Agentic AI: User Empowerment or Enclosure?

Research from arXiv highlights that the impact of Agentic AI, specifically whether it empowers or constrains users, is not solely determined by the technology itself. A comparative analysis of historical cases such as ad blockers, recommender systems, robo-advisors, and email spam governance demonstrates that the interests served by these agentic systems are shaped by technical arrangements like API choices, protocol governance, industry standards, and default configurations. These technical decisions are fundamentally political in nature, indicating that the development and deployment of Agentic AI will involve similar political considerations regarding whose interests are prioritized.

Why it matters

The paper underscores that the design and deployment of Agentic AI represent a critical juncture for governance and policy. Stakeholders must proactively engage in shaping the technical and political frameworks to ensure that these advanced systems align with desired outcomes, rather than inadvertently leading to user disempowerment or market enclosure.

Key insights

• Agentic AI's impact on user empowerment versus enclosure depends on factors beyond the technology itself.
• Decisions regarding whose interests agentic systems serve are resolved through technical arrangements (APIs, protocols, standards, configurations).
• Technical arrangements in agentic systems are inherently political decisions.
• Historical examples of agentic systems (ad blockers, recommender systems, robo-advisors, email spam governance) provide precedent for understanding Agentic AI's trajectory.
• The flexibility of digital agency offered by Agentic AI, from content filtering to price negotiation, necessitates careful consideration of its design parameters.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.06510