Research Papers

Explainable artificial intelligence (XAI) has helped elucidate the internal mechanisms of machine learning algorithms, bolstering their reliability by demonstrating the basis of their predictions. Several XAI models consider causal relationships to explain models by examining the input-output relationships of prediction models and the dependencies between features. The majority of these models have been based their explanations on counterfactual probabilities, assuming that the causal graph is known. However, this assumption complicates the application of such models to real data, given that the causal relationships between features are unknown in most cases. Thus, this study proposed a novel XAI framework that relaxed the constraint that the causal graph is known. This framework leveraged counterfactual probabilities and additional prior information on causal structure, facilitating the integration of a causal graph estimated through causal discovery methods and a black-box classification model. Furthermore, explanatory scores were estimated based on counterfactual probabilities. Numerical experiments conducted employing artificial data confirmed the possibility of estimating the explanatory score more accurately than in the absence of a causal graph. Finally, as an application to real data, we constructed a classification model of credit ratings assigned by Shiga Bank, Shiga prefecture, Japan. We demonstrated the effectiveness of the proposed method in cases where the causal graph is unknown.

Given the impressive capabilities of recent Large Language Models (LLMs), we investigate and benchmark the most popular proprietary and different sized open source models on the task of explicit instruction following in conflicting situations, e.g. overrides. These include the ability of the model to override the knowledge within the weights of the model, the ability to override (or moderate) extracted knowledge in the prompt, and lastly the ability to perform a full jailbreak. Experimentation performed suggest several key findings to improve instruction following - larger models perform the best in following instructions that override internal and contextual instructions, and are obedient, even to a fault. When scaling to longer contexts via rope scaling, a significant buffer needs to be maintained from the edge of the perplexity cliff in order to maintain instruction following capabilities. Finally, we observe improving instruction following, and subsequently instruction overrides/jailbreaks, is fundamentally at odds with the ability of a language model to follow given safety filters or guidelines. Thus, we postulate the most effective approach for safe, trustworthy AI should be dealt external to the LLM itself.

Explainable AI (XAI) has demonstrated the potential to help reinforcement learning (RL) practitioners to understand how RL models work. However, XAI for users who do not have RL expertise (non-RL experts), has not been studied sufficiently. This results in a difficulty for the non-RL experts to participate in the fundamental discussion of how RL models should be designed for an incoming society where humans and AI coexist. Solving such a problem would enable RL experts to communicate with the non-RL experts in producing machine learning solutions that better fit our society. We argue that abstracted trajectories, that depicts transitions between the major states of the RL model, will be useful for non-RL experts to build a mental model of the agents. Our early results suggest that by leveraging a visualization of the abstracted trajectories, users without RL expertise are able to infer the behavior patterns of RL.

Bridging Human Concepts and Computer Vision for Explainable Face Verification

Miriam Doh, Caroline Mazini Rodrigues, Nicolas Boutry +3

With Artificial Intelligence (AI) influencing the decision-making process of sensitive applications such as Face Verification, it is fundamental to ensure the transparency, fairness, and accountability of decisions. Although Explainable Artificial Intelligence (XAI) techniques exist to clarify AI decisions, it is equally important to provide interpretability of these decisions to humans. In this paper, we present an approach to combine computer and human vision to increase the explanation's interpretability of a face verification algorithm. In particular, we are inspired by the human perceptual process to understand how machines perceive face's human-semantic areas during face comparison tasks. We use Mediapipe, which provides a segmentation technique that identifies distinct human-semantic facial regions, enabling the machine's perception analysis. Additionally, we adapted two model-agnostic algorithms to provide human-interpretable insights into the decision-making processes.

Generative artificial intelligence (AI) is poised to reshape the way individuals communicate and interact. While this form of AI has the potential to efficiently make numerous human decisions, there is limited understanding of how individuals respond to its use in social interaction. In particular, it remains unclear how individuals engage with algorithms when the interaction entails consequences for other people. Here, we report the results of a large-scale pre-registered online experiment (N = 3,552) indicating diminished fairness, trust, trustworthiness, cooperation, and coordination by human players in economic twoplayer games, when the decision of the interaction partner is taken over by ChatGPT. On the contrary, we observe no adverse welfare effects when individuals are uncertain about whether they are interacting with a human or generative AI. Therefore, the promotion of AI transparency, often suggested as a solution to mitigate the negative impacts of generative AI on society, shows a detrimental effect on welfare in our study. Concurrently, participants frequently delegate decisions to ChatGPT, particularly when the AI's involvement is undisclosed, and individuals struggle to discern between AI and human decisions.

Counterfactuals have been established as a popular explainability technique which leverages a set of minimal edits to alter the prediction of a classifier. When considering conceptual counterfactuals on images, the edits requested should correspond to salient concepts present in the input data. At the same time, conceptual distances are defined by knowledge graphs, ensuring the optimality of conceptual edits. In this work, we extend previous endeavors on graph edits as counterfactual explanations by conducting a comparative study which encompasses both supervised and unsupervised Graph Neural Network (GNN) approaches. To this end, we pose the following significant research question: should we represent input data as graphs, which is the optimal GNN approach in terms of performance and time efficiency to generate minimal and meaningful counterfactual explanations for black-box image classifiers?

The rising use of Artificial Intelligence (AI) in human detection on Edge camera systems has led to accurate but complex models, challenging to interpret and debug. Our research presents a diagnostic method using Explainable AI (XAI) for model debugging, with expert-driven problem identification and solution creation. Validated on the Bytetrack model in a real-world office Edge network, we found the training dataset as the main bias source and suggested model augmentation as a solution. Our approach helps identify model biases, essential for achieving fair and trustworthy models.

Audio signal segmentation is a key task for automatic audio indexing. It consists of detecting the boundaries of class-homogeneous segments in the signal. In many applications, explainable AI is a vital process for transparency of decision-making with machine learning. In this paper, we propose an explainable multilabel segmentation model that solves speech activity (SAD), music (MD), noise (ND), and overlapped speech detection (OSD) simultaneously. This proxy uses the non-negative matrix factorization (NMF) to map the embedding used for the segmentation to the frequency domain. Experiments conducted on two datasets show similar performances as the pre-trained black box model while showing strong explainability features. Specifically, the frequency bins used for the decision can be easily identified at both the segment level (local explanations) and global level (class prototypes).

Biased datasets are ubiquitous and present a challenge for machine learning. For a number of categories on a dataset that are equally important but some are sparse and others are common, the learning algorithms will favor the ones with more presence. The problem of biased datasets is especially sensitive when dealing with minority people groups. How can we, from biased data, generate algorithms that treat every person equally? This work explores one way to mitigate bias using a debiasing variational autoencoder with experiments on facial expression recognition.

Vision Transformers (ViTs) have achieved state-of-the-art results on various computer vision tasks, including 3D object detection. However, their end-to-end implementation also makes ViTs less explainable, which can be a challenge for deploying them in safety-critical applications, such as autonomous driving, where it is important for authorities, developers, and users to understand the model's reasoning behind its predictions. In this paper, we propose a novel method for generating saliency maps for a DetR-like ViT with multiple camera inputs used for 3D object detection. Our method is based on the raw attention and is more efficient than gradient-based methods. We evaluate the proposed method on the nuScenes dataset using extensive perturbation tests and show that it outperforms other explainability methods in terms of visual quality and quantitative metrics. We also demonstrate the importance of aggregating attention across different layers of the transformer. Our work contributes to the development of explainable AI for ViTs, which can help increase trust in AI applications by establishing more transparency regarding the inner workings of AI models.

Deep neural networks have been widely adopted in numerous domains due to their high performance and accessibility to developers and application-specific end-users. Fundamental to image-based applications is the development of Convolutional Neural Networks (CNNs), which possess the ability to automatically extract features from data. However, comprehending these complex models and their learned representations, which typically comprise millions of parameters and numerous layers, remains a challenge for both developers and end-users. This challenge arises due to the absence of interpretable and transparent tools to make sense of black-box models. There exists a growing body of Explainable Artificial Intelligence (XAI) literature, including a collection of methods denoted Class Activation Maps (CAMs), that seek to demystify what representations the model learns from the data, how it informs a given prediction, and why it, at times, performs poorly in certain tasks. We propose a novel XAI visualization method denoted CAManim that seeks to simultaneously broaden and focus end-user understanding of CNN predictions by animating the CAM-based network activation maps through all layers, effectively depicting from end-to-end how a model progressively arrives at the final layer activation. Herein, we demonstrate that CAManim works with any CAM-based method and various CNN architectures. Beyond qualitative model assessments, we additionally propose a novel quantitative assessment that expands upon the Remove and Debias (ROAD) metric, pairing the qualitative end-to-end network visual explanations assessment with our novel quantitative "yellow brick ROAD" assessment (ybROAD). This builds upon prior research to address the increasing demand for interpretable, robust, and transparent model assessment methodology, ultimately improving an end-user's trust in a given model's predictions.

As AI/ML models, including Large Language Models, continue to scale with massive datasets, so does their consumption of undeniably limited natural resources, and impact on society. In this collaboration between AI, Sustainability, HCI and legal researchers, we aim to enable a transition to sustainable AI development by enabling stakeholders across the AI value chain to assess and quantitfy the environmental and societal impact of AI. We present the ESG Digital and Green Index (DGI), which offers a dashboard for assessing a company's performance in achieving sustainability targets. This includes monitoring the efficiency and sustainable use of limited natural resources related to AI technologies (water, electricity, etc). It also addresses the societal and governance challenges related to AI. The DGI creates incentives for companies to align their pathway with the Sustainable Development Goals (SDGs). The value, challenges and limitations of our methodology and findings are discussed in the paper.