Emotional Prompting: Definition and Examples

Emotional Prompting is a prompt engineering approach that leverages emotional language to positively influence the performance of a large language model (LLM). Rather than formulating purely factual and neutral instructions, this technique adds emotionally charged expressions—such as urgency, personal importance, or stakes—to encourage the model to produce more careful and thorough responses.

This technique is based on a surprising discovery: LLMs, although they do not feel emotions, have been trained on vast corpora of human texts where emotional language is systematically associated with contexts demanding more rigor and attention. Consequently, formulations like 'This is very important for my career' or 'Make sure to give the best possible result' activate statistical patterns that push the model towards higher quality responses.

Research conducted notably by Microsoft and Chinese universities in 2023 demonstrated that adding emotional stimuli to prompts could improve LLM performance by 8% to 115% depending on the task, both in text generation and logical reasoning. These results were observed across several models, including GPT-4, Claude, and LLaMA.

Emotional Prompting should not be confused with manipulating the model. Rather, it is a pragmatic lever that takes advantage of how models have learned to associate certain language registers with certain expected quality levels. It is a complementary tool to other prompting techniques, easy to integrate into any instruction.