Ana Lucic

Ana Lucic, Sheeraz Ahmad, Amanda Furtado Brinhosa et al. · amazon-science

When using medical images for diagnosis, either by clinicians or artificial intelligence (AI) systems, it is important that the images are of high quality. When an image is of low quality, the medical exam that produced the image often needs to be redone. In telemedicine, a common problem is that the quality issue is only flagged once the patient has left the clinic, meaning they must return in order to have the exam redone. This can be especially difficult for people living in remote regions, who make up a substantial portion of the patients at Portal Telemedicina, a digital healthcare organization based in Brazil. In this paper, we report on ongoing work regarding (i) the development of an AI system for flagging and explaining low-quality medical images in real-time, (ii) an interview study to understand the explanation needs of stakeholders using the AI system at OurCompany, and, (iii) a longitudinal user study design to examine the effect of including explanations on the workflow of the technicians in our clinics. To the best of our knowledge, this would be the first longitudinal study on evaluating the effects of XAI methods on end-users -- stakeholders that use AI systems but do not have AI-specific expertise. We welcome feedback and suggestions on our experimental setup.

Garvita Allabadi, Ana Lucic, Siddarth Aananth et al.

Traditional semi-supervised object detection methods assume a fixed set of object classes (in-distribution or ID classes) during training and deployment, which limits performance in real-world scenarios where unseen classes (out-of-distribution or OOD classes) may appear. In such cases, OOD data is often misclassified as ID, thus harming the ID classes accuracy. Open-set methods address this limitation by filtering OOD data to improve ID performance, thereby limiting the learning process to ID classes. We extend this to a more natural open-world setting, where the OOD classes are not only detected but also incorporated into the learning process. Specifically, we explore two key questions: 1) how to accurately detect OOD samples, and, most importantly, 2) how to effectively learn from the OOD samples in a semi-supervised object detection pipeline without compromising ID accuracy. To address this, we introduce an ensemble-based OOD Explorer for detection and classification, and an adaptable semi-supervised object detection framework that integrates both ID and OOD data. Through extensive evaluation on different open-world scenarios, we demonstrate that our method performs competitively against state-of-the-art OOD detection algorithms and also significantly boosts the semi-supervised learning performance for both ID and OOD classes.

Michael Neely, Stefan F. Schouten, Maurits Bleeker et al.

There has been significant debate in the NLP community about whether or not attention weights can be used as an explanation - a mechanism for interpreting how important each input token is for a particular prediction. The validity of "attention as explanation" has so far been evaluated by computing the rank correlation between attention-based explanations and existing feature attribution explanations using LSTM-based models. In our work, we (i) compare the rank correlation between five more recent feature attribution methods and two attention-based methods, on two types of NLP tasks, and (ii) extend this analysis to also include transformer-based models. We find that attention-based explanations do not correlate strongly with any recent feature attribution methods, regardless of the model or task. Furthermore, we find that none of the tested explanations correlate strongly with one another for the transformer-based model, leading us to question the underlying assumption that we should measure the validity of attention-based explanations based on how well they correlate with existing feature attribution explanation methods. After conducting experiments on five datasets using two different models, we argue that the community should stop using rank correlation as an evaluation metric for attention-based explanations. We suggest that researchers and practitioners should instead test various explanation methods and employ a human-in-the-loop process to determine if the explanations align with human intuition for the particular use case at hand.

Maksim Zhdanov, Ana Lucic, Max Welling et al.

We introduce Mosaic, a probabilistic weather forecasting model that addresses two principal sources of spectral degradation in ML-based weather prediction: (1) deterministic training against ensemble means and (2) compressive encoding creating an information bottleneck. Mosaic generates ensemble members through learned functional perturbations and operates on native-resolution grids via block-sparse attention, a hardware-aligned mechanism that captures long-range dependencies at linear cost by sharing keys and values across spatially adjacent queries. At 1.5$°$ resolution with 214M parameters, Mosaic matches or outperforms models trained on 6 times finer data on headline upper-air variables and achieves state-of-the-art results among 1.5$°$ models, producing well-calibrated ensembles whose individual members exhibit near-perfect spectral alignment across all resolved frequencies. A 24-member, 10-day forecast takes under 12 seconds on a single H100 GPU.

Angela van Sprang, Laurens Samson, Ana Lucic et al.

We introduce two new benchmarks REST and REST+(Render-Equivalence Stress Tests) to enable systematic evaluation of cross-modal inconsistency in multimodal large language models (MLLMs). MLLMs are trained to represent vision and language in the same embedding space, yet they cannot perform the same tasks in both modalities. Our benchmarks contain samples with the same semantic information in three modalities (image, text, mixed) and we show that state-of-the-art MLLMs cannot consistently reason over these different modalities. We evaluate 15 MLLMs and find that the degree of modality inconsistency varies substantially, even when accounting for problems with text recognition (OCR). Neither rendering text as image nor rendering an image as text solves the inconsistency. Even if OCR is correct, we find that visual characteristics (text colour and resolution, but not font) and the number of vision tokens have an impact on model performance. Finally, we find that our consistency score correlates with the modality gap between text and images, highlighting a mechanistic interpretation of cross-modal inconsistent MLLMs.

Ana Lucic

Model explainability has become an important problem in machine learning (ML) due to the increased effect that algorithmic predictions have on humans. Explanations can help users understand not only why ML models make certain predictions, but also how these predictions can be changed. In this thesis, we examine the explainability of ML models from three vantage points: algorithms, users, and pedagogy, and contribute several novel solutions to the explainability problem.

Ana Lucic, Maurits Bleeker, Sami Jullien et al.

In this work, we explain the setup for a technical, graduate-level course on Fairness, Accountability, Confidentiality, and Transparency in Artificial Intelligence (FACT-AI) at the University of Amsterdam, which teaches FACT-AI concepts through the lens of reproducibility. The focal point of the course is a group project based on reproducing existing FACT-AI algorithms from top AI conferences and writing a corresponding report. In the first iteration of the course, we created an open source repository with the code implementations from the group projects. In the second iteration, we encouraged students to submit their group projects to the Machine Learning Reproducibility Challenge, resulting in 9 reports from our course being accepted for publication in the ReScience journal. We reflect on our experience teaching the course over two years, where one year coincided with a global pandemic, and propose guidelines for teaching FACT-AI through reproducibility in graduate-level AI study programs. We hope this can be a useful resource for instructors who want to set up similar courses in the future.

Maksim Zhdanov, David Ruhe, Maurice Weiler et al.

We present Clifford-Steerable Convolutional Neural Networks (CS-CNNs), a novel class of $\mathrm{E}(p, q)$-equivariant CNNs. CS-CNNs process multivector fields on pseudo-Euclidean spaces $\mathbb{R}^{p,q}$. They cover, for instance, $\mathrm{E}(3)$-equivariance on $\mathbb{R}^3$ and Poincaré-equivariance on Minkowski spacetime $\mathbb{R}^{1,3}$. Our approach is based on an implicit parametrization of $\mathrm{O}(p,q)$-steerable kernels via Clifford group equivariant neural networks. We significantly and consistently outperform baseline methods on fluid dynamics as well as relativistic electrodynamics forecasting tasks.

Ege Erdogan, Ana Lucic

Sparse autoencoders (SAEs) have proven useful in disentangling the opaque activations of neural networks, primarily large language models, into sets of interpretable features. However, adapting them to domains beyond language, such as scientific data with group symmetries, introduces challenges that can hinder their effectiveness. We show that incorporating such group symmetries into the SAEs yields features more useful in downstream tasks. More specifically, we train autoencoders on synthetic images and find that a single matrix can explain how their activations transform as the images are rotated. Building on this, we develop adaptively equivariant SAEs that can adapt to the base model's level of equivariance. These adaptive SAEs discover features that lead to superior probing performance compared to regular SAEs, demonstrating the value of incorporating symmetries in mechanistic interpretability tools.

Cristian Bodnar, Wessel P. Bruinsma, Ana Lucic et al.

Reliable forecasts of the Earth system are crucial for human progress and safety from natural disasters. Artificial intelligence offers substantial potential to improve prediction accuracy and computational efficiency in this field, however this remains underexplored in many domains. Here we introduce Aurora, a large-scale foundation model for the Earth system trained on over a million hours of diverse data. Aurora outperforms operational forecasts for air quality, ocean waves, tropical cyclone tracks, and high-resolution weather forecasting at orders of magnitude smaller computational expense than dedicated existing systems. With the ability to fine-tune Aurora to diverse application domains at only modest computational cost, Aurora represents significant progress in making actionable Earth system predictions accessible to anyone.

Michael Neely, Stefan F. Schouten, Maurits J. R. Bleeker et al.

By computing the rank correlation between attention weights and feature-additive explanation methods, previous analyses either invalidate or support the role of attention-based explanations as a faithful and plausible measure of salience. To investigate whether this approach is appropriate, we compare LIME, Integrated Gradients, DeepLIFT, Grad-SHAP, Deep-SHAP, and attention-based explanations, applied to two neural architectures trained on single- and pair-sequence language tasks. In most cases, we find that none of our chosen methods agree. Based on our empirical observations and theoretical objections, we conclude that rank correlation does not measure the quality of feature-additive methods. Practitioners should instead use the numerous and rigorous diagnostic methods proposed by the community.

Kim de Bie, Ana Lucic, Hinda Haned

In hybrid human-AI systems, users need to decide whether or not to trust an algorithmic prediction while the true error in the prediction is unknown. To accommodate such settings, we introduce RETRO-VIZ, a method for (i) estimating and (ii) explaining trustworthiness of regression predictions. It consists of RETRO, a quantitative estimate of the trustworthiness of a prediction, and VIZ, a visual explanation that helps users identify the reasons for the (lack of) trustworthiness of a prediction. We find that RETRO-scores negatively correlate with prediction error across 117 experimental settings, indicating that RETRO provides a useful measure to distinguish trustworthy predictions from untrustworthy ones. In a user study with 41 participants, we find that VIZ-explanations help users identify whether a prediction is trustworthy or not: on average, 95.1% of participants correctly select the more trustworthy prediction, given a pair of predictions. In addition, an average of 75.6% of participants can accurately describe why a prediction seems to be (not) trustworthy. Finally, we find that the vast majority of users subjectively experience RETRO-VIZ as a useful tool to assess the trustworthiness of algorithmic predictions.

Ana Lucic, Madhulika Srikumar, Umang Bhatt et al.

Given that there are a variety of stakeholders involved in, and affected by, decisions from machine learning (ML) models, it is important to consider that different stakeholders have different transparency needs. Previous work found that the majority of deployed transparency mechanisms primarily serve technical stakeholders. In our work, we want to investigate how well transparency mechanisms might work in practice for a more diverse set of stakeholders by conducting a large-scale, mixed-methods user study across a range of organizations, within a particular industry such as health care, criminal justice, or content moderation. In this paper, we outline the setup for our study.

Ana Lucic, Maartje ter Hoeve, Gabriele Tolomei et al.

Given the increasing promise of graph neural networks (GNNs) in real-world applications, several methods have been developed for explaining their predictions. Existing methods for interpreting predictions from GNNs have primarily focused on generating subgraphs that are especially relevant for a particular prediction. However, such methods are not counterfactual (CF) in nature: given a prediction, we want to understand how the prediction can be changed in order to achieve an alternative outcome. In this work, we propose a method for generating CF explanations for GNNs: the minimal perturbation to the input (graph) data such that the prediction changes. Using only edge deletions, we find that our method, CF-GNNExplainer, can generate CF explanations for the majority of instances across three widely used datasets for GNN explanations, while removing less than 3 edges on average, with at least 94\% accuracy. This indicates that CF-GNNExplainer primarily removes edges that are crucial for the original predictions, resulting in minimal CF explanations.

Ana Lucic, Harrie Oosterhuis, Hinda Haned et al.

Model interpretability has become an important problem in machine learning (ML) due to the increased effect that algorithmic decisions have on humans. Counterfactual explanations can help users understand not only why ML models make certain decisions, but also how these decisions can be changed. We frame the problem of finding counterfactual explanations as a gradient-based optimization task and extend previous work that could only be applied to differentiable models. In order to accommodate non-differentiable models such as tree ensembles, we use probabilistic model approximations in the optimization framework. We introduce an approximation technique that is effective for finding counterfactual explanations for predictions of the original model and show that our counterfactual examples are significantly closer to the original instances than those produced by other methods specifically designed for tree ensembles.

Ana Lucic, Hinda Haned, Maarten de Rijke

In various business settings, there is an interest in using more complex machine learning techniques for sales forecasting. It is difficult to convince analysts, along with their superiors, to adopt these techniques since the models are considered to be "black boxes," even if they perform better than current models in use. We examine the impact of contrastive explanations about large errors on users' attitudes towards a "black-box'" model. We propose an algorithm, Monte Carlo Bounds for Reasonable Predictions. Given a large error, MC-BRP determines (1) feature values that would result in a reasonable prediction, and (2) general trends between each feature and the target, both based on Monte Carlo simulations. We evaluate on a real dataset with real users by conducting a user study with 75 participants to determine if explanations generated by MC-BRP help users understand why a prediction results in a large error, and if this promotes trust in an automatically-learned model. Our study shows that users are able to answer objective questions about the model's predictions with overall 81.1% accuracy when provided with these contrastive explanations. We show that users who saw MC-BRP explanations understand why the model makes large errors in predictions significantly more than users in the control group. We also conduct an in-depth analysis on the difference in attitudes between Practitioners and Researchers, and confirm that our results hold when conditioning on the users' background.

Ana Lucic, Hinda Haned, Maarten de Rijke

Understanding how "black-box" models arrive at their predictions has sparked significant interest from both within and outside the AI community. Our work focuses on doing this by generating local explanations about individual predictions for tree-based ensembles, specifically Gradient Boosting Decision Trees (GBDTs). Given a correctly predicted instance in the training set, we wish to generate a counterfactual explanation for this instance, that is, the minimal perturbation of this instance such that the prediction flips to the opposite class. Most existing methods for counterfactual explanations are (1) model-agnostic, so they do not take into account the structure of the original model, and/or (2) involve building a surrogate model on top of the original model, which is not guaranteed to represent the original model accurately. There exists a method specifically for random forests; we wish to extend this method for GBDTs. This involves accounting for (1) the sequential dependency between trees and (2) training on the negative gradients instead of the original labels.