Research Papers

Counterfactual explanations are a prominent approach in explainable artificial intelligence (xAI), providing actionable guidance on what input changes would alter a model's prediction to a desired outcome. While early methods primarily focused on minimal feature changes, recent work incorporates additional properties such as sparsity, actionability and plausibility. Despite this progress, fair and systematic evaluation remains challenging. Existing studies often rely on different data splits, predictive models, and evaluation metrics, which limits objective comparison across methods. To fill this gap, we introduce CEL (Counterfactual Explanations Library), a unified library and benchmark for counterfactual explanations designed to support consistent implementation and evaluation. CEL includes 18 datasets of varying size and complexity and provides implementations or reimplementations of 14 widely used counterfactual methods. Using this standardized setup, we conduct a comprehensive quantitative comparison across a variety of methods on datasets that differ in size, number, and types of attributes. The evaluation protocol incorporates multiple complementary metrics capturing validity, coverage, sparsity, proximity, and distributional plausibility, including density- and outlier-based measures to assess the realism of generated counterfactuals. To the best of our knowledge, this is the first comprehensive benchmark that systematically evaluates recent counterfactual explanation methods within a unified and reproducible framework. While prior libraries and benchmarking efforts exist in the literature, many are outdated, limited in scope, or lack consistent evaluation protocols. The proposed benchmark aims to improve reproducibility, enable fair comparison, and establish a workbench for the development of future counterfactual explanation methods.

**Context:** Deep Neural Networks (DNNs) increasingly control Cyber-Physical Systems (CPSs), yet small input perturbations can cause unsafe system-level behavior. Existing approaches often optimize perturbations for individual images and evaluate them only in simulation, limiting their generalizability and practical validity. **Objectives:** This work aims to generate robustness tests that remain effective across operational observations and to evaluate whether the resulting failures transfer from simulation to a physical robot. **Methods:** We propose an explainability-guided multi-objective evolutionary approach that generates sparse perturbations over representative images selected through visual and behavioral clustering. Aggregated Integrated Gradients guide mutations toward influential image regions. We evaluate the approach on a DNN-controlled LeoRover in Gazebo, conduct an ablation study, and validate a stratified subset of perturbations on the physical robot. **Results:** The approach achieved a median success rate of 70.0%, compared with 53.85% for unguided search, and increased median hypervolume from 0.65 to 0.73. Multi-image optimization improved the success rate from 50.0% to 57.5%, while XAI guidance further increased it to 70.0%. In the sim-to-real evaluation, simulation achieved 0.95 precision and 0.67 recall, and simulated and physical failure times showed a significant positive correlation of 0.617. **Conclusion:** Combining multi-image optimization with explainability-guided search improves robustness testing for DNN-controlled robotic systems. Simulation effectively identifies and prioritizes transferable failures, but physical validation remains necessary because some real-world failures are not reproduced in simulation.

Automated detection of vision impairing retina-based ocular conditions from fundus images is important for early screening, timely referral and reducing dependency on specialist-only assessment, for which neural network-based deep learning (DL) models have been widely utilized. However, explainability of the DL frameworks remains a major bottleneck for clinical adoption, particularly when model decisions are not linked to retinal regions that are clinically meaningful. To address this issue, this study presents CounterFundus, a novel CycleGAN-driven counterfactual explainability framework, integrating EfficientNet-B5-based retinal disease detection with visually interpretable disease-to-normal fundus image translation. For each pathological image, the counterfactual yielded by the CycleGAN generator represents an estimated healthy counterpart and the resultant difference map is utilized to localize disease-associated retinal changes. Unlike conventional post-hoc saliency methods, CounterFundus provides counterfactual explanations through visually plausible disease-to-normal retinal translation. Thereafter, to quantify the spatial agreement between counterfactual difference maps and classifier saliency, the Counterfactual-Classifier Alignment Score (CCAS) is introduced, embedding Spearman correlation, binary IoU and pointing accuracy into a single assessment protocol. To this end, EigenCAM-aligned evaluation demonstrates that the generated counterfactual explanations remain spatially consistent with classifier-relevant retinal evidence across all CCAS dimensions. Along with that, ablation studies further confirm that CCAS-filtered counterfactual augmentation improves the downstream classification performance in fundus images, establishing CounterFundus as a clinically-grounded, explainable artificially intelligence (XAI) framework for retinal disease detection.

Artificial intelligence is being deployed across the Global South at a pace matching or exceeding the Global North, yet AI governance has not kept pace, and the gap is far wider in the South. Drawing on a decade of AI audit practice across Latin America, Sub-Saharan Africa, and Asia Pacific (the only fully published second-party audit of a deployed system in the region, Robot Laura in Brazil; two completed but unreleased national audits, of a child-welfare risk model and a public-employment matching algorithm; thirteen Responsible AI Assessments; and a regional landscape analysis), this paper documents patterns in the sparse field of Global South AI evaluation and explains why it so rarely occurs. We count fewer than twenty published second- and third-party audits of deployed systems across the region over the past decade, against hundreds of documented public-sector algorithms and multibillion-dollar national AI investments. We find four cross-cutting patterns: proxy targets that substitute predictability for validity, performance claims that collapse under prevalence analysis, populations scored by models that never saw them in training, and structural bias that persists even after protected attributes are removed. We then draw five lessons for evaluation, regulation, and funding that emerge outside the regulatory, linguistic, and data conditions of the North. We argue the gap is not, at root, a capacity problem but a funding problem: capacity follows funded demand, and no actor is currently required, or funded, to hold deployed systems to account. The actor best placed to close it is the small number of development and philanthropic funders behind most consequential AI in the region, whose funding conditions can require independent evaluation where no regulator yet does. We close with recommendations for funders, governments, and the design of evaluation itself.

Generative AI is changing a basic premise of educational assessment: that submitted work can reliably evidence the human capacities a credential claims to certify. The challenge is not simply whether students use AI, but what remains inferable about learning when some cognitive work has been delegated to a system. This paper develops cognitive stewardship, a framework for AI-mediated assessment that links the learning claim, delegation boundary, evidence standard, and safeguards. We then audit verified public generative AI assessment guidance from 30 universities. Using a pre-specified scoring codebook--a written, source-grounded rubric--four open-weight LLM models applied the rubric as structured coders, with scores averaged to reduce dependence on any single model's bias. The audit shows that public policies are becoming better at classifying AI use than at explaining what evidence and protections preserve credential validity. Boundaries are more visible than evidence standards; safeguards are uneven; and guidance is clearest when AI use resembles final-output substitution rather than feedback, access, verification, or professional workflow. The takeaway is that permission categories are necessary but insufficient. Universities need policies that make the certification logic visible: what learners may delegate, what they must still demonstrate, and how institutions will protect fair evidence rather than merely monitor AI use.

Validating Explainable Artificial Intelligence (XAI) methods in medical imaging requires ground-truth data with known locations of informative features. However, current approaches rely on expert annotations, which are prone to labeling errors, or on hand-crafted artificial perturbations superimposed onto healthy images to mimic lesions or malignant features, which lack clinical realism. We present Local Label-Informed Feature Transfer (LLIFT), a framework for generating semi-synthetic brain magnetic resonance images with realistic lesions placed in user-controlled regions, which does not require pixel-level lesion annotations during training. We implement LLIFT with two generative paradigms: LLIFT-GAN, a custom GAN that learns pathological features from binary class labels alone, and LLIFT-DM, a diffusion-based inpainting pipeline conditioned on bounding-box masks via ControlNet. Both approaches are evaluated on brain magnetic resonance imaging data derived from the Human Connectome Project. In evaluations, both achieve Fréchet Inception Distance scores, with respect to the real pathological distribution, that are comparable to the inter-class reference between healthy and pathological images in the given dataset. Furthermore, qualitative inspection confirms the realism of lesion structures. The resulting benchmark datasets provide spatially controlled ground truth data for evaluating XAI methods in medical imaging.

Clinical prediction models in intensive care degrade as data shift, and regulators such as the FDA ask developers to specify bounded update protocols in advance. Evaluations of such protocols assume that apparent drift reflects clinical practice rather than changes in how the database records care. We test whether update strategies can be evaluated on unaudited ICU data, and whether a parameter-bounded update gives up accuracy or attribution stability relative to full fine-tuning once the data are audited. We audit MIMIC-IV 3.1 (2008-2022) for recording artifacts and define a cleaning protocol. On the cleaned cohort, a label-free classifier two-sample test gates updates across three eras. We compare a bounded update, which retrains only treatment-context and fusion parameters, with full fine-tuning for two model families and two targets under a pre-specified non-inferiority margin of 0.015 AUROC, and measure attribution stability against the source model. In 2020-2022 about half of ICU stays have no input or procedure-event records, and a vasopressor ItemID migration makes the 2017-2019 label rate appear to fall when it actually rises. On the cleaned cohort, the gate detects drift in every era, including in physiology-only features, with no false alarms on null data. Bounded updating is non-inferior or superior in 10 of 12 primary cells, but this depends on the target, the model family and learning-rate tuning, and adaptation gains over never updating are small. At a shared learning rate, bounded updates keep physiological attributions more stable in neural models, clearly for vasopressor and less clearly for intubation; with separately tuned rates, full updates are more stable. Audits should precede update evaluation, and the update boundary should be chosen per target.

Concept erasure aims to remove semantic concepts from a trained generative model and is increasingly important for responsible AI deployment. However, verifying whether a model has robustly removed targeted concepts remains a critical challenge. Existing evaluation methods are typically pre-defined and static, failing to expose vulnerabilities under diverse natural-language probes and challenging conditions. Moreover, manually designed evaluation strategies can be biased and difficult to scale. We posit that concept erasure evaluation is best formulated as an adaptive hypothesis search, operationalised by agents that iteratively propose, critique, and verify tests to systematically expand coverage of failure modes. To this end, we propose Stress Testing Agents for Concept Erasure (STACE), a framework that autonomously stress-tests concept-erased models using multiple Large Language Model (LLM) agents, by iteratively generating and verifying stress-testing hypotheses grounded by external knowledge. We also introduce a suite of metrics for assessing the performance and efficiency of LLM-agent-powered stress-testing frameworks. Our extensive experiments show that STACE outperforms five LLM-based evaluation baselines on four concept categories. Further analysis across two T2I models, six concept erasure approaches, and various erasure strengths show that STACE is robust for different settings. We also show that STACE can be adapted beyond concept erasure evaluation to other problem domains, such as LLM jailbreaking. Our code is available anonymously.

Large language models (LLMs) are increasingly adopted for network automation, yet their output quality and inference cost can vary substantially across LLM families. We present HuGLEN, a stepwise evaluation pipeline that uses an LLM-as-a-judge together with a small set of expert ratings to enable scalable and reproducible comparison of candidate LLMs, and to rank them using a quality efficiency score (QES). We demonstrate HuGLEN for translating outputs from an explainable artificial intelligence (XAI) model for the optical network quality of transmission (QoT) estimation task into operator-friendly explanations. Our results show that a medium-sized LLM (12B parameters) achieves the highest QES, indicating the best trade-off between explanation quality and efficiency. Overall, HuGLEN reduces the human-labeling burden while supporting consistent model selection for operator-facing automation tasks.

The rapid improvement of AI systems has intensified debate about humanity's economic, political, social, and existential future. As AI reshapes expectations about what lies ahead, policy choices and institutional responses will play a crucial role in determining who benefits, who bears the costs, and whether the most serious risks can be mitigated. Africa remains relatively overlooked in these discussions, partly because it is largely a consumer rather than a producer of frontier AI systems, and also because many countries on the continent continue to face pressing development challenges. Given the catch-up imperative, governments across Africa are eager to embrace AI as a means for accelerating economic transformation. Drawing on speeches, press releases, public statements, and national AI strategies/frameworks, this essay argues that while African governments are highly attentive to AI's opportunities, they devote comparatively little attention to AI safety.

This study contributes toward development of an Automated Data Processing (ADP) framework designed to evaluate and reinforce optimal machine learning model-feature combinations for predictive tasks in fused deposition modeling (FDM) process datasets. The methodology is centered around a reinforcement learning-inspired policy updating mechanism, where multiple machine learning models are trained on both full feature sets and feature subsets selected through Shapley-based Explainable AI (SHAP XAI) across 217 datasets. At each episode, the framework assesses the predictive accuracy and F1-scores of each model-feature pair, computes a scalar reward, and updates $Q$ values to guide future model selection. SHAP XAI feature importance was employed to generate reduced yet informative feature subsets to enable the framework to explore performance with dimensionality. The policy was shown to evolve over multiple episodes, with reward distributions used to visualize performance stability. Overall, results indicate that leveraging the ADP framework through XAI algorithms successfully converges toward optimal model-feature configurations with improved accuracy and stability. Specifically, the proposed framework improves the test-set AUC from 0.9248 to 0.9731 and increases the mean reward value by more than fifty percent compared with the baseline full-feature configuration.

Can LLM Code Explanations Adapt to Diverse Problem-Solvers' Needs?

Andrew Anderson, David Piorkowski, Justin Weisz +2

Large language model (LLM) code explanations can support people in solving code-related problems, yet prior work has shown that people have diverse problem-solving styles. If explanations fail to meet people's problem-solving needs, they may be less productive in their occupations and miss opportunities to learn and grow. Although some research has examined how LLMs can adapt their outputs to a user's age or expertise, no prior work has examined how LLMs can adapt their code explanations to people's problem-solving styles. To address this gap, we developed prompts from an established inclusive design method that considers 5 types of problem-solving styles, and we generated 1,072 code explanations from six open-weight LLMs. Using natural language processing techniques, we uncovered a taxonomy of 13 linguistic adaptations, with each adaptation supported by evidence from the literature, the prompts, or the LLMs' outputs. They also show which LLMs adapted their code explanations more frequently than others. This paper is the first to investigate problem-solving style adaptations in LLM code explanation, contributing two problem-solving adaptation approaches: declarative statements for each adaptation and 10 problem-solving style prompts.