Research Papers

Surging interest in deep learning from high-stakes domains has precipitated concern over the inscrutable nature of black box neural networks. Explainable AI (XAI) research has led to an abundance of explanation algorithms for these black boxes. Such post hoc explainers produce human-comprehensible explanations, however, their fidelity with respect to the model is not well understood - explanation evaluation remains one of the most challenging issues in XAI. In this paper, we ask a targeted but important question: can popular feature-additive explainers (e.g., LIME, SHAP, SHAPR, MAPLE, and PDP) explain feature-additive predictors? Herein, we evaluate such explainers on ground truth that is analytically derived from the additive structure of a model. We demonstrate the efficacy of our approach in understanding these explainers applied to symbolic expressions, neural networks, and generalized additive models on thousands of synthetic and several real-world tasks. Our results suggest that all explainers eventually fail to correctly attribute the importance of features, especially when a decision-making process involves feature interactions.

A Uniform Language to Explain Decision Trees

Marcelo Arenas, Pablo Barcelo, Diego Bustamante +2

The formal XAI community has studied a plethora of interpretability queries aiming to understand the classifications made by decision trees. However, a more uniform understanding of what questions we can hope to answer about these models, traditionally deemed to be easily interpretable, has remained elusive. In an initial attempt to understand uniform languages for interpretability, Arenas et al. (2021) proposed FOIL, a logic for explaining black-box ML models, and showed that it can express a variety of interpretability queries. However, we show that FOIL is limited in two important senses: (i) it is not expressive enough to capture some crucial queries, and (ii) its model agnostic nature results in a high computational complexity for decision trees. In this paper, we carefully craft two fragments of first-order logic that allow for efficiently interpreting decision trees: Q-DT-FOIL and its optimization variant OPT-DT-FOIL. We show that our proposed logics can express not only a variety of interpretability queries considered by previous literature, but also elegantly allows users to specify different objectives the sought explanations should optimize for. Using finite model-theoretic techniques, we show that the different ingredients of Q-DT-FOIL are necessary for its expressiveness, and yet that queries in Q-DT-FOIL can be evaluated with a polynomial number of queries to a SAT solver, as well as their optimization versions in OPT-DT-FOIL. Besides our theoretical results, we provide a SAT-based implementation of the evaluation for OPT-DT-FOIL that is performant on industry-size decision trees.

The AI model has surpassed human players in the game of Go, and it is widely believed that the AI model has encoded new knowledge about the Go game beyond human players. In this way, explaining the knowledge encoded by the AI model and using it to teach human players represent a promising-yet-challenging issue in explainable AI. To this end, mathematical supports are required to ensure that human players can learn accurate and verifiable knowledge, rather than specious intuitive analysis. Thus, in this paper, we extract interaction primitives between stones encoded by the value network for the Go game, so as to enable people to learn from the value network. Experiments show the effectiveness of our method.

Artificial Intelligence (AI) is often an integral part of modern decision support systems. The best-performing predictive models used in AI-based decision support systems lack transparency. Explainable Artificial Intelligence (XAI) aims to create AI systems that can explain their rationale to human users. Local explanations in XAI can provide information about the causes of individual predictions in terms of feature importance. However, a critical drawback of existing local explanation methods is their inability to quantify the uncertainty associated with a feature's importance. This paper introduces an extension of a feature importance explanation method, Calibrated Explanations, previously only supporting classification, with support for standard regression and probabilistic regression, i.e., the probability that the target is above an arbitrary threshold. The extension for regression keeps all the benefits of Calibrated Explanations, such as calibration of the prediction from the underlying model with confidence intervals, uncertainty quantification of feature importance, and allows both factual and counterfactual explanations. Calibrated Explanations for standard regression provides fast, reliable, stable, and robust explanations. Calibrated Explanations for probabilistic regression provides an entirely new way of creating probabilistic explanations from any ordinary regression model, allowing dynamic selection of thresholds. The method is model agnostic with easily understood conditional rules. An implementation in Python is freely available on GitHub and for installation using both pip and conda, making the results in this paper easily replicable.

We address the task of probabilistic anomaly attribution in the black-box regression setting, where the goal is to compute the probability distribution of the attribution score of each input variable, given an observed anomaly. The training dataset is assumed to be unavailable. This task differs from the standard XAI (explainable AI) scenario, since we wish to explain the anomalous deviation from a black-box prediction rather than the black-box model itself. We begin by showing that mainstream model-agnostic explanation methods, such as the Shapley values, are not suitable for this task because of their ``deviation-agnostic property.'' We then propose a novel framework for probabilistic anomaly attribution that allows us to not only compute attribution scores as the predictive mean but also quantify the uncertainty of those scores. This is done by considering a generative process for perturbations that counter-factually bring the observed anomalous observation back to normalcy. We introduce a variational Bayes algorithm for deriving the distributions of per variable attribution scores. To the best of our knowledge, this is the first probabilistic anomaly attribution framework that is free from being deviation-agnostic.

For Better or Worse: The Impact of Counterfactual Explanations' Directionality on User Behavior in xAI

Ulrike Kuhl, André Artelt, Barbara Hammer

Counterfactual explanations (CFEs) are a popular approach in explainable artificial intelligence (xAI), highlighting changes to input data necessary for altering a model's output. A CFE can either describe a scenario that is better than the factual state (upward CFE), or a scenario that is worse than the factual state (downward CFE). However, potential benefits and drawbacks of the directionality of CFEs for user behavior in xAI remain unclear. The current user study (N=161) compares the impact of CFE directionality on behavior and experience of participants tasked to extract new knowledge from an automated system based on model predictions and CFEs. Results suggest that upward CFEs provide a significant performance advantage over other forms of counterfactual feedback. Moreover, the study highlights potential benefits of mixed CFEs improving user performance compared to downward CFEs or no explanations. In line with the performance results, users' explicit knowledge of the system is statistically higher after receiving upward CFEs compared to downward comparisons. These findings imply that the alignment between explanation and task at hand, the so-called regulatory fit, may play a crucial role in determining the effectiveness of model explanations, informing future research directions in xAI. To ensure reproducible research, the entire code, underlying models and user data of this study is openly available: https://github.com/ukuhl/DirectionalAlienZoo

HiTZ@Antidote: Argumentation-driven Explainable Artificial Intelligence for Digital Medicine

Rodrigo Agerri, Iñigo Alonso, Aitziber Atutxa +9

Providing high quality explanations for AI predictions based on machine learning is a challenging and complex task. To work well it requires, among other factors: selecting a proper level of generality/specificity of the explanation; considering assumptions about the familiarity of the explanation beneficiary with the AI task under consideration; referring to specific elements that have contributed to the decision; making use of additional knowledge (e.g. expert evidence) which might not be part of the prediction process; and providing evidence supporting negative hypothesis. Finally, the system needs to formulate the explanation in a clearly interpretable, and possibly convincing, way. Given these considerations, ANTIDOTE fosters an integrated vision of explainable AI, where low-level characteristics of the deep learning process are combined with higher level schemes proper of the human argumentation capacity. ANTIDOTE will exploit cross-disciplinary competences in deep learning and argumentation to support a broader and innovative view of explainable AI, where the need for high-quality explanations for clinical cases deliberation is critical. As a first result of the project, we publish the Antidote CasiMedicos dataset to facilitate research on explainable AI in general, and argumentation in the medical domain in particular.

A recent trend in explainable AI research has focused on surrogate modeling, where neural networks are approximated as simpler ML algorithms such as kernel machines. A second trend has been to utilize kernel functions in various explain-by-example or data attribution tasks. In this work, we combine these two trends to analyze approximate empirical neural tangent kernels (eNTK) for data attribution. Approximation is critical for eNTK analysis due to the high computational cost to compute the eNTK. We define new approximate eNTK and perform novel analysis on how well the resulting kernel machine surrogate models correlate with the underlying neural network. We introduce two new random projection variants of approximate eNTK which allow users to tune the time and memory complexity of their calculation. We conclude that kernel machines using approximate neural tangent kernel as the kernel function are effective surrogate models, with the introduced trace NTK the most consistent performer. Open source software allowing users to efficiently calculate kernel functions in the PyTorch framework is available (https://github.com/pnnl/projection\_ntk).

Electronic health records (EHRs) serve as an essential data source for the envisioned artificial intelligence (AI)-driven transformation in healthcare. However, clinician biases reflected in EHR notes can lead to AI models inheriting and amplifying these biases, perpetuating health disparities. This study investigates the impact of stigmatizing language (SL) in EHR notes on mortality prediction using a Transformer-based deep learning model and explainable AI (XAI) techniques. Our findings demonstrate that SL written by clinicians adversely affects AI performance, particularly so for black patients, highlighting SL as a source of racial disparity in AI model development. To explore an operationally efficient way to mitigate SL's impact, we investigate patterns in the generation of SL through a clinicians' collaborative network, identifying central clinicians as having a stronger impact on racial disparity in the AI model. We find that removing SL written by central clinicians is a more efficient bias reduction strategy than eliminating all SL in the entire corpus of data. This study provides actionable insights for responsible AI development and contributes to understanding clinician behavior and EHR note writing in healthcare.

Explaining RL Decisions with Trajectories

Shripad Vilasrao Deshmukh, Arpan Dasgupta, Balaji Krishnamurthy +4

Explanation is a key component for the adoption of reinforcement learning (RL) in many real-world decision-making problems. In the literature, the explanation is often provided by saliency attribution to the features of the RL agent's state. In this work, we propose a complementary approach to these explanations, particularly for offline RL, where we attribute the policy decisions of a trained RL agent to the trajectories encountered by it during training. To do so, we encode trajectories in offline training data individually as well as collectively (encoding a set of trajectories). We then attribute policy decisions to a set of trajectories in this encoded space by estimating the sensitivity of the decision with respect to that set. Further, we demonstrate the effectiveness of the proposed approach in terms of quality of attributions as well as practical scalability in diverse environments that involve both discrete and continuous state and action spaces such as grid-worlds, video games (Atari) and continuous control (MuJoCo). We also conduct a human study on a simple navigation task to observe how their understanding of the task compares with data attributed for a trained RL policy. Keywords -- Explainable AI, Verifiability of AI Decisions, Explainable RL.

Towards Explainable AI Writing Assistants for Non-native English Speakers

Yewon Kim, Mina Lee, Donghwi Kim +1

We highlight the challenges faced by non-native speakers when using AI writing assistants to paraphrase text. Through an interview study with 15 non-native English speakers (NNESs) with varying levels of English proficiency, we observe that they face difficulties in assessing paraphrased texts generated by AI writing assistants, largely due to the lack of explanations accompanying the suggested paraphrases. Furthermore, we examine their strategies to assess AI-generated texts in the absence of such explanations. Drawing on the needs of NNESs identified in our interview, we propose four potential user interfaces to enhance the writing experience of NNESs using AI writing assistants. The proposed designs focus on incorporating explanations to better support NNESs in understanding and evaluating the AI-generated paraphrasing suggestions.

With the advent of deep learning, text generation language models have improved dramatically, with text at a similar level as human-written text. This can lead to rampant misinformation because content can now be created cheaply and distributed quickly. Automated claim verification methods exist to validate claims, but they lack foundational data and often use mainstream news as evidence sources that are strongly biased towards a specific agenda. Current claim verification methods use deep neural network models and complex algorithms for a high classification accuracy but it is at the expense of model explainability. The models are black-boxes and their decision-making process and the steps it took to arrive at a final prediction are obfuscated from the user. We introduce a novel claim verification approach, namely: ExClaim, that attempts to provide an explainable claim verification system with foundational evidence. Inspired by the legal system, ExClaim leverages rationalization to provide a verdict for the claim and justifies the verdict through a natural language explanation (rationale) to describe the model's decision-making process. ExClaim treats the verdict classification task as a question-answer problem and achieves a performance of 0.93 F1 score. It provides subtasks explanations to also justify the intermediate outcomes. Statistical and Explainable AI (XAI) evaluations are conducted to ensure valid and trustworthy outcomes. Ensuring claim verification systems are assured, rational, and explainable is an essential step toward improving Human-AI trust and the accessibility of black-box systems.