| Ph. Mylonas, A. Krouska, C. Troussas, C. Sgouropoulou |
| Explainable AI for Personalized Image and Video Content Analysis in Social Media |
| The 3rd International Conference on Foundation and Large Language Models (FLLM2025), 25-28 November 2025, Vienna, Austria |
|
ABSTRACT
|
| This paper presents a hybrid explainable AI framework for personalized image and video content analysis in social media, designed to balance recommendation accuracy with transparency. Our approach integrates Deep Learning modules for multimedia feature extraction, using a ResNet-50 backbone for images and an LSTM for temporal video analysis, with an interpretable Machine Learning component based on decision trees and rule-based reasoning. An NLP-driven explanation generator adapts the style and complexity of justifications to the user¢s technical profile, producing faithful and accessible rationales. We evaluate our framework on a combined dataset of Instagram and YouTube public data, complemented by a proprietary multimedia interaction corpus. Comparative experiments against 3 representative baselines. i.e., collaborative filtering with template explanations, the NARRE review-based explainer, and a deep learning–only recommender, show that our model achieves superior recommendation quality (F1-score = 0.91) and explanation clarity (mean user rating = 4.3/5). Objective faithfulness analysis using SHAP confirms that 87% of generated explanations accurately reflect the model¢s decision process. Results demonstrate that incorporating interpretable reasoning into multimodal recommendation not only improves user trust, but also enhances ranking performance, making this approach a promising direction for transparent and user-centered AI in social media environments.
|
| 25 November , 2025 |
| Ph. Mylonas, A. Krouska, C. Troussas, C. Sgouropoulou, "Explainable AI for Personalized Image and Video Content Analysis in Social Media", The 3rd International Conference on Foundation and Large Language Models (FLLM2025), 25-28 November 2025, Vienna, Austria |
[  PDF] [
BibTex] [
Print] [
 Back] |
