# Prompting AI Art: An Investigation into the Creative Skill of Prompt Engineering

**Authors:** Jonas Oppenlaender, Rhema Linder, Johanna Silvennoinen  
**arXiv:** [2303.13534](https://arxiv.org/abs/2303.13534)  
**Date:** 2023-03-13  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.13534v1/

## Key points

- Prompt engineering is a novel creative skill for creating prompt-based art
- Participants were able to assess the quality of prompts and corresponding images, with their ability increasing based on experience and interest in art
- Participants demonstrated proficiency in writing prompts using rich descriptive language
- Their prompts lacked specific vocabulary needed to apply a certain style to generated images, suggesting that prompt engineering requires expertise and practice
- Ten recommendations are provided for conducting experimental research on text-to-image generation and prompt engineering
- The study speculates on four possible futures of prompt engineering
- Contributes to understanding prompt engineering as a creative skill with implications for AI art creation.

## Summary

In the paper titled "Prompting AI Art: An Investigation into the Creative Skill of Prompt Engineering," authors Jonas Oppenlaender, Rhema Linder and Johanna Silvennoinen explore the concept of prompt engineering as a novel creative skill for creating prompt-based art. They investigate whether untrained participants can recognize the quality of prompts, write prompts themselves and improve their prompts. To do this, three studies were conducted with participants recruited from a crowdsourcing platform. The results showed that participants were able to assess the quality of prompts and corresponding images; this ability increased based on their experience and interest in art. Additionally, participants demonstrated proficiency in writing prompts using rich descriptive language. However, it was observed that their prompts lacked specific vocabulary needed to apply a certain style to the generated images despite being instructed to generate artworks. This suggests that prompt engineering is a learned skill which requires expertise and practice. Based on their findings and experience with running studies using a paid crowd, the authors provide ten recommendations for conducting experimental research on text-to-image generation and prompt engineering which aim to enhance understanding of prompt engineering and open up avenues for future research in this field. The paper concludes by speculating on four possible futures of prompt engineering. In summary, this study contributes to our understanding of prompt engineering as a creative skill and highlights its potential implications for AI art creation.