Independent research

Small, noisy steps toward better evidence.

StocGrad studies how people and organizations behave — and builds the quantitative and AI-assisted tools that make that research faster, more careful, and easier to reproduce.

FocusBehavioral & organizational science
MethodsQuantitative · Computational
Fig. 1. Each estimate is noisy; the trend is not. Stochastic gradient descent in miniature — and a fair description of how research makes progress.
mini-batch estimaterunning averageoptimum
Research areas

Three lines of inquiry.

Our work sits where substantive questions about people meet the methods used to answer them.

01

Behavior at work

How individuals and teams think, decide, and adapt inside organizations — and what shapes those patterns over time.

02

Measurement & modeling

Latent-variable, longitudinal, and multilevel models — with attention to what a measure really captures.

03

AI for scholarship

Careful use of language models to read, organize, and check the research literature at scale — with humans in the loop.

Approach

Borrowed from optimization, applied to research.

θ₀

Start from the literature

Good questions begin with a close, structured reading of what is already known.

∇̂

Expect noise

Any single study is an estimate. We design for uncertainty rather than around it.

η

Take measured steps

Small, frequent, checkable iterations beat rare leaps of faith.

θ*

Converge in the open

Methods, decisions, and corrections should be reproducible by someone else.

Status

Work in progress, by design.

Selected papers, working notes, and tools will appear here as they are ready.

Currently iterating
Papers · forthcoming
Notes · forthcoming
Tools · forthcoming