1. médialab Sciences Po
  2. News
  3. Measurement validation and machine learning in the social sciences

Measurement validation and machine learning in the social sciences

Alex Kindel will be presenting at the next FacSem which will take place on Thursday, January 9th

Event, Research Seminar

Salle du Conseil, 13 rue de l'Université, 75007 Paris

Sciences Po's Vice Presidence for Research invites you to the FacSem with Alex Kindel, Assistant Professor at the médialab.

He will present his research on the topic "Measurement validation and machine learning in the social sciences" with Lou Safra.

Abstract

"Many research problems in the social sciences require measuring hypothetical, abstract, theoretical, conceptual, implicit, or otherwise latent features of the social world. In the 20th century, social scientists developed methods for constructing and validating latent measures primarily using data describing individuals’ responses to tests and surveys (factor analysis, item response theory). Over the past two or three decades, advances in computer science (greater data availability, increased computing power, more complex models) have expanded the empirical scope of this measurement approach, and it is increasingly common for social scientists to incorporate measures of latent variation in a corpus of media (text, images, audio, video) into empirical social research. This talk will discuss some of the validation challenges in this domain through an extended example: the use of word association models to measure conceptual associations."

Alex KINDEL studies social measurement. Most of his work is concerned with the history and practice of methodology (i.e. rules and rulemaking about knowledge) in the domain of attaching numbers to what people do, say, and think.

Lou SAFRA, Assistant Professor at the Centre for Political Research at Sciences Po (CEVIPOF), will be the discutant.

This FacSem will be held in Salle du Conseil, 13 rue de l'Université, 5th floor.

Mandatory registration: marina.abelskaiagraziani@sciencespo.fr