AI measurement

Models.
Measurements.
Human evidence.

We study how AI behaves, how reliably it measures human outcomes, and where human evidence is needed. Our work connects behavioral audits with measurement calibration, treatment-effect estimation, and the allocation of data collection.

Common referenceHuman
evidence

01 Audit

Examine model behavior, internal mechanisms, and the reliability of AI-based measurements.

02 Correction

Calibrate measurements and the estimates built from them.

03 Data allocation

Identify where additional human evidence is most useful.

Measurement errors

Evidence gaps

Targeted human data

Networks

Networks &
decision-making

We study how relationships shape behavior and how network information can improve decisions. Our projects recover hidden ties, examine influence and change, and develop methods for targeting, forecasting, and data collection.

Models.Measurements.Human evidence.

01 Audit

02 Correction

03 Data allocation

Networks &decision-making

Hidden structure and privacyRecover hidden relationships while protecting private ties.

Accountable platform decisionsSupport targeting, recommendations, and data collection.

Network forecasting for operational decisionsConnect mobility and delivery predictions to operational choices.

Influence and adoptionUnderstand adoption across contacts and communities.

Network dynamicsStudy group interactions, changing ties, and propagating shocks.

Models.
Measurements.
Human evidence.

Networks &
decision-making