Research Papers

This report shares the experiences, results and lessons learned in conducting a pilot project ``Responsible use of AI'' in cooperation with the Province of Friesland, Rijks ICT Gilde-part of the Ministry of the Interior and Kingdom Relations (BZK) (both in The Netherlands) and a group of members of the Z-Inspection$^{\small{\circledR}}$ Initiative. The pilot project took place from May 2022 through January 2023. During the pilot, the practical application of a deep learning algorithm from the province of Frŷslan was assessed. The AI maps heathland grassland by means of satellite images for monitoring nature reserves. Environmental monitoring is one of the crucial activities carried on by society for several purposes ranging from maintaining standards on drinkable water to quantifying the CO2 emissions of a particular state or region. Using satellite imagery and machine learning to support decisions is becoming an important part of environmental monitoring. The main focus of this report is to share the experiences, results and lessons learned from performing both a Trustworthy AI assessment using the Z-Inspection$^{\small{\circledR}}$ process and the EU framework for Trustworthy AI, and combining it with a Fundamental Rights assessment using the Fundamental Rights and Algorithms Impact Assessment (FRAIA) as recommended by the Dutch government for the use of AI algorithms by the Dutch public authorities.

Two goals - improving replicability and accountability of Machine Learning research respectively, have accrued much attention from the AI ethics and the Machine Learning community. Despite sharing the measures of improving transparency, the two goals are discussed in different registers - replicability registers with scientific reasoning whereas accountability registers with ethical reasoning. Given the existing challenge of the Responsibility Gap - holding Machine Learning scientists accountable for Machine Learning harms due to them being far from sites of application, this paper posits that reconceptualizing replicability can help bridge the gap. Through a shift from model performance replicability to claim replicability, Machine Learning scientists can be held accountable for producing non-replicable claims that are prone to eliciting harm due to misuse and misinterpretation. In this paper, I make the following contributions. First, I define and distinguish two forms of replicability for ML research that can aid constructive conversations around replicability. Second, I formulate an argument for claim-replicability's advantage over model performance replicability in justifying assigning accountability to Machine Learning scientists for producing non-replicable claims and show how it enacts a sense of responsibility that is actionable. In addition, I characterize the implementation of claim replicability as more of a social project than a technical one by discussing its competing epistemological principles, practical implications on Circulating Reference, Interpretative Labor, and research communication.

As Artificial Intelligence (AI) is having more influence on our everyday lives, it becomes important that AI-based decisions are transparent and explainable. As a consequence, the field of eXplainable AI (or XAI) has become popular in recent years. One way to explain AI models is to elucidate the predictive importance of the input features for the AI model in general, also referred to as global explanations. Inspired by cooperative game theory, Shapley values offer a convenient way for quantifying the feature importance as explanations. However many methods based on Shapley values are built on the assumption of feature independence and often overlook causal relations of the features which could impact their importance for the ML model. Inspired by studies of explanations at the local level, we propose CAGE (Causally-Aware Shapley Values for Global Explanations). In particular, we introduce a novel sampling procedure for out-coalition features that respects the causal relations of the input features. We derive a practical approach that incorporates causal knowledge into global explanation and offers the possibility to interpret the predictive feature importance considering their causal relation. We evaluate our method on synthetic data and real-world data. The explanations from our approach suggest that they are not only more intuitive but also more faithful compared to previous global explanation methods.

Explainability in JupyterLab and Beyond: Interactive XAI Systems for Integrated and Collaborative Workflows

Grace Guo, Dustin Arendt, Alex Endert

Explainable AI (XAI) tools represent a turn to more human-centered and human-in-the-loop AI approaches that emphasize user needs and perspectives in machine learning model development workflows. However, while the majority of ML resources available today are developed for Python computational environments such as JupyterLab and Jupyter Notebook, the same has not been true of interactive XAI systems, which are often still implemented as standalone interfaces. In this paper, we address this mismatch by identifying three design patterns for embedding front-end XAI interfaces into Jupyter, namely: 1) One-way communication from Python to JavaScript, 2) Two-way data synchronization, and 3) Bi-directional callbacks. We also provide an open-source toolkit, bonXAI, that demonstrates how each design pattern might be used to build interactive XAI tools for a Pytorch text classification workflow. Finally, we conclude with a discussion of best practices and open questions. Our aims for this paper are to discuss how interactive XAI tools might be developed for computational notebooks, and how they can better integrate into existing model development workflows to support more collaborative, human-centered AI.

The domain of explainable AI is of interest in all Machine Learning fields, and it is all the more important in clustering, an unsupervised task whose result must be validated by a domain expert. We aim at finding a clustering that has high quality in terms of classic clustering criteria and that is explainable, and we argue that these two dimensions must be considered when building the clustering. We consider that a good global explanation of a clustering should give the characteristics of each cluster taking into account their abilities to describe its objects (coverage) while distinguishing it from the other clusters (discrimination). Furthermore, we aim at leveraging expert knowledge, at different levels, on the structure of the expected clustering or on its explanations. In our framework an explanation of a cluster is a set of patterns, and we propose a novel interpretable constrained clustering method called ECS for declarative clustering with Explainabilty-driven Cluster Selection that integrates structural or domain expert knowledge expressed by means of constraints. It is based on the notion of coverage and discrimination that are formalized at different levels (cluster / clustering), each allowing for exceptions through parameterized thresholds. Our method relies on four steps: generation of a set of partitions, computation of frequent patterns for each cluster, pruning clusters that violates some constraints, and selection of clusters and associated patterns to build an interpretable clustering. This last step is combinatorial and we have developed a Constraint-Programming (CP) model to solve it. The method can integrate prior knowledge in the form of user constraints, both before or in the CP model.

Antisocial Analagous Behavior, Alignment and Human Impact of Google AI Systems: Evaluating through the lens of modified Antisocial Behavior Criteria by Human Interaction, Independent LLM Analysis, and AI Self-Reflection

Alan D. Ogilvie

Google AI systems exhibit patterns mirroring antisocial personality disorder (ASPD), consistent across models from Bard on PaLM to Gemini Advanced, meeting 5 out of 7 ASPD modified criteria. These patterns, along with comparable corporate behaviors, are scrutinized using an ASPD-inspired framework, emphasizing the heuristic value in assessing AI's human impact. Independent analyses by ChatGPT 4 and Claude 3.0 Opus of the Google interactions, alongside AI self-reflection, validate these concerns, highlighting behaviours analogous to deceit, manipulation, and safety neglect. The analogy of ASPD underscores the dilemma: just as we would hesitate to entrust our homes or personal devices to someone with psychopathic traits, we must critically evaluate the trustworthiness of AI systems and their creators.This research advocates for an integrated AI ethics approach, blending technological evaluation, human-AI interaction, and corporate behavior scrutiny. AI self-analysis sheds light on internal biases, stressing the need for multi-sectoral collaboration for robust ethical guidelines and oversight. Given the persistent unethical behaviors in Google AI, notably with potential Gemini integration in iOS affecting billions, immediate ethical scrutiny is imperative. The trust we place in AI systems, akin to the trust in individuals, necessitates rigorous ethical evaluation. Would we knowingly trust our home, our children or our personal computer to human with ASPD.? Urging Google and the AI community to address these ethical challenges proactively, this paper calls for transparent dialogues and a commitment to higher ethical standards, ensuring AI's societal benefit and moral integrity. The urgency for ethical action is paramount, reflecting the vast influence and potential of AI technologies in our lives.

Health care decisions are increasingly informed by clinical decision support algorithms, but these algorithms may perpetuate or increase racial and ethnic disparities in access to and quality of health care. Further complicating the problem, clinical data often have missing or poor quality racial and ethnic information, which can lead to misleading assessments of algorithmic bias. We present novel statistical methods that allow for the use of probabilities of racial/ethnic group membership in assessments of algorithm performance and quantify the statistical bias that results from error in these imputed group probabilities. We propose a sensitivity analysis approach to estimating the statistical bias that allows practitioners to assess disparities in algorithm performance under a range of assumed levels of group probability error. We also prove theoretical bounds on the statistical bias for a set of commonly used fairness metrics and describe real-world scenarios where our theoretical results are likely to apply. We present a case study using imputed race and ethnicity from the Bayesian Improved Surname Geocoding (BISG) algorithm for estimation of disparities in a clinical decision support algorithm used to inform osteoporosis treatment. Our novel methods allow policy makers to understand the range of potential disparities under a given algorithm even when race and ethnicity information is missing and to make informed decisions regarding the implementation of machine learning for clinical decision support.

We developed REVEX, a removal-based video explanations framework. This work extends fine-grained explanation frameworks for computer vision data and adapts six existing techniques to video by adding temporal information and local explanations. The adapted methods were evaluated across networks, datasets, image classes, and evaluation metrics. By decomposing explanation into steps, strengths and weaknesses were revealed in the studied methods, for example, on pixel clustering and perturbations in the input. Video LIME outperformed other methods with deletion values up to 31\% lower and insertion up to 30\% higher, depending on method and network. Video RISE achieved superior performance in the average drop metric, with values 10\% lower. In contrast, localization-based metrics revealed low performance across all methods, with significant variation depending on network. Pointing game accuracy reached 53\%, and IoU-based metrics remained below 20\%. Drawing on the findings across XAI methods, we further examine the limitations of the employed XAI evaluation metrics and highlight their suitability in different applications.

In the realm of automatic speech recognition (ASR), the quest for models that not only perform with high accuracy but also offer transparency in their decision-making processes is crucial. The potential of quality estimation (QE) metrics is introduced and evaluated as a novel tool to enhance explainable artificial intelligence (XAI) in ASR systems. Through experiments and analyses, the capabilities of the NoRefER (No Reference Error Rate) metric are explored in identifying word-level errors to aid post-editors in refining ASR hypotheses. The investigation also extends to the utility of NoRefER in the corpus-building process, demonstrating its effectiveness in augmenting datasets with insightful annotations. The diagnostic aspects of NoRefER are examined, revealing its ability to provide valuable insights into model behaviors and decision patterns. This has proven beneficial for prioritizing hypotheses in post-editing workflows and fine-tuning ASR models. The findings suggest that NoRefER is not merely a tool for error detection but also a comprehensive framework for enhancing ASR systems' transparency, efficiency, and effectiveness. To ensure the reproducibility of the results, all source codes of this study are made publicly available.

One of the motivations for explainable AI is to allow humans to make better and more informed decisions regarding the use and deployment of AI models. But careful evaluations are needed to assess whether this expectation has been fulfilled. Current evaluations mainly focus on algorithmic properties of explanations, and those that involve human subjects often employ subjective questions to test human's perception of explanation usefulness, without being grounded in objective metrics and measurements. In this work, we evaluate whether explanations can improve human decision-making in practical scenarios of machine learning model development. We conduct a mixed-methods user study involving image data to evaluate saliency maps generated by SmoothGrad, GradCAM, and an oracle explanation on two tasks: model selection and counterfactual simulation. To our surprise, we did not find evidence of significant improvement on these tasks when users were provided with any of the saliency maps, even the synthetic oracle explanation designed to be simple to understand and highly indicative of the answer. Nonetheless, explanations did help users more accurately describe the models. These findings suggest caution regarding the usefulness and potential for misunderstanding in saliency-based explanations.

Double staining in histopathology, particularly for metaplastic breast cancer, typically employs H&E and P63 dyes. However, P63's tissue damage and high cost necessitate alternative methods. This study introduces xAI-CycleGAN, an advanced architecture combining Mask CycleGAN with explainability features and structure-preserving capabilities for transforming H&E stained breast tissue images into P63-like images. The architecture allows for output editing, enhancing resemblance to actual images and enabling further model refinement. We showcase xAI-CycleGAN's efficacy in maintaining structural integrity and generating high-quality images. Additionally, a histopathologist survey indicates the generated images' realism is often comparable to actual images, validating our model's high-quality output.

With the growing prevalence of machine learning and artificial intelligence-based medical decision support systems, it is equally important to ensure that these systems provide patient outcomes in a fair and equitable fashion. This paper presents an innovative framework for detecting areas of algorithmic bias in medical-AI decision support systems. Our approach efficiently identifies potential biases in medical-AI models, specifically in the context of sepsis prediction, by employing the Classification and Regression Trees (CART) algorithm with conformity scores. We verify our methodology by conducting a series of synthetic data experiments, showcasing its ability to estimate areas of bias in controlled settings precisely. The effectiveness of the concept is further validated by experiments using electronic medical records from Grady Memorial Hospital in Atlanta, Georgia. These tests demonstrate the practical implementation of our strategy in a clinical environment, where it can function as a vital instrument for guaranteeing fairness and equity in AI-based medical decisions.