Abstract
This study investigates the effects of occlusions on the fairness of face recognition systems, particularly focusing on demographic biases. Using the Racial Faces in the Wild (RFW) dataset and synthetically added realistic occlusions, we evaluate their effect on the performance of face recognition models trained on the BUPT-Balanced and BUPT-GlobalFace datasets. We note increases in the dispersion of FMR, FNMR, and accuracy alongside decreases in fairness according to Equalized Odds, Demographic Parity, STD of Accuracy, and Fairness Discrepancy Rate. Additionally, we utilize a pixel attribution method to understand the importance of occlusions in model predictions, proposing a new metric, Face Occlusion Impact Ratio (FOIR), that quantifies the extent to which occlusions affect model performance across different demographic groups. Our results indicate that occlusions exacerbate existing demographic biases, with models placing higher importance on occlusions in an unequal fashion across demographics.

Abstract
Face recognition applications have grown in parallel with the size of datasets, complexity of deep learning models and computational power. However, while deep learning models evolve to become more capable and computational power keeps increasing, the datasets available are being retracted and removed from public access. Privacy and ethical concerns are relevant topics within these domains. Through generative artificial intelligence, researchers have put efforts into the development of completely synthetic datasets that can be used to train face recognition systems. Nonetheless, the recent advances have not been sufficient to achieve performance comparable to the state-of-the-art models trained on real data. To study the drift between the performance of models trained on real and synthetic datasets, we leverage a massive attribute classifier (MAC) to create annotations for four datasets: two real and two synthetic. From these annotations, we conduct studies on the distribution of each attribute within all four datasets. Additionally, we further inspect the differences between real and synthetic datasets on the attribute set. When comparing through the Kullback-Leibler divergence we have found differences between real and synthetic samples. Interestingly enough, we have verified that while real samples suffice to explain the synthetic distribution, the opposite could not be further from being true.

Abstract
Machine Learning has been overtaken by a growing necessity to explain and understand decisions made by trained models as regulation and consumer awareness have increased. Alongside understanding the inner workings of a model comes the task of verifying how adequately we can model a problem with the learned functions. Traditional global assessment functions lack the granularity required to understand local differences in performance in different regions of the feature space, where the model can have problems adapting. Residual Analysis adds a layer of model understanding by interpreting prediction residuals in an exploratory manner. However, this task can be unfeasible for high-dimensionality datasets through hypotheses and visualizations alone. In this work, we use weak interpretable learners to identify regions of high prediction error in the feature space. We achieve this by examining the absolute residuals of predictions made by trained regressors. This methodology retains the interpretability of the identified regions. It allows practitioners to have tools to formulate hypotheses surrounding model failure on particular regions for future model tunning, data collection, or data augmentation on critical cohorts of data. We present a way of including information on different levels of model uncertainty in the feature space through the use of locally fitted Model Agnostic Prediction Intervals (MAPIE) in the identified regions, comparing this approach with other common forms of conformal predictions which do not take into account findings from weak segment identification, by assessing local and global coverage of the prediction intervals. To demonstrate the practical application of our approach, we present a real-world industry use case in the context of inbound retention call-centre operations for a Telecom Provider to determine optimal pairing between a customer and an available assistant through the prediction of contracted revenue.

Abstract
This study investigates the use of SHAP (SHapley Additive exPlanations) values as an explainable artificial intelligence (xAI) technique applied on a facial attribute classification task. We analyse the consistency of SHAP value distributions across diverse classifier architectures that share the same feature extractor, revealing that key features driving attribute classification remain stable regardless of classifier architecture. Our findings highlight the challenges in interpreting SHAP values at the individual sample level, as their reliability depends on the model’s ability to learn distinct class-specific features; models exploiting inter-class correlations yield less representative SHAP explanations. Furthermore, pixel-level SHAP analysis reveals that superior classification accuracy does not necessarily equate to meaningful semantic understanding; notably, despite FaceNet exhibiting lower performance than CLIP, it demonstrated a more nuanced grasp of the underlying class attributes. Finally, we address the computational scalability of SHAP, demonstrating that KernelExplainer becomes infeasible for high-dimensional pixel data, whereas DeepExplainer and GradientExplainer offer more practical alternatives with trade-offs. Our results suggest that SHAP is most effective for small to medium feature sets or tabular data, providing interpretable and computationally manageable explanations.