Research Papers

Proceedings of ICML 2021 Workshop on Theoretic Foundation, Criticism, and Application Trend of Explainable AI

Quanshi Zhang, Tian Han, Lixin Fan +5

This is the Proceedings of ICML 2021 Workshop on Theoretic Foundation, Criticism, and Application Trend of Explainable AI. Deep neural networks (DNNs) have undoubtedly brought great success to a wide range of applications in computer vision, computational linguistics, and AI. However, foundational principles underlying the DNNs' success and their resilience to adversarial attacks are still largely missing. Interpreting and theorizing the internal mechanisms of DNNs becomes a compelling yet controversial topic. This workshop pays a special interest in theoretic foundations, limitations, and new application trends in the scope of XAI. These issues reflect new bottlenecks in the future development of XAI.

Modern data analytics underpinned by machine learning techniques has become a key enabler to the automation of data-led decision making. As an important branch of state-of-the-art data analytics, business process predictions are also faced with a challenge in regard to the lack of explanation to the reasoning and decision by the underlying `black-box' prediction models. With the development of interpretable machine learning techniques, explanations can be generated for a black-box model, making it possible for (human) users to access the reasoning behind machine learned predictions. In this paper, we aim to present an approach that allows us to use model explanations to investigate certain reasoning applied by machine learned predictions and detect potential issues with the underlying methods thus enhancing trust in business process prediction models. A novel contribution of our approach is the proposal of model inspection that leverages both the explanations generated by interpretable machine learning mechanisms and the contextual or domain knowledge extracted from event logs that record historical process execution. Findings drawn from this work are expected to serve as a key input to developing model reliability metrics and evaluation in the context of business process predictions.

Towards Model-informed Precision Dosing with Expert-in-the-loop Machine Learning

Yihuang Kang, Yi-Wen Chiu, Ming-Yen Lin +2

Machine Learning (ML) and its applications have been transforming our lives but it is also creating issues related to the development of fair, accountable, transparent, and ethical Artificial Intelligence. As the ML models are not fully comprehensible yet, it is obvious that we still need humans to be part of algorithmic decision-making processes. In this paper, we consider a ML framework that may accelerate model learning and improve its interpretability by incorporating human experts into the model learning loop. We propose a novel human-in-the-loop ML framework aimed at dealing with learning problems that the cost of data annotation is high and the lack of appropriate data to model the association between the target tasks and the input features. With an application to precision dosing, our experimental results show that the approach can learn interpretable rules from data and may potentially lower experts' workload by replacing data annotation with rule representation editing. The approach may also help remove algorithmic bias by introducing experts' feedback into the iterative model learning process.

Towards Fully Interpretable Deep Neural Networks: Are We There Yet?

Sandareka Wickramanayake, Wynne Hsu, Mong Li Lee

Despite the remarkable performance, Deep Neural Networks (DNNs) behave as black-boxes hindering user trust in Artificial Intelligence (AI) systems. Research on opening black-box DNN can be broadly categorized into post-hoc methods and inherently interpretable DNNs. While many surveys have been conducted on post-hoc interpretation methods, little effort is devoted to inherently interpretable DNNs. This paper provides a review of existing methods to develop DNNs with intrinsic interpretability, with a focus on Convolutional Neural Networks (CNNs). The aim is to understand the current progress towards fully interpretable DNNs that can cater to different interpretation requirements. Finally, we identify gaps in current work and suggest potential research directions.

Explaining Graph Neural Networks predictions to end users of AI applications in easily understandable terms remains an unsolved problem. In particular, we do not have well developed methods for automatically evaluating explanations, in ways that are closer to how users consume those explanations. Based on recent application trends and our own experiences in real world problems, we propose automatic evaluation approaches for GNN Explanations.

Adversarial images highlight how vulnerable modern image classifiers are to perturbations outside of their training set. Human oversight might mitigate this weakness, but depends on humans understanding the AI well enough to predict when it is likely to make a mistake. In previous work we have found that humans tend to assume that the AI's decision process mirrors their own. Here we evaluate if methods from explainable AI can disrupt this assumption to help participants predict AI classifications for adversarial and standard images. We find that both saliency maps and examples facilitate catching AI errors, but their effects are not additive, and saliency maps are more effective than examples.

Explaining the decision of a multi-modal decision-maker requires to determine the evidence from both modalities. Recent advances in XAI provide explanations for models trained on still images. However, when it comes to modeling multiple sensory modalities in a dynamic world, it remains underexplored how to demystify the mysterious dynamics of a complex multi-modal model. In this work, we take a crucial step forward and explore learnable explanations for audio-visual recognition. Specifically, we propose a novel space-time attention network that uncovers the synergistic dynamics of audio and visual data over both space and time. Our model is capable of predicting the audio-visual video events, while justifying its decision by localizing where the relevant visual cues appear, and when the predicted sounds occur in videos. We benchmark our model on three audio-visual video event datasets, comparing extensively to multiple recent multi-modal representation learners and intrinsic explanation models. Experimental results demonstrate the clear superior performance of our model over the existing methods on audio-visual video event recognition. Moreover, we conduct an in-depth study to analyze the explainability of our model based on robustness analysis via perturbation tests and pointing games using human annotations.

The field of explainable AI has recently seen an explosion in the number of explanation methods for highly non-linear deep neural networks. The extent to which such methods -- that are often proposed and tested in the domain of computer vision -- are appropriate to address the explainability challenges in NLP is yet relatively unexplored. In this work, we consider Contextual Decomposition (CD) -- a Shapley-based input feature attribution method that has been shown to work well for recurrent NLP models -- and we test the extent to which it is useful for models that contain attention operations. To this end, we extend CD to cover the operations necessary for attention-based models. We then compare how long distance subject-verb relationships are processed by models with and without attention, considering a number of different syntactic structures in two different languages: English and Dutch. Our experiments confirm that CD can successfully be applied for attention-based models as well, providing an alternative Shapley-based attribution method for modern neural networks. In particular, using CD, we show that the English and Dutch models demonstrate similar processing behaviour, but that under the hood there are consistent differences between our attention and non-attention models.

We Haven't Gone Paperless Yet: Why the Printing Press Can Help Us Understand Data and AI

Julian Posada, Nicholas Weller, Wendy H. Wong

How should we understand the social and political effects of the datafication of human life? This paper argues that the effects of data should be understood as a constitutive shift in social and political relations. We explore how datafication, or quantification of human and non-human factors into binary code, affects the identity of individuals and groups. This fundamental shift goes beyond economic and ethical concerns, which has been the focus of other efforts to explore the effects of datafication and AI. We highlight that technologies such as datafication and AI (and previously, the printing press) both disrupted extant power arrangements, leading to decentralization, and triggered a recentralization of power by new actors better adapted to leveraging the new technology. We use the analogy of the printing press to provide a framework for understanding constitutive change. The printing press example gives us more clarity on 1) what can happen when the medium of communication drastically alters how information is communicated and stored; 2) the shift in power from state to private actors; and 3) the tension of simultaneously connecting individuals while driving them towards narrower communities through algorithmic analyses of data.

Deep Learning has become a very valuable tool in different fields, and no one doubts the learning capacity of these models. Nevertheless, since Deep Learning models are often seen as black boxes due to their lack of interpretability, there is a general mistrust in their decision-making process. To find a balance between effectiveness and interpretability, Explainable Artificial Intelligence (XAI) is gaining popularity in recent years, and some of the methods within this area are used to generate counterfactual explanations. The process of generating these explanations generally consists of solving an optimization problem for each input to be explained, which is unfeasible when real-time feedback is needed. To speed up this process, some methods have made use of autoencoders to generate instant counterfactual explanations. Recently, a method called Deep Guided Counterfactual Explanations (DGCEx) has been proposed, which trains an autoencoder attached to a classification model, in order to generate straightforward counterfactual explanations. However, this method does not ensure that the generated counterfactual instances are close to the data manifold, so unrealistic counterfactual instances may be generated. To overcome this issue, this paper presents Distribution Aware Deep Guided Counterfactual Explanations (DA-DGCEx), which adds a term to the DGCEx cost function that penalizes out of distribution counterfactual instances.

Faithful and Plausible Explanations of Medical Code Predictions

Zach Wood-Doughty, Isabel Cachola, Mark Dredze

Machine learning models that offer excellent predictive performance often lack the interpretability necessary to support integrated human machine decision-making. In clinical medicine and other high-risk settings, domain experts may be unwilling to trust model predictions without explanations. Work in explainable AI must balance competing objectives along two different axes: 1) Explanations must balance faithfulness to the model's decision-making with their plausibility to a domain expert. 2) Domain experts desire local explanations of individual predictions and global explanations of behavior in aggregate. We propose to train a proxy model that mimics the behavior of the trained model and provides fine-grained control over these trade-offs. We evaluate our approach on the task of assigning ICD codes to clinical notes to demonstrate that explanations from the proxy model are faithful and replicate the trained model behavior.

Explainable Artificial Intelligence (XAI) is a rising field in AI. It aims to produce a demonstrative factor of trust, which for human subjects is achieved through communicative means, which Machine Learning (ML) algorithms cannot solely produce, illustrating the necessity of an extra layer producing support to the model output. When approaching the medical field, we can see challenges arise when dealing with the involvement of human-subjects, the ideology behind trusting a machine to tend towards the livelihood of a human poses an ethical conundrum - leaving trust as the basis of the human-expert in acceptance to the machines decision. The aim of this paper is to apply XAI methods to demonstrate the usability of explainable architectures as a tertiary layer for the medical domain supporting ML predictions and human-expert opinion, XAI methods produce visualization of the feature contribution towards a given models output on both a local and global level. The work in this paper uses XAI to determine feature importance towards high-dimensional data-driven questions to inform domain-experts of identifiable trends with a comparison of model-agnostic methods in application to ML algorithms. The performance metrics for a glass-box method is also provided as a comparison against black-box capability for tabular data. Future work will aim to produce a user-study using metrics to evaluate human-expert usability and opinion of the given models.