Algorithmic Prompt Generation for Diverse Human-like Teaming and Communication with Large Language Models
arXiv:2504.03991v2 Announce Type: replace Abstract: Understanding how humans collaborate and communicate in teams is essential for improving human-agent teaming and AI-assisted decision-making. However, relying solely on data from large-scale user studies is impractical due to logistical, ethical, and practical constraints, necessitating synthetic models of multiple diverse human behaviors. Recently, agents powered by Large Language Models (LLMs) have been shown to emulate human-like behavior in social settings. But, obtaining a large set of diverse behaviors requires manual effort in the form
The increasing sophistication of LLMs allows for more effective emulation of human behavior, making synthetic team models a viable alternative to resource-intensive user studies.
This development addresses critical constraints in human-AI collaboration research, accelerating the development of more effective and diverse human-agent teams and AI-assisted decision-making systems.
The reliance on expensive and logistically challenging large-scale human user studies for AI team dynamics research can now be significantly reduced, replaced by scalable 'algorithmic prompt generation' for synthetic behaviors.
- · AI research and development
- · Companies building AI-assisted decision systems
- · Defense and aerospace (for human-agent teaming)
- · Academia focused on human-computer interaction
- · Traditional user study providers
- · Organizations heavily reliant on manual data collection for team dynamics
More robust and generalizable AI teaming models will emerge, improving human-AI collaboration across various sectors.
The ability to simulate diverse human behaviors at scale will accelerate AI development by enabling rapid iteration and testing of human-AI interfaces.
Ethical considerations around the 'reality' and 'autonomy' of synthetic AI agents might become more prominent as LLM-generated behaviors become indistinguishable from human input.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL