EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

A Speech Act baseline for Evaluating AI-generated Headlines: Four Criteria from Pakistani Print Media

Maria Saleem, Misbah Afsheen Khan

Journal of Arts and Linguistics Studies · 26.10.2026 · 4 (10) · 597-621

DOI: 10.71281/jals.v4i10.863 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Arts and Linguistics Studies, online erschienen am 26.10.2026. Verfasst von Maria Saleem und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Newsrooms have started to leverage generative AI to write and propose headlines, but the metrics typically applied to these types of text are mostly related to accuracy, bias, and fluency, and don't offer much information about whether a headline has the right kind of act in the right tone. Judging that includes a description of the purpose of human edited headlines. This study brings one for the English-language press of Pakistan and is based on it a set of preliminary criteria for the review of AI-generated headlines. Hundred headlines were coded between January 2024 and August 2026 in Dawn, The News International, The Express Tribune and The Nation based on Speech Act Theory for each of the four topics, Searle categories, valence and three scales of reader effect. Of the newspaper headlines, 88% featured representatives and a majority of all newspapers featured representatives. The valence of the newspaper differed (Cramér's V = .28, p = .011) with Dawn having the most negative valence, an emotive intensity which differed with a small to medium effect (ε² = .09). These differences were reduced when security, crime and disaster news was removed (p = .51 and p = .33). High emotive intensity (V = .48): none of the 30 Negative headlines were Least Emotive and none of the 20 Neutral headlines were Highly Emotive. The four criteria are: force fits genre, valence fits topic, intensity fits valence, register fits newspaper/topic. The criterion is proposals, which is not yet tested with machine generated headlines. This study employs speech act theory to examine news headlines, generative AI, and its impact on the pragmatic evaluation of Pakistani English-language media, using content analysis as its research method.The research method used in this study is content analysis, which is a method for analyzing texts.The research method used in this study is content analysis that is a type of text analysis.

Zitieren

APA

Saleem, M., & Khan, M. A. (2026). A Speech Act baseline for Evaluating AI-generated Headlines: Four Criteria from Pakistani Print Media. Journal of Arts and Linguistics Studies, 4(10), 597-621. https://doi.org/10.71281/jals.v4i10.863

Vancouver

Saleem M, Khan MA. A Speech Act baseline for Evaluating AI-generated Headlines: Four Criteria from Pakistani Print Media. Journal of Arts and Linguistics Studies. 2026;4(10):597-621. doi:10.71281/jals.v4i10.863

BibTeX
@article{Saleem2026,
  author = {Saleem, Maria and Khan, Misbah Afsheen},
  title = {A Speech Act baseline for Evaluating AI-generated Headlines: Four Criteria from Pakistani Print Media},
  journal = {Journal of Arts and Linguistics Studies},
  year = {2026},
  volume = {4},
  number = {10},
  pages = {597--621},
  doi = {10.71281/jals.v4i10.863}
}
Kostenlos starten →