Prompting the Future: AI-Guided Episodic Thinking to Support Long-Term Goals

Abstract

People often struggle with long-term goals because short-term rewards feel more immediate, making motivation and self-regulation difficult. Episodic Future Thinking (EFT) addresses this by prompting individuals to vividly imagine positive future events linked to their goals. However, current EFT interventions require time-consuming cue creation and offer limited delivery formats, reducing scalability and everyday relevance. To address these limitations, we developed a mobile application that uses large language models (LLMs) to automatically generate personalized EFT scenarios. The app lowers user effort while maintaining psychological impact through multimodal delivery (text, audio, and image). In a six-day feasibility study with twelve participants, we measured changes in delay discounting, delayed gratification, motivation, and user experience. Results showed a significant increase in intrinsic motivation, with qualitative feedback highlighting good usability and flexible modality preferences. Together, these findings demonstrate short-term feasibility and provide early design considerations for longer-term evaluations of AI-assisted EFT. Figure 1: