Story Arena: A Multi-Agent Environment for Envisioning the Future of Software Engineering

TL;DR

This paper investigates how AI can help envision the future of software engineering by using a multi-agent 'writer's room' to generate a design fiction about AI's impact on the profession. It introduces Story Arena, a Python-based BeeAI workflow that orchestrates ideation, consensus, and collaborative writing among diverse large language models, including an AI agent representing AI's own perspective. The short story and accompanying reflections probe trust, transparency, ownership, augmentation vs replacement, and human–AI collaboration in a high-scale software setting, highlighting methodological considerations for evaluating AI-authored prose. The work demonstrates a novel approach to using design fictions to explore socioethical implications of AI in software engineering and suggests directions for alignment, governance, and human-centric AI design.

Abstract

What better way to understand the impact of AI on software engineering than to ask AI itself? We constructed Story Arena, a multi-agent "writer's room" in which multiple AI agents, independently imbued with a position statement on the future of software engineering, converse with each other to develop a shared vision. They then use this shared vision to collaboratively construct a design fiction that depicts this vision in narrative form. We present "The Code of Trust," a short fiction that investigates themes of human comprehension, trust, content ownership, augmentation vs. replacement, and uncertain futures in human-AI co-creation.