Research Papers

The advent of advanced AI underscores the urgent need for comprehensive safety evaluations, necessitating collaboration across communities (i.e., AI, software engineering, and governance). However, divergent practices and terminologies across these communities, combined with the complexity of AI systems-of which models are only a part-and environmental affordances (e.g., access to tools), obstruct effective communication and comprehensive evaluation. This paper proposes a framework for AI system evaluation comprising three components: 1) harmonised terminology to facilitate communication across communities involved in AI safety evaluation; 2) a taxonomy identifying essential elements for AI system evaluation; 3) a mapping between AI lifecycle, stakeholders, and requisite evaluations for accountable AI supply chain. This framework catalyses a deeper discourse on AI system evaluation beyond model-centric approaches.

In the dynamic and ever-changing domain of Unmanned Aerial Vehicles (UAVs), the utmost importance lies in guaranteeing resilient and lucid security measures. This study highlights the necessity of implementing a Zero Trust Architecture (ZTA) to enhance the security of unmanned aerial vehicles (UAVs), hence departing from conventional perimeter defences that may expose vulnerabilities. The Zero Trust Architecture (ZTA) paradigm requires a rigorous and continuous process of authenticating all network entities and communications. The accuracy of our methodology in detecting and identifying unmanned aerial vehicles (UAVs) is 84.59\%. This is achieved by utilizing Radio Frequency (RF) signals within a Deep Learning framework, a unique method. Precise identification is crucial in Zero Trust Architecture (ZTA), as it determines network access. In addition, the use of eXplainable Artificial Intelligence (XAI) tools such as SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) contributes to the improvement of the model's transparency and interpretability. Adherence to Zero Trust Architecture (ZTA) standards guarantees that the classifications of unmanned aerial vehicles (UAVs) are verifiable and comprehensible, enhancing security within the UAV field.

Despite its technological breakthroughs, eXplainable Artificial Intelligence (XAI) research has limited success in producing the {\em effective explanations} needed by users. In order to improve XAI systems' usability, practical interpretability, and efficacy for real users, the emerging area of {\em Explainable Interfaces} (EIs) focuses on the user interface and user experience design aspects of XAI. This paper presents a systematic survey of 53 publications to identify current trends in human-XAI interaction and promising directions for EI design and development. This is among the first systematic survey of EI research.

Generative AI, exemplified by models like transformers, has opened up new possibilities in various domains but also raised concerns about fairness, transparency and reliability, especially in fields like medicine and law. This paper emphasizes the urgency of ensuring fairness and quality in these domains through generative AI. It explores using cryptographic techniques, particularly Zero-Knowledge Proofs (ZKPs), to address concerns regarding performance fairness and accuracy while protecting model privacy. Applying ZKPs to Machine Learning models, known as ZKML (Zero-Knowledge Machine Learning), enables independent validation of AI-generated content without revealing sensitive model information, promoting transparency and trust. ZKML enhances AI fairness by providing cryptographic audit trails for model predictions and ensuring uniform performance across users. We introduce snarkGPT, a practical ZKML implementation for transformers, to empower users to verify output accuracy and quality while preserving model privacy. We present a series of empirical results studying snarkGPT's scalability and performance to assess the feasibility and challenges of adopting a ZKML-powered approach to capture quality and performance fairness problems in generative AI models.

There are concerns about the reliability and safety of artificial intelligence (AI) based on sub-symbolic neural networks because its decisions cannot be explained explicitly. This is the black box problem of modern AI. At the same time, symbolic AI has the nature of a white box and is able to ensure the reliability and safety of its decisions. However, several problems prevent the widespread use of symbolic AI: the opacity of mathematical models and natural language terms, the lack of a unified ontology, and the combinatorial explosion of search capabilities. To solve the black-box problem of AI, we propose eXplicitly eXplainable AI (XXAI) - a fully transparent white-box AI based on deterministic logical cellular automata whose rules are derived from the first principles of the general theory of the relevant domain. In this case, the general theory of the domain plays the role of a knowledge base for deriving the inferences of the cellular automata. A cellular automaton implements parallel multi-level logical inference at all levels of organization - from local interactions of the element base to the system as a whole. Our verification of several ecological hypotheses sets a precedent for the successful implementation of the proposed solution. XXAI is able to automatically verify the reliability, security and ethics of sub-symbolic neural network solutions in both the final and training phases. In this article, we present precedents for the successful implementation of XXAI, the theoretical and methodological foundations for its further development, and discuss prospects for the future.

Motivation: We explored how explainable artificial intelligence (XAI) can help to shed light into the inner workings of neural networks for protein function prediction, by extending the widely used XAI method of integrated gradients such that latent representations inside of transformer models, which were finetuned to Gene Ontology term and Enzyme Commission number prediction, can be inspected too. Results: The approach enabled us to identify amino acids in the sequences that the transformers pay particular attention to, and to show that these relevant sequence parts reflect expectations from biology and chemistry, both in the embedding layer and inside of the model, where we identified transformer heads with a statistically significant correspondence of attribution maps with ground truth sequence annotations (e.g. transmembrane regions, active sites) across many proteins. Availability and Implementation: Source code can be accessed at https://github.com/markuswenzel/xai-proteins .

A New Perspective on Evaluation Methods for Explainable Artificial Intelligence (XAI)

Timo Speith, Markus Langer

Within the field of Requirements Engineering (RE), the increasing significance of Explainable Artificial Intelligence (XAI) in aligning AI-supported systems with user needs, societal expectations, and regulatory standards has garnered recognition. In general, explainability has emerged as an important non-functional requirement that impacts system quality. However, the supposed trade-off between explainability and performance challenges the presumed positive influence of explainability. If meeting the requirement of explainability entails a reduction in system performance, then careful consideration must be given to which of these quality aspects takes precedence and how to compromise between them. In this paper, we critically examine the alleged trade-off. We argue that it is best approached in a nuanced way that incorporates resource availability, domain characteristics, and considerations of risk. By providing a foundation for future research and best practices, this work aims to advance the field of RE for AI.

Explanations in XAI are typically developed by AI experts and focus on algorithmic transparency and the inner workings of AI systems. Research has shown that such explanations do not meet the needs of users who do not have AI expertise. As a result, explanations are often ineffective in making system decisions interpretable and understandable. We aim to strengthen a socio-technical view of AI by following a Human-Centered Explainable Artificial Intelligence (HC-XAI) approach, which investigates the explanation needs of end-users (i.e., subject matter experts and lay users) in specific usage contexts. One of the most influential works in this area is the XAI Question Bank (XAIQB) by Liao et al. The authors propose a set of questions that end-users might ask when using an AI system, which in turn is intended to help developers and designers identify and address explanation needs. Although the XAIQB is widely referenced, there are few reports of its use in practice. In particular, it is unclear to what extent the XAIQB sufficiently captures the explanation needs of end-users and what potential problems exist in the practical application of the XAIQB. To explore these open questions, we used the XAIQB as the basis for analyzing 12 think-aloud software explorations with subject matter experts. We investigated the suitability of the XAIQB as a tool for identifying explanation needs in a specific usage context. Our analysis revealed a number of explanation needs that were missing from the question bank, but that emerged repeatedly as our study participants interacted with an AI system. We also found that some of the XAIQB questions were difficult to distinguish and required interpretation during use. Our contribution is an extension of the XAIQB with 11 new questions. In addition, we have expanded the descriptions of all new and existing questions to facilitate their use.

In recent years, the community of 'explainable artificial intelligence' (XAI) has created a vast body of methods to bridge a perceived gap between model 'complexity' and 'interpretability'. However, a concrete problem to be solved by XAI methods has not yet been formally stated. As a result, XAI methods are lacking theoretical and empirical evidence for the 'correctness' of their explanations, limiting their potential use for quality-control and transparency purposes. At the same time, Haufe et al. (2014) showed, using simple toy examples, that even standard interpretations of linear models can be highly misleading. Specifically, high importance may be attributed to so-called suppressor variables lacking any statistical relation to the prediction target. This behavior has been confirmed empirically for a large array of XAI methods in Wilming et al. (2022). Here, we go one step further by deriving analytical expressions for the behavior of a variety of popular XAI methods on a simple two-dimensional binary classification problem involving Gaussian class-conditional distributions. We show that the majority of the studied approaches will attribute non-zero importance to a non-class-related suppressor feature in the presence of correlated noise. This poses important limitations on the interpretations and conclusions that the outputs of these XAI methods can afford.

Explainable AI (XAI) techniques have been widely used to help explain and understand the output of deep learning models in fields such as image classification and Natural Language Processing. Interest in using XAI techniques to explain deep learning-based automatic speech recognition (ASR) is emerging. but there is not enough evidence on whether these explanations can be trusted. To address this, we adapt a state-of-the-art XAI technique from the image classification domain, Local Interpretable Model-Agnostic Explanations (LIME), to a model trained for a TIMIT-based phoneme recognition task. This simple task provides a controlled setting for evaluation while also providing expert annotated ground truth to assess the quality of explanations. We find a variant of LIME based on time partitioned audio segments, that we propose in this paper, produces the most reliable explanations, containing the ground truth 96% of the time in its top three audio segments.

Alzheimer's Disease (AD) is the world leading cause of dementia, a progressively impairing condition leading to high hospitalization rates and mortality. To optimize the diagnostic process, numerous efforts have been directed towards the development of deep learning approaches (DL) for the automatic AD classification. However, their typical black box outline has led to low trust and scarce usage within clinical frameworks. In this work, we propose two state-of-the art DL models, trained respectively on structural MRI (ResNet18) and brain connectivity matrixes (BC-GCN-SE) derived from diffusion data. The models were initially evaluated in terms of classification accuracy. Then, results were analyzed using an Explainable Artificial Intelligence (XAI) approach (Grad-CAM) to measure the level of interpretability of both models. The XAI assessment was conducted across 132 brain parcels, extracted from a combination of the Harvard-Oxford and AAL brain atlases, and compared to well-known pathological regions to measure adherence to domain knowledge. Results highlighted acceptable classification performance as compared to the existing literature (ResNet18: TPRmedian = 0.817, TNRmedian = 0.816; BC-GCN-SE: TPRmedian = 0.703, TNRmedian = 0.738). As evaluated through a statistical test (p < 0.05) and ranking of the most relevant parcels (first 15%), Grad-CAM revealed the involvement of target brain areas for both the ResNet18 and BC-GCN-SE models: the medial temporal lobe and the default mode network. The obtained interpretabilities were not without limitations. Nevertheless, results suggested that combining different imaging modalities may result in increased classification performance and model reliability. This could potentially boost the confidence laid in DL models and favor their wide applicability as aid diagnostic tools.

Counterfactual explanations have become a mainstay of the XAI field. This particularly intuitive statement allows the user to understand what small but necessary changes would have to be made to a given situation in order to change a model prediction. The quality of a counterfactual depends on several criteria: realism, actionability, validity, robustness, etc. In this paper, we are interested in the notion of robustness of a counterfactual. More precisely, we focus on robustness to counterfactual input changes. This form of robustness is particularly challenging as it involves a trade-off between the robustness of the counterfactual and the proximity with the example to explain. We propose a new framework, CROCO, that generates robust counterfactuals while managing effectively this trade-off, and guarantees the user a minimal robustness. An empirical evaluation on tabular datasets confirms the relevance and effectiveness of our approach.