Emotional AI in Reflective Writing
Optimizing Reflective Writing Through Human-Centered Prompt Engineering: Leveraging Emotional Agents

Topic
This thesis explores how emotional experiences influence reflective writing and investigates how AI-based emotional agents can support individuals during the reflection process. By combining theories of reflective learning and emotional design, the study aims to inform the development of emotionally intelligent AI tools that enhance cognitive and emotional engagement in digital writing contexts.
Relevance
As reflective writing becomes increasingly integrated into corporate learning, education, and self-development platforms, understanding how emotional support can improve this process is vital. Practitioners benefit from insights into how human-centered AI design can make reflection more accessible, empathetic, and productive for users in high-stress or emotionally complex environments.
Results
Participants reported that emotional clarity, timing, and writing conditions significantly affected their reflective depth. Most found emotional agents potentially useful—especially for emotion naming, cognitive prompts, and non-intrusive support. However, ethical concerns about emotional data usage emerged. The findings highlight design tensions between helpfulness and emotional safety in AI-based reflective tools.
Implications for Practitioners
- Design AI tools that offer emotionally intelligent, optional writing prompts.
- Integrate emotional awareness without overwhelming or distracting users.
- Allow user control over emotional tracking and data transparency.
- Apply insights in HR, coaching, e-learning, and digital health platforms.
- Train practitioners on integrating such tools ethically and sensitively.
Methods
The study employed a qualitative design using semi-structured interviews with individuals from diverse backgrounds experienced in reflective writing. The interview guide was developed based on Gibbs’ Reflective Cycle, emotional design theory (Norman, 2004), and affective computing principles (Picard, 1997). Data was analyzed thematically to identify emotional barriers, enablers, and user preferences for AI-supported reflection. The approach ensured depth, context sensitivity, and exploratory value for the development of emotionally adaptive digital systems.