Date of Award

Summer 2018

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Psychology

Committee Director

Richard N. Landers

Committee Member

Konstantin P. Cigularov

Committee Member

Ryan L. Klinger

Abstract

Natural language processing techniques can be used to analyze text and speech data. These techniques have been applied within many domains to date but have only recently been examined in the domain of personnel assessment. By linking workplace-relevant constructs such as general cognitive ability (GCA) to natural language processing outcomes such as word counts, a foundation for language-based psychological assessment of those abilities can be laid. Over 400 participants were recruited through Amazon Mechanical Turk to write cognitively demanding essays and complete a battery of cognitive tests. Essays were analyzed using Linguistic Inquiry and Word Count (LIWC). Structural equation modeling was used to examine the relationship between GCA and word count categories as well as the relationship between broad cognitive abilities and word count categories. Latent GCA added incremental prediction of unique word use over latent verbal ability and incremental prediction of preposition use over latent short-term memory. Although not statistically significant, latent GCA and latent verbal ability related to various LIWC word count categories the strongest out of the abilities measured, yielding small to medium effect sizes in both positive and negative directions. Latent short-term memory and latent fluid reasoning were weakly related or unrelated to the LIWC word count categories observed. Word counting approaches to natural language processing may partially express GCA and latent verbal ability, but not latent short-term memory and latent fluid reasoning in cognitively demanding essay contexts.

DOI

10.25777/jw6c-ec97

ISBN

9780438455627

Share

COinS