Reproduce “Build Agent Advocates, Not Platform Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sayash Kapoor, Noam Kolt, Seth Lazar
Researchers from Princeton University, Mozilla, Hebrew University, and Australian National University analyze how Language Model Agents (LMAs), if controlled by platform companies, will intensify surveillance, lock-in, and market manipulation. The work proposes "agent advocates," user-controlled LMAs designed to operate exclusively in user interests, supported by open models, interoperability standards, and market regulation to enhance individual autonomy.
Key insights
- The fundamental choice of who controls AI agents—platforms or users—will determine whether advanced AI technology exacerbates or alleviates existing digital harms.
- "Platform agents," as the default trajectory, are predicted to intensify surveillance, manipulative market design, user lock-in, and illegitimate governance.
- "Agent advocates" can actively mitigate these harms by enabling user data control, fostering competitive markets, reducing switching costs, and empowering user autonomy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.