Abstract
Reflective journaling has long been used as a method to enhance self-insight, emotional resilience, and mental well-being. However, traditional guided journaling tools that rely on standardized prompts often lead to reduced engagement and motivation in daily use. This study presents MindFeed, an AI-assisted reflective journaling tool that generates personalized journaling prompts based on users’ own social media content. The system combines three core models, including BLIP-2 for image captioning, RoBERTa for emotion classification, and GPT-4 for prompt generation, to transform user-generated Instagram or Snapchat stories into personally relevant journaling prompts to facilitate self-reflection. A user study (N = 20) compared MindFeed with a traditional guided journaling tool. The results showed that the AI-generated prompts matched traditional ones in engagement but varied in their ability to elicit deep personal insight. The qualitative results highlighted that prompt clarity, openness, alignment, and perceived authenticity were key factors influencing the user experience.

This publication has 44 references indexed in Scilit: