Mission
Produce evidence about when AI-assisted teaching improves scientific learning and practice, when it introduces error, and how learners and educators can distinguish the two.
ABOUT SANTAROS LABS
A Nonprofit Lab for Learning and Scientific Practice
Santaros Labs is a nonprofit research lab studying how learners, educators, and scientific teams use AI to build research judgment, generate hypotheses, analyze data, review code, synthesize literature, and evaluate uncertainty.
Our Purpose and Direction
Vision
AI-assisted learning and research that expand scientific capability while preserving learner agency, accountability, reproducibility, multiple valid methods, and expert judgment.
A Nonprofit Mission With Visible Boundaries
Santaros Labs operates for scientific and public benefit rather than private distribution. Resources are directed to research staff, compute, data stewardship, independent review, replication, and dissemination.
We do not represent contributions as tax deductible or claim institutional accreditation, ethics determinations, registrations, or completed findings unless they are verified and published for the relevant entity or study.
RESEARCH INTEGRITY
Responsible From Question to Claim
Protocol before interpretation
For confirmatory work, the question, comparison, outcomes, exclusions, and analysis decisions are specified before outcomes are interpreted.
Documented oversight
Human-participant work does not begin until the responsible institution or authorized review body documents the required determination and safeguards.
Traceable evidence
We preserve sources, transformations, code, model details, and decision logs so the analytical path can be audited and computational results can be reproduced where feasible.
Research Practices and Public Commitments
Nonprofit research mission
Resources support scientific work, research infrastructure, and public-interest outputs rather than private distribution.
Honest study status
Concept, protocol drafting, review, registration, recruitment, analysis, and completed work are labeled separately. Plans are not findings.
Reproducible by design
Planned study artifacts include versioned protocols, analysis environments, and structured records that link sources, transformations, tools, and decisions.
Public-interest outputs
Where privacy, consent, licensing, and security permit, we publish methods, code, instruments, and negative results.
RESEARCH PARTNERSHIPS
Bring a Question Worth Testing
We welcome conversations with principal investigators, nonprofit institutes, universities, open-source communities, and funders working on trustworthy AI for science.
Methods, governance, and authorship expectations are discussed before a project begins.