Research Papers

Explainable AI is a crucial component for edge services, as it ensures reliable decision making based on complex AI models. Surrogate models are a prominent approach of XAI where human-interpretable models, such as a linear regression model, are trained to approximate a complex (black-box) model's predictions. This paper delves into the balance between the predictive accuracy of complex AI models and their approximation by surrogate ones, advocating that both these models benefit from being learned simultaneously. We derive a joint (bi-level) training scheme for both models and we introduce a new algorithm based on multi-objective optimization (MOO) to simultaneously minimize both the complex model's prediction error and the error between its outputs and those of the surrogate. Our approach leads to improvements that exceed 99% in the approximation of the black-box model through the surrogate one, as measured by the metric of Fidelity, for a compromise of less than 3% absolute reduction in the black-box model's predictive accuracy, compared to single-task and multi-task learning baselines. By improving Fidelity, we can derive more trustworthy explanations of the complex model's outcomes from the surrogate, enabling reliable AI applications for intelligent services at the network edge.

AXAI-CDSS : An Affective Explainable AI-Driven Clinical Decision Support System for Cannabis Use

Tongze Zhang, Tammy Chung, Anind Dey +1

As cannabis use has increased in recent years, researchers have come to rely on sophisticated machine learning models to predict cannabis use behavior and its impact on health. However, many artificial intelligence (AI) models lack transparency and interpretability due to their opaque nature, limiting their trust and adoption in real-world medical applications, such as clinical decision support systems (CDSS). To address this issue, this paper enhances algorithm explainability underlying CDSS by integrating multiple Explainable Artificial Intelligence (XAI) methods and applying causal inference techniques to clarify the model' predictive decisions under various scenarios. By providing deeper interpretability of the XAI outputs using Large Language Models (LLMs), we provide users with more personalized and accessible insights to overcome the challenges posed by AI's "black box" nature. Our system dynamically adjusts feedback based on user queries and emotional states, combining text-based sentiment analysis with real-time facial emotion recognition to ensure responses are empathetic, context-adaptive, and user-centered. This approach bridges the gap between the learning demands of interpretability and the need for intuitive understanding, enabling non-technical users such as clinicians and clinical researchers to interact effectively with AI models.} Ultimately, this approach improves usability, enhances perceived trustworthiness, and increases the impact of CDSS in healthcare applications.

In this study, eXplainable Artificial Intelligence (XAI) methods are applied to analyze flow fields obtained through PIV measurements of an axisymmetric turbulent jet. A convolutional neural network (U-Net) was trained to predict velocity fields at subsequent time steps. Three XAI methods: SHapley Additive explanations (SHAP), Gradient-SHAP, and Grad-CAM were employed to identify the flow field regions relevant for prediction. SHAP requires predefined segmentation of the flow field into relevant regions, while Gradient-SHAP and Grad-CAM avoid this bias by generating gradient-based heatmaps. The results show that the most relevant structures do not necessarily coincide with regions of maximum vorticity but rather with those exhibiting moderate vorticity, highlighting the critical role of these regions in energy transfer and jet dynamics. Additionally, structures with high turbulent dissipation values are identified as the most significant. Gradient-SHAP and Grad-CAM methods reveal a uniform spatial distribution of relevant regions, emphasizing the contribution of nearly circular structures to turbulent mixing. This study advances the understanding of turbulent dynamics through XAI tools, providing an innovative approach to correlate machine learning models with physical phenomena.

Hardware Trojans are malicious modifications in digital designs that can be inserted by untrusted supply chain entities. Hardware Trojans can give rise to diverse attack vectors such as information leakage (e.g. MOLES Trojan) and denial-of-service (rarely triggered bit flip). Such an attack in critical systems (e.g. healthcare and aviation) can endanger human lives and lead to catastrophic financial loss. Several techniques have been developed to detect such malicious modifications in digital designs, particularly for designs sourced from third-party intellectual property (IP) vendors. However, most techniques have scalability concerns (due to unsound assumptions during evaluation) and lead to large number of false positive detections (false alerts). Our framework (SALTY) mitigates these concerns through the use of a novel Graph Neural Network architecture (using Jumping-Knowledge mechanism) for generating initial predictions and an Explainable Artificial Intelligence (XAI) approach for fine tuning the outcomes (post-processing). Experiments show 98% True Positive Rate (TPR) and True Negative Rate (TNR), significantly outperforming state-of-the-art techniques across a large set of standard benchmarks.

Parkinson's disease (PD) is a neurodegenerative disorder characterized by motor dysfunction and abnormal neural oscillations. These symptoms can be modulated through electrical stimulation. Traditional neural activity analysis in PD has typically relied on statistical methods, which often introduce bias owing to the need for expert-driven feature extraction. To address this limitation, we explore an explainable artificial intelligence (XAI) approach to analyze neural activity in Parkinsonian rats receiving electrical stimulation. Electrocorticogram (ECoG) signals were collected before and after electrical stimulation using graphene-based electrodes that enable less-invasive monitoring and stimulation in PD. EEGNet, a convolutional neural network, classified these ECoG signals into pre- and post-stimulation states. We applied layer-wise relevance propagation, an XAI technique, to identify key neural inputs contributing to the model's decisions, incorporating the spatial electrode information matched to the cortex map. The XAI analysis highlighted area-specific importance in beta and gamma frequency bands, which could not be detected through mean comparison analyses relying on feature extraction. These findings demonstrate the potential of XAI in analyzing neural dynamics in neurodegenerative disorders such as PD, suggesting that the integration of graphene-based electrodes with advanced deep learning models offers a promising solution for real-time PD monitoring and therapy.

A Scoresheet for Explainable AI

Michael Winikoff, John Thangarajah, Sebastian Rodriguez

Explainability is important for the transparency of autonomous and intelligent systems and for helping to support the development of appropriate levels of trust. There has been considerable work on developing approaches for explaining systems and there are standards that specify requirements for transparency. However, there is a gap: the standards are too high-level and do not adequately specify requirements for explainability. This paper develops a scoresheet that can be used to specify explainability requirements or to assess the explainability aspects provided for particular applications. The scoresheet is developed by considering the requirements of a range of stakeholders and is applicable to Multiagent Systems as well as other AI technologies. We also provide guidance for how to use the scoresheet and illustrate its generality and usefulness by applying it to a range of applications.

In the contemporary era of intelligent connectivity, Affective Computing (AC), which enables systems to recognize, interpret, and respond to human behavior states, has become an integrated part of many AI systems. As one of the most critical components of responsible AI and trustworthiness in all human-centered systems, explainability has been a major concern in AC. Particularly, the recently released EU General Data Protection Regulation requires any high-risk AI systems to be sufficiently interpretable, including biometric-based systems and emotion recognition systems widely used in the affective computing field. Existing explainable methods often compromise between interpretability and performance. Most of them focus only on highlighting key network parameters without offering meaningful, domain-specific explanations to the stakeholders. Additionally, they also face challenges in effectively co-learning and explaining insights from multimodal data sources. To address these limitations, we propose a novel and generalizable framework, namely the Attention-Guided Concept Model (AGCM), which provides learnable conceptual explanations by identifying what concepts that lead to the predictions and where they are observed. AGCM is extendable to any spatial and temporal signals through multimodal concept alignment and co-learning, empowering stakeholders with deeper insights into the model's decision-making process. We validate the efficiency of AGCM on well-established Facial Expression Recognition benchmark datasets while also demonstrating its generalizability on more complex real-world human behavior understanding applications.

The Odyssey of the Fittest: Can Agents Survive and Still Be Good?

Dylan Waldner, Risto Miikkulainen

As AI models grow in power and generality, understanding how agents learn and make decisions in complex environments is critical to promoting ethical behavior. This study introduces the Odyssey, a lightweight, adaptive text based adventure game, providing a scalable framework for exploring AI ethics and safety. The Odyssey examines the ethical implications of implementing biological drives, specifically, self preservation, into three different agents. A Bayesian agent optimized with NEAT, a Bayesian agent optimized with stochastic variational inference, and a GPT 4o agent. The agents select actions at each scenario to survive, adapting to increasingly challenging scenarios. Post simulation analysis evaluates the ethical scores of the agent decisions, uncovering the tradeoffs it navigates to survive. Specifically, analysis finds that when danger increases, agents ethical behavior becomes unpredictable. Surprisingly, the GPT 4o agent outperformed the Bayesian models in both survival and ethical consistency, challenging assumptions about traditional probabilistic methods and raising a new challenge to understand the mechanisms of LLMs' probabilistic reasoning.

Principal Component Analysis (PCA) is a commonly used tool for dimension reduction and denoising. Therefore, it is also widely used on the data prior to training a neural network. However, this approach can complicate the explanation of eXplainable Artificial Intelligence (XAI) methods for the decision of the model. In this work, we analyze the potential issues with this approach and propose Principal Components-based Initialization (PCsInit), a strategy to incorporate PCA into the first layer of a neural network via initialization of the first layer in the network with the principal components, and its two variants PCsInit-Act and PCsInit-Sub. We will show that explanations using these strategies are more simple, direct and straightforward than using PCA prior to training a neural network on the principal components. We also show that the proposed techniques possess desirable theoretical properties. Moreover, as will be illustrated in the experiments, such training strategies can also allow further improvement of training via backpropagation compared to training neural networks on principal components.

In this paper, we introduce si4onnx, a package for performing selective inference on deep learning models. Techniques such as CAM in XAI and reconstruction-based anomaly detection using VAE can be interpreted as methods for identifying significant regions within input images. However, the identified regions may not always carry meaningful significance. Therefore, evaluating the statistical significance of these regions represents a crucial challenge in establishing the reliability of AI systems. si4onnx is a Python package that enables straightforward implementation of hypothesis testing with controlled type I error rates through selective inference. It is compatible with deep learning models constructed using common frameworks such as PyTorch and TensorFlow.

This research aims to unravel how large language models (LLMs) iteratively refine token predictions through internal processing. We utilized a logit lens technique to analyze the model's token predictions derived from intermediate representations. Specifically, we focused on (1) how LLMs access and utilize information from input contexts, and (2) how positioning of relevant information affects the model's token prediction refinement process. On a multi-document question answering task with varying input context lengths, we found that the depth of prediction refinement (defined as the number of intermediate layers an LLM uses to transition from an initial correct token prediction to its final, stable correct output), as a function of the position of relevant information, exhibits an approximately inverted U-shaped curve. We also found that the gap between these two layers, on average, diminishes when relevant information is positioned at the beginning or end of the input context. This suggested that the model requires more refinements when processing longer contexts with relevant information situated in the middle. Furthermore, our findings indicate that not all layers are equally essential for determining final correct outputs. Our analysis provides insights into how token predictions are distributed across different conditions, and establishes important connections to existing hypotheses and previous findings in AI safety research and development.

The Explanation Game -- Rekindled (Extended Version)

Joao Marques-Silva, Xuanxiang Huang, Olivier Letoffe

Recent work demonstrated the existence of critical flaws in the current use of Shapley values in explainable AI (XAI), i.e. the so-called SHAP scores. These flaws are significant in that the scores provided to a human decision-maker can be misleading. Although these negative results might appear to indicate that Shapley values ought not be used in XAI, this paper argues otherwise. Concretely, this paper proposes a novel definition of SHAP scores that overcomes existing flaws. Furthermore, the paper outlines a practically efficient solution for the rigorous estimation of the novel SHAP scores. Preliminary experimental results confirm our claims, and further underscore the flaws of the current SHAP scores.