The Good, the Bad, and the Uncanny: Investigating Diversity Aspects of LLM-Generated Personas for Requirements Engineering -- Summary

Key Points

• The aim is to explore the diversity aspects of personas generated by large language models (LLMs) in the context of requirements engineering.
• Qualitative analysis of LLM-generated personas
• Comparative evaluation of persona diversity
• Framework development for assessing persona effectiveness
• Identified strengths and weaknesses of LLM-generated personas
• Highlighted implications for user representation
• Demonstrated the need for improved diversity in generated personas