AI evaluation and psychometrics

Psychometrics · AI evaluation

I research, design and build. Then I check what works.

I evaluate AI models and design tools for research and everyday work.

Explore my work Get in touch

Selected work

Questions that become projects.

Research / AI evaluation

Silicon Invariance

Can a synthetic population stand in for a human survey?

A psychometric audit of six models, examining structure, comparability and prediction against a human reference.

Explore the case

Reversing the response labels

4 of 6

models kept a positive profile correlation (+0.50 to +0.94) after label reversal, so they did not follow the new labels.
Within the models and conditions studied. Article in preparation.

Research / Comparability

PISA and invariance

What makes comparisons between education systems defensible, and how an adaptive design changes the analysis.

Explore the case

Tool / Psychometrics

EvalIA

Six IRT models, independent numerical checks and an explicit path from data to reproducible R code.

Explore the case

System / Creative production

Cinefilosos Studio

A production tool for a podcast: organize the work, prepare pieces and trace their claims back to the episode.

Explore the case

All nine projects (Spanish) Articles and guides (Spanish)