners to perceive inequality across groups, identify and attribute sources of inequality, and mitigate inequality with automatic allocation simulations and constraint-satisfying recommendations. We demonstrate and evaluate the usage and usefulness of IF-City on a real neighborhood in New York City, US, with practicing urban planners from multiple countries, and discuss generalizing our findings, application, and framework to other use cases and applications of fair allocation.

Efficient Decompositional Rule Extraction for Deep Neural Networks

Mateo Espinosa Zarlenga, Zohreh Shams, Mateja Jamnik

In recent years, there has been significant work on increasing both interpretability and debuggability of a Deep Neural Network (DNN) by extracting a rule-based model that approximates its decision boundary. Nevertheless, current DNN rule extraction methods that consider a DNN's latent space when extracting rules, known as decompositional algorithms, are either restricted to single-layer DNNs or intractable as the size of the DNN or data grows. In this paper, we address these limitations by introducing ECLAIRE, a novel polynomial-time rule extraction algorithm capable of scaling to both large DNN architectures and large training datasets. We evaluate ECLAIRE on a wide variety of tasks, ranging from breast cancer prognosis to particle detection, and show that it consistently extracts more accurate and comprehensible rule sets than the current state-of-the-art methods while using orders of magnitude less computational resources. We make all of our methods available, including a rule set visualisation interface, through the open-source REMIX library (https://github.com/mateoespinosa/remix).

From Philosophy to Interfaces: an Explanatory Method and a Tool Inspired by Achinstein's Theory of Explanation

Francesco Sovrano, Fabio Vitali

We propose a new method for explanations in Artificial Intelligence (AI) and a tool to test its expressive power within a user interface. In order to bridge the gap between philosophy and human-computer interfaces, we show a new approach for the generation of interactive explanations based on a sophisticated pipeline of AI algorithms for structuring natural language documents into knowledge graphs, answering questions effectively and satisfactorily. Among the mainstream philosophical theories of explanation we identified one that in our view is more easily applicable as a practical model for user-centric tools: Achinstein's Theory of Explanation. With this work we aim to prove that the theory proposed by Achinstein can be actually adapted for being implemented into a concrete software application, as an interactive process answering questions. To this end we found a way to handle the generic (archetypal) questions that implicitly characterise an explanatory processes as preliminary overviews rather than as answers to explicit questions, as commonly understood. To show the expressive power of this approach we designed and implemented a pipeline of AI algorithms for the generation of interactive explanations under the form of overviews, focusing on this aspect of explanations rather than on existing interfaces and presentation logic layers for question answering. We tested our hypothesis on a well-known XAI-powered credit approval system by IBM, comparing CEM, a static explanatory tool for post-hoc explanations, with an extension we developed adding interactive explanations based on our model. The results of the user study, involving more than 100 participants, showed that our proposed solution produced a statistically relevant improvement on effectiveness (U=931.0, p=0.036) over the baseline, thus giving evidence in favour of our theory.

Expressive Explanations of DNNs by Combining Concept Analysis with ILP

Johannes Rabold, Gesina Schwalbe, Ute Schmid

Explainable AI has emerged to be a key component for black-box machine learning approaches in domains with a high demand for reliability or transparency. Examples are medical assistant systems, and applications concerned with the General Data Protection Regulation of the European Union, which features transparency as a cornerstone. Such demands require the ability to audit the rationale behind a classifier's decision. While visualizations are the de facto standard of explanations, they come short in terms of expressiveness in many ways: They cannot distinguish between different attribute manifestations of visual features (e.g. eye open vs. closed), and they cannot accurately describe the influence of absence of, and relations between features. An alternative would be more expressive symbolic surrogate models. However, these require symbolic inputs, which are not readily available in most computer vision tasks. In this paper we investigate how to overcome this: We use inherent features learned by the network to build a global, expressive, verbal explanation of the rationale of a feed-forward convolutional deep neural network (DNN). The semantics of the features are mined by a concept analysis approach trained on a set of human understandable visual concepts. The explanation is found by an Inductive Logic Programming (ILP) method and presented as first-order rules. We show that our explanation is faithful to the original black-box model. The code for our experiments is available at https://github.com/mc-lovin-mlem/concept-embeddings-and-ilp/tree/ki2020.

A framework is presented to extract and understand decision-making information from a deep neural network (DNN) classifier of jet substructure tagging techniques. The general method studied is to provide expert variables that augment inputs ("eXpert AUGmented" variables, or XAUG variables), then apply layerwise relevance propagation (LRP) to networks both with and without XAUG variables. The XAUG variables are concatenated with the intermediate layers after network-specific operations (such as convolution or recurrence), and used in the final layers of the network. The results of comparing networks with and without the addition of XAUG variables show that XAUG variables can be used to interpret classifier behavior, increase discrimination ability when combined with low-level features, and in some cases capture the behavior of the classifier completely. The LRP technique can be used to find relevant information the network is using, and when combined with the XAUG variables, can be used to rank features, allowing one to find a reduced set of features that capture part of the network performance. In the studies presented, adding XAUG variables to low-level DNNs increased the efficiency of classifiers by as much as 30-40\%. In addition to performance improvements, an approach to quantify numerical uncertainties in the training of these DNNs is presented.

Should We Trust (X)AI? Design Dimensions for Structured Experimental Evaluations

Fabian Sperrle, Mennatallah El-Assady, Grace Guo +3

This paper systematically derives design dimensions for the structured evaluation of explainable artificial intelligence (XAI) approaches. These dimensions enable a descriptive characterization, facilitating comparisons between different study designs. They further structure the design space of XAI, converging towards a precise terminology required for a rigorous study of XAI. Our literature review differentiates between comparative studies and application papers, revealing methodological differences between the fields of machine learning, human-computer interaction, and visual analytics. Generally, each of these disciplines targets specific parts of the XAI process. Bridging the resulting gaps enables a holistic evaluation of XAI in real-world scenarios, as proposed by our conceptual model characterizing bias sources and trust-building. Furthermore, we identify and discuss the potential for future work based on observed research gaps that should lead to better coverage of the proposed model.

Future of sustainable fashion lies in adoption of AI for a better understanding of consumer shopping behaviour and using this understanding to further optimize product design, development and sourcing to finally reduce the probability of overproducing inventory. Explainability and interpretability are highly effective in increasing the adoption of AI based tools in creative domains like fashion. In a fashion house, stakeholders like buyers, merchandisers and financial planners have a more quantitative approach towards decision making with primary goals of high sales and reduced dead inventory. Whereas, designers have a more intuitive approach based on observing market trends, social media and runways shows. Our goal is to build an explainable new product forecasting tool with capabilities of interventional analysis such that all the stakeholders (with competing goals) can participate in collaborative decision making process of new product design, development and launch.

With the widespread integration of AI in everyday and critical technologies, it seems inevitable to witness increasing instances of failure in AI systems. In such cases, there arises a need for technical investigations that produce legally acceptable and scientifically indisputable findings and conclusions on the causes of such failures. Inspired by the domain of cyber forensics, this paper introduces the need for the establishment of AI Forensics as a new discipline under AI safety. Furthermore, we propose a taxonomy of the subfields under this discipline, and present a discussion on the foundational challenges that lay ahead of this new research area.

In this study, we conduct a resume-screening experiment (N=528) where people collaborate with simulated AI models exhibiting race-based preferences (bias) to evaluate candidates for 16 high and low status occupations. Simulated AI bias approximates factual and counterfactual estimates of racial bias in real-world AI systems. We investigate people's preferences for White, Black, Hispanic, and Asian candidates (represented through names and affinity groups on quality-controlled resumes) across 1,526 scenarios and measure their unconscious associations between race and status using implicit association tests (IATs), which predict discriminatory hiring decisions but have not been investigated in human-AI collaboration. When making decisions without AI or with AI that exhibits no race-based preferences, people select all candidates at equal rates. However, when interacting with AI favoring a particular group, people also favor those candidates up to 90% of the time, indicating a significant behavioral shift. The likelihood of selecting candidates whose identities do not align with common race-status stereotypes can increase by 13% if people complete an IAT before conducting resume screening. Finally, even if people think AI recommendations are low quality or not important, their decisions are still vulnerable to AI bias under certain circumstances. This work has implications for people's autonomy in AI-HITL scenarios, AI and work, design and evaluation of AI hiring systems, and strategies for mitigating bias in collaborative decision-making tasks. In particular, organizational and regulatory policy should acknowledge the complex nature of AI-HITL decision making when implementing these systems, educating people who use them, and determining which are subject to oversight.