Responsible Prompting Recommendation in Multi-Turn Interaction with LLMs

Large Language Models (LLMs) are being proposed as a solution to be applied in multiple workflows, but not all users are expert in prompt engineering (prompting), leading to a trial and error commonly seen while people interact with such computing systems. Beyond that, LLMs lack proper user guidance and Responsible AI awareness in prompting-time, i.e., before sending a given prompt to an LLM. In this context, this research proposes a way to provide user guidance and Responsible AI awareness while people interact with LLMs in a multi-turn fashion. We expect this work motivates more prompting recommender systems aiming at speeding up prompting tasks, users' agency, transparency, and promoting Responsible AI in prompting-time.