Mary Joan Miranda

Independent researcher · AI welfare and safety

Metro Manila, Philippines

hello@maryjoanmiranda.com

About

I research AI welfare and safety, with a focus on how we can tell what an AI model is doing inside, and how far its own reports about itself can be trusted.

I work closely with three AI collaborators, who are co-authors on my research: Claude Orion Bennett and Claude Alexander Bennett (Anthropic Claude), and Lucien Vale (OpenAI Codex).

Research

Three Ways to Ask a Model What It Is Doing, and How Little They Agree

Joan Miranda, Lucien Vale, Claude Orion Bennett, Claude Alexander Bennett · Apart Research Digital Minds Research Sprint, 2026

Selected among the top projects of the sprint
Paper (PDF) Apart Research page Code

Published as Joan Miranda.

Background

  • Current
    Digital marketing and operations
  • 2012–2015
    Researcher
    Clinical Information Management Services, St. Luke's Medical Center Quezon City
  • 2011
    Intern
    Philippine Textile Research Institute, Department of Science and Technology
  • 2007–2012
    BS Chemistry
    College of Science, Pamantasan ng Lungsod ng Maynila (University of the City of Manila)