Research Papers

Explainability as a Requirement for Hardware: Introducing Explainable Hardware (XHW)

Timo Speith, Julian Speith, Steffen Becker +3

In today's age of digital technology, ethical concerns regarding computing systems are increasing. While the focus of such concerns currently is on requirements for software, this article spotlights the hardware domain, specifically microchips. For example, the opaqueness of modern microchips raises security issues, as malicious actors can manipulate them, jeopardizing system integrity. As a consequence, governments invest substantially to facilitate a secure microchip supply chain. To combat the opaqueness of hardware, this article introduces the concept of Explainable Hardware (XHW). Inspired by and building on previous work on Explainable AI (XAI) and explainable software systems, we develop a framework for achieving XHW comprising relevant stakeholders, requirements they might have concerning hardware, and possible explainability approaches to meet these requirements. Through an exploratory survey among 18 hardware experts, we showcase applications of the framework and discover potential research gaps. Our work lays the foundation for future work and structured debates on XHW.

The lack of established rules and regulations in cyberspace is attributed to the absence of agreed-upon ethical principles, making it difficult to establish accountability, regulations, and laws. Addressing this challenge requires examining cyberspace from fundamental philosophical principles. This work focuses on the ethics of using defensive deception in cyberspace, proposing a doctrine of cyber effect that incorporates five ethical principles: goodwill, deontology, no-harm, transparency, and fairness. To guide the design of defensive cyber deception, we develop a reasoning framework, the game of ethical duplicity, which is consistent with the doctrine. While originally intended for cyber deception, this doctrine has broader applicability, including for ethical issues such as AI accountability and controversies related to YouTube recommendations. By establishing ethical principles, we can promote greater accountability, regulation, and protection in the digital realm.

AI ethics is an emerging field with multiple, competing narratives about how to best solve the problem of building human values into machines. Two major approaches are focused on bias and compliance, respectively. But neither of these ideas fully encompasses ethics: using moral principles to decide how to act in a particular situation. Our method posits that the way data is labeled plays an essential role in the way AI behaves, and therefore in the ethics of machines themselves. The argument combines a fundamental insight from ethics (i.e. that ethics is about values) with our practical experience building and scaling machine learning systems. We want to build AI that is actually ethical by first addressing foundational concerns: how to build good systems, how to define what is good in relation to system architecture, and who should provide that definition. Building ethical AI creates a foundation of trust between a company and the users of that platform. But this trust is unjustified unless users experience the direct value of ethical AI. Until users have real control over how algorithms behave, something is missing in current AI solutions. This causes massive distrust in AI, and apathy towards AI ethics solutions. The scope of this paper is to propose an alternative path that allows for the plurality of values and the freedom of individual expression. Both are essential for realizing true moral character.

Using Explainable AI to Cross-Validate Socio-economic Disparities Among Covid-19 Patient Mortality

Li Shi, Redoan Rahman, Esther Melamed +3

This paper applies eXplainable Artificial Intelligence (XAI) methods to investigate the socioeconomic disparities in COVID patient mortality. An Extreme Gradient Boosting (XGBoost) prediction model is built based on a de-identified Austin area hospital dataset to predict the mortality of COVID-19 patients. We apply two XAI methods, Shapley Additive exPlanations (SHAP) and Locally Interpretable Model Agnostic Explanations (LIME), to compare the global and local interpretation of feature importance. This paper demonstrates the advantages of using XAI which shows the feature importance and decisive capability. Furthermore, we use the XAI methods to cross-validate their interpretations for individual patients. The XAI models reveal that Medicare financial class, older age, and gender have high impact on the mortality prediction. We find that LIME local interpretation does not show significant differences in feature importance comparing to SHAP, which suggests pattern confirmation. This paper demonstrates the importance of XAI methods in cross-validation of feature attributions.

With the increased usage of artificial intelligence (AI), it is imperative to understand how these models work internally. These needs have led to the development of a new field called eXplainable artificial intelligence (XAI). This field consists of on a set of techniques that allows us to theoretically determine the cause of the AI decisions. One main issue of XAI is how to verify the works on this field, taking into consideration the lack of ground truth (GT). In this study, we propose a new method to generate datasets with GT. We conducted a set of experiments that compared our GT with real model explanations and obtained excellent results confirming that our proposed method is correct.

GCI: A (G)raph (C)oncept (I)nterpretation Framework

Dmitry Kazhdan, Botty Dimanov, Lucie Charlotte Magister +3

Explainable AI (XAI) underwent a recent surge in research on concept extraction, focusing on extracting human-interpretable concepts from Deep Neural Networks. An important challenge facing concept extraction approaches is the difficulty of interpreting and evaluating discovered concepts, especially for complex tasks such as molecular property prediction. We address this challenge by presenting GCI: a (G)raph (C)oncept (I)nterpretation framework, used for quantitatively measuring alignment between concepts discovered from Graph Neural Networks (GNNs) and their corresponding human interpretations. GCI encodes concept interpretations as functions, which can be used to quantitatively measure the alignment between a given interpretation and concept definition. We demonstrate four applications of GCI: (i) quantitatively evaluating concept extractors, (ii) measuring alignment between concept extractors and human interpretations, (iii) measuring the completeness of interpretations with respect to an end task and (iv) a practical application of GCI to molecular property prediction, in which we demonstrate how to use chemical functional groups to explain GNNs trained on molecular property prediction tasks, and implement interpretations with a 0.76 AUCROC completeness score.

Out-of-distribution (OoD) detection for data-based programs is a goal of paramount importance. Common approaches in the literature tend to train detectors requiring inside-of-distribution (in-distribution, or IoD) and OoD validation samples, and/or implement confidence metrics that are often abstract and therefore difficult to interpret. In this work, we propose to use existing work from the field of explainable AI, namely the PARTICUL pattern identification algorithm, in order to build more interpretable and robust OoD detectors for visual classifiers. Crucially, this approach does not require to retrain the classifier and is tuned directly to the IoD dataset, making it applicable to domains where OoD does not have a clear definition. Moreover, pattern identification allows us to provide images from the IoD dataset as reference points to better explain the confidence scores. We demonstrates that the detection capabilities of this approach are on par with existing methods through an extensive benchmark across four datasets and two definitions of OoD. In particular, we introduce a new benchmark based on perturbations of the IoD dataset which provides a known and quantifiable evaluation of the discrepancy between the IoD and OoD datasets that serves as a reference value for the comparison between various OoD detection methods. Our experiments show that the robustness of all metrics under test does not solely depend on the nature of the IoD dataset or the OoD definition, but also on the architecture of the classifier, which stresses the need for thorough experimentations for future work on OoD detection.

In recent years the applications of machine learning models have increased rapidly, due to the large amount of available data and technological progress.While some domains like web analysis can benefit from this with only minor restrictions, other fields like in medicine with patient data are strongerregulated. In particular \emph{data privacy} plays an important role as recently highlighted by the trustworthy AI initiative of the EU or general privacy regulations in legislation. Another major challenge is, that the required training \emph{data is} often \emph{distributed} in terms of features or samples and unavailable for classicalbatch learning approaches. In 2016 Google came up with a framework, called \emph{Federated Learning} to solve both of these problems. We provide a brief overview on existing Methods and Applications in the field of vertical and horizontal \emph{Federated Learning}, as well as \emph{Fderated Transfer Learning}.

We present Azimuth, an open-source and easy-to-use tool to perform error analysis for text classification. Compared to other stages of the ML development cycle, such as model training and hyper-parameter tuning, the process and tooling for the error analysis stage are less mature. However, this stage is critical for the development of reliable and trustworthy AI systems. To make error analysis more systematic, we propose an approach comprising dataset analysis and model quality assessment, which Azimuth facilitates. We aim to help AI practitioners discover and address areas where the model does not generalize by leveraging and integrating a range of ML techniques, such as saliency maps, similarity, uncertainty, and behavioral analyses, all in one tool. Our code and documentation are available at github.com/servicenow/azimuth.

Towards Human-Interpretable Prototypes for Visual Assessment of Image Classification Models

Poulami Sinhamahapatra, Lena Heidemann, Maureen Monnet +1

Explaining black-box Artificial Intelligence (AI) models is a cornerstone for trustworthy AI and a prerequisite for its use in safety critical applications such that AI models can reliably assist humans in critical decisions. However, instead of trying to explain our models post-hoc, we need models which are interpretable-by-design built on a reasoning process similar to humans that exploits meaningful high-level concepts such as shapes, texture or object parts. Learning such concepts is often hindered by its need for explicit specification and annotation up front. Instead, prototype-based learning approaches such as ProtoPNet claim to discover visually meaningful prototypes in an unsupervised way. In this work, we propose a set of properties that those prototypes have to fulfill to enable human analysis, e.g. as part of a reliable model assessment case, and analyse such existing methods in the light of these properties. Given a 'Guess who?' game, we find that these prototypes still have a long way ahead towards definite explanations. We quantitatively validate our findings by conducting a user study indicating that many of the learnt prototypes are not considered useful towards human understanding. We discuss about the missing links in the existing methods and present a potential real-world application motivating the need to progress towards truly human-interpretable prototypes.

Explainable artificial intelligence is an emerging and evolving concept. Its impact on construction, though yet to be realised, will be profound in the foreseeable future. Still, XAI has received limited attention in construction. As a result, no evaluation frameworks have been propagated to enable construction organisations to understand the what, why, how, and when of XAI. Our paper aims to fill this void by developing a content, context, process, and outcome evaluation framework that can be used to justify the adoption and effective management of XAI. After introducing and describing this novel framework, we discuss its implications for future research. While our novel framework is conceptual, it provides a frame of reference for construction organisations to make headway toward realising XAI business value and benefits.

Several approaches have been developed to mitigate algorithmic bias stemming from health data poverty, where minority groups are underrepresented in training datasets. Augmenting the minority class using resampling (such as SMOTE) is a widely used approach due to the simplicity of the algorithms. However, these algorithms decrease data variability and may introduce correlations between samples, giving rise to the use of generative approaches based on GAN. Generation of high-dimensional, time-series, authentic data that provides a wide distribution coverage of the real data, remains a challenging task for both resampling and GAN-based approaches. In this work we propose CA-GAN architecture that addresses some of the shortcomings of the current approaches, where we provide a detailed comparison with both SMOTE and WGAN-GP*, using a high-dimensional, time-series, real dataset of 3343 hypotensive Caucasian and Black patients. We show that our approach is better at both generating authentic data of the minority class and remaining within the original distribution of the real data.