{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_53106"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_53106","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"Towards Trustworthy Graph Modeling : Addressing Fairness and Transparency in Social Network Analysis","abstract":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","abstract_html":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","abstract_has_math":false,"creators":["Santos, Francisco"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Tan, Pang-Ning","Esfahanian, Abdol-Hossein","Frank, Kenneth","Ross, Arun"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T18:59:37Z","subjects":["Computer science"],"languages":["eng"],"rights":["In Copyright"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:53106","local:Santos_grad.msu_0128D_22117"],"render_values":[{"text":"etd:53106","href":null,"code":true},{"text":"local:Santos_grad.msu_0128D_22117","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/q3zw-sn95","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/q3zw-sn95","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_53106","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Tan, Pang-Ning","Esfahanian, Abdol-Hossein","Frank, Kenneth","Ross, Arun"]},{"key":"dc:creator","label":"Author","values":["Santos, Francisco"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["In Copyright"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:53106","local:Santos_grad.msu_0128D_22117","https://doi.org/doi:10.25335/q3zw-sn95"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","Machine learning is increasingly being used for crucial decision-making tasks across diverse domains, from recidivism prediction and credit approval to college admission and employee recruitment. As these algorithms influence high-stakes outcomes that can profoundly affect individuals’ lives, there is a growing demand to ensure that AI systems are not only accurate but also ethical, transparent, and fair. This imperative has led to the emergence of the trustworthy AI paradigm, which promotes the development of systems grounded in four foundational principles: fairness, transparency, prevention of harm, and human control. These principles aim to ensure that AI systems behave reliably, treat individuals equitably, and provide understandable reasoning behind their decisions.This thesis targets two core principles of trustworthy AI—fairness and transparency—in the context of graph-based machine learning. Graphs are a natural way to represent relational data and are widely used in tasks such as node classification, ranking, and influence modeling. While Graph Neural Networks (GNNs) have advanced the state of the art in these areas by integrating node features with graph topology in a nonlinear way through a message-passing scheme that enables information sharing among neighboring nodes, they also introduce critical challenges. Specifically, GNNs can amplify disparities by propagating existing biases embedded in graph topology or node attributes, which may result in unequal outcomes across demographic groups. Moreover, GNNs are often harder to interpret, making it challenging to understand the internal logic behind their predictions. This lack of interpretability can lead to the loss of valuable insights about how decisions are made and how information flows within the network.To overcome these challenges, this thesis proposes novel machine learning algorithms that explicitly promote fairness and interpretability in GNNs for node classification, ranking, representation learning, and influence propagation problems. First, we introduce FACS-GCN (Fairness-Aware Cost-Sensitive Graph Convolutional Network), a node classification method that combines cost-sensitive learning with adversarial training to jointly address structural bias and class imbalance in graph data. Second, we propose DeepFairRank, a multi-objective node ranking algorithm that integrates both group and individual fairness criteria while incorporating acceptability scores to generate context-aware, equitable rankings of nodes. Third, we develop FairGRUNT (Fair Graph Representation Learning for Unseen Sensitive Attribute Groups), a framework that handles incomplete demographic data using disentangled representations and hard-label estimation to extend fairness to unseen subpopulations. Finally, we present TRIMIS (Temporal Recurrence for Modeling Influence and Micro-level Individual Susceptibility), an interpretable influence propagation model that infers personalized susceptibility and influence scores from observed cascades, enabling transparent reasoning about how information and behaviors spread in networks. Together, these contributions advance the broader goal of trustworthy AI by developing fair and interpretable graph-based models that are both effective and socially responsible. By tackling key limitations in bias mitigation and model transparency, this thesis lays the foundation for more accountable and trustworthy machine learning systems in networked environments.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (121 p.) : ill."]},{"key":"dc:title","label":"Title","values":["Towards Trustworthy Graph Modeling : Addressing Fairness and Transparency in Social Network Analysis"]}]}],"canonical_facts":{"dc:contributor":["Tan, Pang-Ning","Esfahanian, Abdol-Hossein","Frank, Kenneth","Ross, Arun"],"dc:creator":["Santos, Francisco"],"dc:date":["2026"],"dc:description":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","Machine learning is increasingly being used for crucial decision-making tasks across diverse domains, from recidivism prediction and credit approval to college admission and employee recruitment. As these algorithms influence high-stakes outcomes that can profoundly affect individuals’ lives, there is a growing demand to ensure that AI systems are not only accurate but also ethical, transparent, and fair. This imperative has led to the emergence of the trustworthy AI paradigm, which promotes the development of systems grounded in four foundational principles: fairness, transparency, prevention of harm, and human control. These principles aim to ensure that AI systems behave reliably, treat individuals equitably, and provide understandable reasoning behind their decisions.This thesis targets two core principles of trustworthy AI—fairness and transparency—in the context of graph-based machine learning. Graphs are a natural way to represent relational data and are widely used in tasks such as node classification, ranking, and influence modeling. While Graph Neural Networks (GNNs) have advanced the state of the art in these areas by integrating node features with graph topology in a nonlinear way through a message-passing scheme that enables information sharing among neighboring nodes, they also introduce critical challenges. Specifically, GNNs can amplify disparities by propagating existing biases embedded in graph topology or node attributes, which may result in unequal outcomes across demographic groups. Moreover, GNNs are often harder to interpret, making it challenging to understand the internal logic behind their predictions. This lack of interpretability can lead to the loss of valuable insights about how decisions are made and how information flows within the network.To overcome these challenges, this thesis proposes novel machine learning algorithms that explicitly promote fairness and interpretability in GNNs for node classification, ranking, representation learning, and influence propagation problems. First, we introduce FACS-GCN (Fairness-Aware Cost-Sensitive Graph Convolutional Network), a node classification method that combines cost-sensitive learning with adversarial training to jointly address structural bias and class imbalance in graph data. Second, we propose DeepFairRank, a multi-objective node ranking algorithm that integrates both group and individual fairness criteria while incorporating acceptability scores to generate context-aware, equitable rankings of nodes. Third, we develop FairGRUNT (Fair Graph Representation Learning for Unseen Sensitive Attribute Groups), a framework that handles incomplete demographic data using disentangled representations and hard-label estimation to extend fairness to unseen subpopulations. Finally, we present TRIMIS (Temporal Recurrence for Modeling Influence and Micro-level Individual Susceptibility), an interpretable influence propagation model that infers personalized susceptibility and influence scores from observed cascades, enabling transparent reasoning about how information and behaviors spread in networks. Together, these contributions advance the broader goal of trustworthy AI by developing fair and interpretable graph-based models that are both effective and socially responsible. By tackling key limitations in bias mitigation and model transparency, this thesis lays the foundation for more accountable and trustworthy machine learning systems in networked environments.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (121 p.) : ill."],"dc:identifier":["etd:53106","local:Santos_grad.msu_0128D_22117","https://doi.org/doi:10.25335/q3zw-sn95"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["In Copyright"],"dc:subject":["Computer science"],"dc:title":["Towards Trustworthy Graph Modeling : Addressing Fairness and Transparency in Social Network Analysis"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:37Z"}