On September 22, we hosted another session of Constructor Talks featuring Dr. Denis Federyakin, Research Fellow at Johannes Gutenberg University Mainz and Goethe University Frankfurt.
Since ChatGPT’s release in late 2022, the landscape of information problem-solving has transformed dramatically. Yet harnessing the power of LLMs requires meticulous prompting—a skill that continues to evolve rapidly. Dr. Federyakin presented a groundbreaking insight: despite fundamental differences between how LLMs and humans process information, a finite set of psychological principles underlies all effective prompt engineering techniques.
Through an extensive literature review, Dr. Federyakin identified nine broad groups of psychological principles that simplify and systematize prompting approaches—from chain-of-thought reasoning and role-based prompting to leveraging self-reflection, managing cognitive load, and fostering creative divergent thinking.
The presentation bridged cognitive science and practical AI techniques, offering attendees a unified framework for understanding why certain prompting strategies work across diverse information problem-solving tasks.
A stimulating discussion followed, with participants gaining deeper insight into how psychology can inform more effective LLM interactions.