Social Noise Classification Model Integration with Explainable AI (XAI) Using SHAP and LIME

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science and Info Sys

Date of Award

Spring 5-1-2026

Abstract

The rapid use of social media platforms as sources of information and news makes it a fertile environment to spread misinformation. Provided studies attempt to enhance the classification of identifying misinformation. However, due to the nature of misinformation, what might be misinformation for someone may not be the same for someone else. So, the previous studies proposed that the social noise concept attempts to address the online users’ behavior of participation in sharing information and unintentionally spreading misinformation on an online platform, which in turn contributes to social noise. Their studies focused on defining the construct and identifying the keywords associated with each construct. There is a need to enhance the classification of social noise on social media platforms such as X. This study aims to enhance the classification of social noise on tweets related to Ukraine hashtags by integrating Deep Learning (DL) models such as LSTM, CNN, BERT and XLM-RoBERTa. Additionally, this research incorporates explainable AI (XAI) methods, namely Shapley Additive Explanations (SHAP) and Local Interpretable Model - Agnostic Explanation (LIME), to uncover insights into the decision-making process of the models, addressing their black-box characteristics and making it possible to understand v the rationale behind the model’s decisions. This process of analysis will provide an explanation of the performance of each model, provide transparency, and achieve high accuracy. Keywords: Social noise, Deep learning, XAI, SHAP, LIME

Advisor

Omar El Ariss

Subject Categories

Computer Sciences | Physical Sciences and Mathematics

Link to Full Text

Share

COinS