Artificial Intelligence Based Prediction of Social Media Engagement among Young Adults: A Literature Review

Social media has become an important part of everyday communication, information sharing, entertainment, and social interaction among young adults. At the same time, artificial intelligence (AI) is increasingly used by social media platforms to recommend content, personalize feeds, rank posts, and influence what users see and interact with. AI and machine-learning methods are also being used by researchers to understand and predict user engagement. However, the available research is spread across different areas. Some studies focus on psychological and behavioral factors among young adults, while others concentrate on prediction models, temporal patterns, recommendation algorithms, or explainable AI. This integrative review brings these research streams together. Analysis was performed on twenty-four studies from 2017 to 2026. The studies have been compared based on their population, platforms, data, engagement indicators, behavioral indicators, predictive variables, methodology, machine learning models, results, and limitations. The literature was organized into six themes: the meaning and measurement of social media engagement; behavioral and psychological factors among young adults; AI and machine-learning methods for engagement prediction; temporal, contextual, content, and multimodal features; the influence of AI recommendation systems; and explainability, privacy, and responsible AI. The review shows that engagement is not a single behavior. Passive viewing, liking, commenting, sharing, and content creation have different characteristics and may require different prediction strategies. Studies also show that personality, boredom, information overload, emotion regulation, social influence, previous activity, content characteristics, posting time, and algorithmic recommendations can affect engagement. Machine learning algorithms such as Random Forest and Gradient Boosting are appropriate for structured data whereas LSTMs and other deep learning algorithms are appropriate for sequential data. Recently, there has also been some attention paid to the usage of Transformers and other graph-based algorithms. Regardless of these advancements, not many studies have been carried out which integrate specific behaviors of young adults with multidimensional engagement, temporal and contextual data, AI predictions, and explainability. Moreover, cross platform validation, privacy, fairness, and explainability of the models is yet to be taken into consideration. Thus, this review suggests a framework for the use of AI from young adult-centric and context-aware perspective.