Research Intern - Inference Economics and Human Agency

Microsoft Microsoft · Big Tech · Redmond, WA +1 · Applied Sciences

Research intern to conduct empirical research on how human oversight shapes the economic return of AI-assisted work. This includes designing controlled experiments, developing session-level evaluation frameworks that link inference cost to output quality and human effort, and analyzing how interface design choices affect user confidence, reliance, and decision quality during AI-assisted tasks. The role involves collaboration with the MADE team and preparing a submission-ready research manuscript.

What you'd actually do

  1. Conduct empirical research on how human oversight shapes the economic return of AI-assisted work.
  2. Design controlled experiments.
  3. Develop session-level evaluation frameworks that link inference cost to output quality and human effort.
  4. Analyze how interface design choices affect user confidence, reliance, and decision quality during AI-assisted tasks.
  5. Collaborate with the MADE team in the Office of the CTO and prepare a submission-ready research manuscript.

Skills

Required

  • Currently enrolled in a PhD program in Computer Science, Cognitive Science, Human-Computer Interaction, Information Systems, Behavioral Economics, or a related STEM field
  • At least 1 year of research experience in one or more of the following: human-AI interaction, decision science, metacognition, AI/LLM evaluation, or computational economics

Nice to have

  • Experience designing and running human-subjects experiments (IRB-compliant protocols, participant recruitment, controlled task studies)
  • Familiarity with LLM agent architectures, prompt engineering, or multi-step AI orchestration systems
  • Background in confidence calibration, signal detection theory, or metacognitive measurement
  • Experience with mixed-methods research combining quantitative behavioral data and qualitative interview analysis
  • Proficiency in Python and statistical analysis tools (R, pandas, scipy, or equivalent)
  • Published or in-progress research in CHI, CSCW, FAccT, AAAI, NeurIPS, or related venues
  • Familiarity with inference cost modeling, token economics, or cloud compute pricing

What the JD emphasized

  • PhD program in Computer Science, Cognitive Science, Human-Computer Interaction, Information Systems, Behavioral Economics, or a related STEM field
  • 1 year of research experience in one or more of the following: human-AI interaction, decision science, metacognition, AI/LLM evaluation, or computational economics
  • human-subjects experiments
  • LLM agent architectures, prompt engineering, or multi-step AI orchestration systems
  • confidence calibration, signal detection theory, or metacognitive measurement
  • mixed-methods research combining quantitative behavioral data and qualitative interview analysis
  • Python and statistical analysis tools (R, pandas, scipy, or equivalent)
  • Published or in-progress research in CHI, CSCW, FAccT, AAAI, NeurIPS, or related venues
  • inference cost modeling, token economics, or cloud compute pricing

Other signals

  • human oversight shapes economic return of AI-assisted work
  • designing controlled experiments
  • developing session-level evaluation frameworks
  • analyzing interface design choices