[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84976-en":3,"doc-seo-84976-105":29,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84976,7971461740909,"Levi","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Unraveling Machine Behavior by Multi-Level Bias Analysis and Detection","Bias in machine learning, especially in computer vision applications like biometrics, raises ethical, legal, and practical concerns. This study examines how bias appears and propagates inside neural networks using a multi-level lens: learned latent representations, layer activations, and model parameters. A three-part detection taxonomy is proposed: SpaceBias, ActivationBias, and WeightBias, enabling interpretation of where disparity is encoded within network architecture. Experiments cover gender classification on DiveFace and digit classification on colored-MNIST with controlled bias severity, evaluating over 127,000 trained models.","arXiv :2607 .07236v 1 [ cs .CV] 8 Jul 2026  \nUnraveling Machine Behavior by Multi-Level Bias Analysis and Detection:  \nMethodology and Application to Computer Vision  \nIgnacio Sernaa,∗, Aythami Moralesb,c , Julian Fierrezb  \na Center for Humans and Machines, Max Planck Institute for Human Development, Berlin, Germany b BiometricsAI, UAM, Madrid, Spain  \nc Department of Mathematics, ULPGC, Spain  \nAbstract  \nBias in machine learning, particularly in computer vision applications such as biometrics, is a critical issue with profound ethical, legal, and practical implications. This study investigates the presence and propagation of bias within Neural Networks through a comprehensive multi-level analysis spanning the learned latent space, layer activations, and the network’s parameters. Based on this taxonomy, we propose three bias detection approaches: 1) SpaceBias (new method), which characterizes the latent space prior to the final classification layer using neighbor-probability distributions and quantifies bias with the two-sample Kolmogorov–Smirnov test on the per-group distributions. 2) ActivationBias (extension of the existing method InsideBias), which analyzes the activations of neural network filtersand quantifies bias via a Mann–Whitney U test, based on the observed fact that underrepresented groups exhibit lower activation levels in the final convolutional layers. 3) WeightBias (extension of the existing method IFBiD), which uses a secondary neural network trained to identify biased patterns directly in the parameters of task-specific models. Unlike conventional methods, which assess neural network outcomes and treat the model as a black box, our proposed techniques provide insight into how biases manifest within the network architecture itself at different levels, offering a more nuanced and detailed understanding. Experiments are conducted on two complementary applications: gender classification in the DiveFace dataset (72,000 face images) and digit classification on a colored-MNIST benchmark with controlled bias severity. In total, more than 127,000 models with varying degrees and types of bias were trained and evaluated. The severity sweep shows that the internal disparity, and with it the detection performance, decreases smoothly as the training distribution approaches balance. The results highlight the importance of methods that provide deeper insight into the behavior of AI models.  \nKeywords: Bias Detection, Face Biometrics, Neural Networks, Convolution, Neuron Activation, Latent Space  \n1. Introduction  \nArtificial Intelligence (AI), particularly through the development of neural networks and machine learning algorithms, has catalyzed transformative changes across diverse domains. From optimizing logistics to predicting market trends, AI’s ability to process massive volumes of data and find overlooked patterns has led to more accurate predictions, quicker responses, and informed decision-making. However, the integration of AI into critical decision-making processes also brings significant ethical and practical challenges to the forefront, especially with respect to fairness, transparency, and accountability.  \nThe sphere of influence of AI algorithms spans a wide spectrum of domains, from the financial sector to the criminal justice system. In finance, AI-driven algorithms have revolutionized risk assessment, enabling lenders to make informed decisions about creditworthiness with unprecedented accuracy. Similarly, in law enforcement, the  \n∗ Corresponding author: [serna@mpib-berlin.mpg.de](serna@mpib-berlin.mpg.de)  \nability of AI to analyze patterns and anomalies has taken predictive policing and resource allocation to new levels of efficiency [1] .  \nHowever, as AI’s role in shaping decisions expands, a number of ethical complexities also unfold. A fundamental challenge lies in ensuring that the algorithms that guide decision-making are grounded in fairness, transparency, and accountability. The opacity of dee","cbCaicMRRTkAqPlz","https://ap.wps.com/l/cbCaicMRRTkAqPlz","pdf",1318885,1,29,"English","en",105,"# Introduction\n## Problem of bias and fairness challenges\n## Motivation and research objective\n## Related work and gap\n# Proposed multi-level bias detection framework\n## SpaceBias\n## ActivationBias\n## WeightBias\n# Experimental evaluation and results\n## DiveFace gender classification\n## Colored-MNIST digit classification with controlled severity\n## Bias severity analysis and insights","[{\"question\":\"What multi-level aspects of a neural network does the study analyze for bias detection?\",\"answer\":\"The approach examines bias across the learned latent space, layer activations, and the network parameters, providing a view of how bias emerges throughout model internals rather than only at outputs.\"},{\"question\":\"What are SpaceBias, ActivationBias, and WeightBias?\",\"answer\":\"SpaceBias detects bias in the latent space before the final classification layer using neighbor-probability distributions and a two-sample Kolmogorov–Smirnov test. ActivationBias analyzes filter activations using a Mann–Whitney U test, based on lower activations for underrepresented groups. WeightBias trains a secondary network to identify biased patterns directly in the model parameters.\"},{\"question\":\"How were the proposed methods evaluated and what does the bias severity trend show?\",\"answer\":\"Experiments use DiveFace for gender classification and a colored-MNIST benchmark for digit classification under controlled bias severity. With increasing closeness of the training distribution to balance, internal disparity decreases smoothly, and detection performance declines accordingly.\"}]",1784199922,73,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"unraveling-machine-behavior-by-multi-level-bias-analysis-and-detection","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/unraveling-machine-behavior-by-multi-level-bias-analysis-and-detection/84976/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What multi-level aspects of a neural network does the study analyze for bias detection?","Question",{"text":74,"@type":75},"The approach examines bias across the learned latent space, layer activations, and the network parameters, providing a view of how bias emerges throughout model internals rather than only at outputs.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"What are SpaceBias, ActivationBias, and WeightBias?",{"text":79,"@type":75},"SpaceBias detects bias in the latent space before the final classification layer using neighbor-probability distributions and a two-sample Kolmogorov–Smirnov test. ActivationBias analyzes filter activations using a Mann–Whitney U test, based on lower activations for underrepresented groups. WeightBias trains a secondary network to identify biased patterns directly in the model parameters.",{"name":81,"@type":72,"acceptedAnswer":82},"How were the proposed methods evaluated and what does the bias severity trend show?",{"text":83,"@type":75},"Experiments use DiveFace for gender classification and a colored-MNIST benchmark for digit classification under controlled bias severity. With increasing closeness of the training distribution to balance, internal disparity decreases smoothly, and detection performance declines accordingly.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":105,"slug":137},19,"General","general"]