Daniel Muise

Head Scientist, Outset  ·  Stanford PhD  ·  former founder

Computational social science, applied to simulations.

I'm a computational social scientist. I lead the Science organization at Outset, working primarily within the Simulations Lab.

Focus areas

Measurement of human behavior, applied to industry research and development.

Digital twins. At Outset I lead the science org and scientific practice within the Simulations Lab. We design LLM-based twins of real people, measure how accurate they are, and rapidly ship improvements to their fidelity (against the humans they represent).

Latent assumptions in media measurement. My PhD work used continuous smartphone screen recording to investigate the latent assumptions underlying large cross-sectional work in political polarization.

The human experience of time. I'm interested in the role of time in lived experience, in information consumption, and in the decay of preferences and memory. Each bears on how public opinion gets measured, and motivates being selective about which data types answer which questions in social science.

Selected publications

See all →

Major projects

Efforts I've led or co-led.

Outset Simulations Lab

The central focus of the science org at Outset: digital twin design, accuracy measurement, and fidelity research.

2026 – present

Screenlake

Founded and ran a venture-backed startup that monitored engagement with brands and political figures across a cumulative panel of 100,000 smartphones worldwide, with privacy compliance via distributed computing. IP sold in early 2026.

2022 – 2026 · the story

Princeton-SMART

An open-source smartphone screen-recording and analytics tool, built with Justin Cornelius, now hosted by the Princeton Research Accelerator.

2025 · at the Research Accelerator

Digitox

Maintained and grew Digitox, a top-10 Play Store app for screen-time self-management: personal analytics and self-control tools to limit smartphone use. Now run by Generation Lab.

2019 – 2022

CDF.PSIdekick

An R package on CRAN from the Harvard Privacy Tools project, with Kobbi Nissim and Georgios Kellaris. A test bench for differentially private algorithms: generate summary statistics and CDFs under a privacy budget, then measure how much accuracy each algorithm gives up for the same guarantee.

2016 · CRAN

Recent talks

See all →

  • Talk2026-11

    Digital twin composition

  • Talk2026-10

    Digital twins are as much a data problem as a model problem

  • Webinar2026
    Online

    Outset webinar on digital twins (second of two)

  • Webinar2026
    Online

    Outset webinar on digital twins (first of two)

Other

Daniel Muise outdoors among trees

Music and performance. I enjoy musical performance and critical listening. I play several instruments and have been on stage as a lead in musicals, back in high school. My toddler is now my main audience.

Mandalay FM. Pre-COVID, I spent over a year as a radio host at one of the largest radio stations in Myanmar. I was recognized on the street once by a taxi driver, and I miss the role. It taught me how media works beyond measurement.

Languages. I'm a long-time learner of Mandarin Chinese, currently learning Urdu, and fascinated by linguistics.

Redwoods. I support the preservation of redwood trees, and I'm growing one in my garden.