Smartphone Usage and Next App Prediction for Smart Interventions: A Quantum LSTM Approach

This work focuses on analyzing Smartphone Problematic Usage by predicting the next mobile app a user is likely to engage with, using advanced Artificial Intelligence (AI) algorithms. The project examines app usage patterns across various mobile applications and leverages Deep Learning models to forecast user interaction dynamics and engagement patterns for mobile apps, identifying apps that contribute to excessive use. By incorporating Quantum Cognition, which applies Quantum Probability Theory to human decision-making, this research explores a comparison between Quantum Deep Learning (qLSTM) and Traditional Deep Learning (LSTM). The qLSTM model, designed to align more closely with human cognitive processes, demonstrates superior accuracy and computational efficiency, effectively predicting app usage and flagging problematic apps. To mitigate excessive usage, the system triggers timely interventions, warning users when they engage with apps for extended periods. The study employs a real-world dataset of 90,717 app usage records and evaluates the performance of qLSTM and LSTM models using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The results demonstrate that qLSTM achieves an MAE of 3.170, while LSTM yields an MAE of 4.839, with the quantum approach consistently outperforming traditional methods. Additionally, this work discusses the ethical considerations of user intervention, weighing the benefits of reducing overuse against respecting user autonomy. The research highlights the potential of Quantum-Enhanced Machine Learning to address challenges in digital well-being and lays the groundwork for future advancements in quantum-based predictive analytics.