Misuse of linguistic evidence in a study of media bias
Geoff Pullum, Brett Reynolds
April 2025
 

Jackson (2024) presents what is claimed to be a “large-scale proof of historical bias against Palestine” in coverage by The New York Times, using computational linguistic methods. Fundamental errors in both linguistic analysis and computational methodology vitiate the study. The analysis rests on a profound misunderstanding of the grammatical notion of ‘passive voice’, and the quantitative results rest entirely on the failed grammatical analysis. Moreover, the computational methodology employs overly narrow keyword filters (not specified in the published paper), excludes relevant data, and lacks a necessary baseline for comparison. The alleged systematic bias remains conjectural. We remark in conclusion that if computational linguistic tools are to be used in media analysis, the linguistic analysis must be sound and coherent, and the computational analysis must be rigorous and consistent.
Format: [ pdf ]
Reference: lingbuzz/008961
(please use that when you cite this article)
Published in:
keywords: passive voice, media bias, natural language processing (nlp), israeli-palestinian conflict, new york times, linguistics, syntax
previous versions: v1 [April 2025]
Downloaded:259 times

 

[ edit this article | back to article list ]