Researchers Draw Critical Distinction Between AI Automation and True Agency

A new arXiv paper distinguishes between "agentic" systems (LLM tools with engineered workflows) and "agentive" systems (AI with internalized autonomy), arguing genuine agency requires self-directed goal-setting, identity evolution, and self-regulation rather than external scaffolding. The authors propose a Goal-Identity-Configurator architecture and examine safety and controllability implications as AI systems gain greater autonomy.

Why it matters: As companies deploy LLM-based agents, clarifying the difference between scripted automation and true autonomy is essential for both building safer systems and assessing real versus speculative AI risks.