Page 21 - Tvorjenje pomena in večrazsežna – večkodna pismenost
P. 21
Simona Kranjc Meaning Without an Author? A Comparative Discourse
University of Ljubljana, Slovenia Analysis of Human- and Machine-Generated Commentary
simona.kranjc@ff.uni-lj.si
©2026SimonaKranjc This presentation explores how meaning is constructed in human-written and
machine-generated texts through a comparative discourse analysis of commen-
taries published in response to the release of a book by Branko Grims (2024). The
analysis is based on the journalistic commentary ‘Grimsove pravljice,’ published
in the weekly Mladina (Volk, 2024), which serves as an example of a clearly po-
sitioned and ideologically marked public response. To enable comparison, three
additional commentaries on the same topic were produced using three different
large language models, all responding to the same communicative task with a
defined audience, genre, and length.
The analysis is based on an understanding of meaning as a discursive effect rather
than a fixed property of text (Fairclough, 1995; van Dijk, 2008; Gumperz, 1982). It
compares the four commentaries across several key dimensions: how a stance
is articulated and maintained, how the audience is addressed, which rhetori-
cal strategies are employed, and how arguments develop semantically over the
course of the text. Particular attention is paid to the tension between authorial
intentionality and automated text generation, as well as to the ways in which
implicit meanings and ideological orientations emerge in machine-generated
commentaries despite the absence of a human author who could be held discur-
sively accountable.
The findings suggest that machine-generated commentaries can successfully
reproduce the formal and generic features of journalistic commentary and can
present arguments that are coherent and internally consistent. At the same time,
notable differences emerge when these texts are compared with the human
commentary. Machine-generated texts tend to adopt more cautious pragmatic
strategies, take fewer discursive risks, and display a more stable but also less dy-
namic evaluative stance. As a result, they often appear less conflict-driven and
semantically less open.
By highlighting these differences, the presentation contributes to ongoing dis-
cussionsabouttheroleofartificialintelligenceincontemporarymeaning-making
practices (Bender et al., 2021) and raises broader questions about authorship, re-
sponsibility, and the place of the human beings within multilayered and multi-
modal forms of literacy (Kress, 2010).
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers
of stochastic parrots: Can language models be too big? In FAccT ’25: Proceed-
ings of the ACM Conference on Fairness, Accountability, and Transparency (pp.
610–623). Association for Computing Machinery.
Fairclough,N.(1995). Criticaldiscourseanalysis:Thecriticalstudyoflanguage. Long-
man.
Gumperz, J. J. (1982). Discourse strategies. Cambridge University Press.
Meaning-Making, Multiliteracies
Grims, B. (2024). Zmaga dobrega: vodnik za boj proti kulturnemu marksizmu. M.
and Multimodality
Abstracts of the International Dragoš Grims.
Symposium Kress, G. (2010). Multimodality:Asocialsemioticapproachtocontemporarycommu-
Koper, 19–20 March 2026 nication. Routledge.
Volk, L. (2024, 16 February). Grimsove pravljice. Mladina.
Van Dijk, T. A. (2008). Discourse and power. Palgrave Macmillan.
https://doi.org/10.26493/978-961-293-565-8.18 21
