Review Reports - The Efficiency–Verification Paradox of Generative AI Use: Exploring the Tension Between Productivity and Accuracy

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an interesting article and well-written. Its aim is to examine how
generative AI is experienced by journalists and teachers within the contemporary information ecosystem by analysing 10 interviews based on the argument the "Efficiency–Verification Paradox of Generative AI"

To my opinion, the article presents a good introduction and a good literature review that could benefit from more research conducted on the field. However, there is a question raised of what field since it deals with different fields, journalism and education... It is a bit strange to mix these two totally different fields in one research, however, I think the author(s) has found a way to combine them.

One problem is the small amount of participants, especially considering that they are divided in two (journalists and educators). There is also a problem concerning the different way the material was gathered, since some were written and other oral which brings out a question of using not equivalent data, as written responses tend to be shorter and lack interviewer follow-up. It could be to specify how many participants used which mode, and include that as a limitation of the study (there are not any limitations at the end of the manuscript). Also, there is an issue of how reflective analysis was performed (there is not a single reference on the method used) and whether the author is mature enough or involved enough in the fields under study to perform such a method.

There is one praticipant stating "I don’t use AI tools… The sentences they produce are usually sterile and full of clichés.", however, one out of five is 20% of the participants which is not a negligible number. And in the rest of the manuscript I am not sure I have read anything regarding the specific participant. I am pointing this out since at the discussion and conclusions section there seems to be a generalization of the findings that it is not in accordance with the small sample, including someone who is against using AI.

Another issue is that you mention education in general, however, there are only interviews with informatics teachers in primary schools which also limits the generalization of the findings.

In the discussion section there is more a repetition of the results rather than their interpretation and the critical aspect of the researcher(s). I think this part should be smaller in lenght and more precise to the aspects derived by the research.

Lastly Figure 1 is not properly presented.

I would recommend a closer study of the manuscript so it will be come publishable.

Author Response

Response to Reviewers

I would like to sincerely thank the reviewers for their careful reading of the manuscript and for their constructive and insightful comments. I greatly appreciate the opportunity to revise the manuscript. The reviewers’ suggestions helped me clarify the theoretical foundation of the study, strengthen the empirical and methodological justification, improve the interpretation of the findings, and make the scope and limitations of the study more explicit. In the revised manuscript, I have addressed all comments point by point. The major revisions include: (1) strengthening and expanding the empirical literature review; (2) introducing the Social Construction of Technology (SCOT) framework as the principal theoretical perspective; (3) incorporating human oversight as a complementary analytical concept; (4) clarifying the data collection procedures, including the use of written responses; (5) strengthening the methodological justification of reflexive thematic analysis; (6) explicitly addressing the study limitations; (7) revising the interpretation of the participant who reported not using AI; (8) narrowing claims concerning teachers to informatics teachers in Croatian primary schools; (9) substantially revising the Discussion to focus on interpretation rather than repetition of results; and (10) removal of Figure 1, leaving only written explanation. (11) Considering that Reviewer also raised concerns about and problematized the use of the term “education in general,” I suggest that, in addition to the proposed revisions to the text, and provided that the editors and reviewers agree, I also change the title by removing the term „education,” as it is too broad. The new title would be:

„The Efficiency-Verification Paradox of Generative AI Use: Exploring the Tension Between Productivity and Accuracy“.

A revised version of the paper is attached. Changes in the manuscript are highlighted in red.

Response to Reviewer 1

Comment 1: The combination of journalism and education

Reviewer comment:
The article examines two different fields, journalism and education. It is somewhat unusual to combine these two fields in one study, although the reviewer acknowledges that the author have found a way to connect them. The reviewer asks, implicitly, for greater clarification of the field and rationale for comparison.

Response:
Thank you for this important observation. I agree that journalism and education represent distinct professional fields and that their combination requires a clear conceptual justification. In the revised manuscript, I have clarified that the study does not treat journalism and education as a single professional field. Rather, they are examined as two interconnected professional domains within the contemporary information ecosystem that perform complementary functions in relation to information. Journalists are primarily involved in the production, verification, contextualisation, and public dissemination of information, whereas teachers contribute to the interpretation of information and the development of students’ abilities to critically evaluate information. The comparison therefore focuses not on the similarity of the professions themselves, but on their shared relationship with AI-mediated information practices, professional judgement, verification, and responsibility. This rationale has been strengthened in the Introduction, Literature Review, Theoretical Framework, Discussion, and Conclusions. The revised theoretical framework, based primarily on the Social Construction of Technology perspective, further provides a common analytical basis for examining how different professional groups negotiate the meaning, appropriate uses, and boundaries of GenAI.

Comment 2: Small sample and two different modes of data collection

Reviewer comment:
The sample is small, particularly because it is divided into two professional groups. The reviewer also raises concerns about the use of both oral and written responses, noting that written responses may be shorter and do not permit interviewer follow-up. The reviewer recommends specifying how many participants used each mode and including this as a limitation.

Response:
I agree with this comment and have clarified the data collection procedure. The revised manuscript now explicitly states that three of the ten participants provided written responses: one journalist and two teachers, while the remaining seven participants participated in oral interviews. I have clarified that the written participants received and completed the same semi-structured interview protocol used in the oral interviews. I also acknowledge that the two modes of data collection are not fully equivalent. In particular, written responses do not provide the same opportunity for spontaneous follow-up questions, clarification, or elaboration as synchronous interviews. This issue has now been explicitly acknowledged as a limitation of the study. I have also revised the methodological discussion to avoid implying that all data were collected through identical interactional conditions. The revised manuscript therefore provides greater transparency concerning the nature and limitations of the dataset.

Comment 3: Reflexive thematic analysis and researcher reflexivity

Reviewer comment:
The reviewer asks how reflexive thematic analysis was conducted, notes that there was no methodological reference, and raises a question regarding whether the researcher was sufficiently experienced or involved in the fields under study to conduct such analysis.

Response:
I appreciate this comment and have substantially strengthened the methodological explanation. I now identifiy reflexive thematic analysis (RTA) as the analytical approach and cites Braun and Clarke (2006, 2021). The revised manuscript explains the six recursive phases of reflexive thematic analysis, including familiarisation, coding, development of initial themes, review and refinement of themes, definition and naming of themes, and production of the analytical report. I have also clarified that coding was conducted inductively while remaining informed by the research questions and theoretical framework, and that codes and themes were repeatedly reviewed and refined across the two professional groups. The revised manuscript explicitly acknowledges that qualitative interpretation is shaped by the researcher’s prior knowledge, theoretical assumptions, and disciplinary perspective. The researcher’s interpretative role is therefore treated as part of the analytical process rather than as something that can be eliminated.

Comment 4: The participant who does not use AI

Reviewer comment:
The reviewer notes that one journalist reported not using AI and argues that this represents 20% of the journalist sample and should not be overlooked. The reviewer also expresses concern that the Discussion and Conclusions may generalise findings despite this divergent position.

Response:
I agree with this observation and have revised the manuscript accordingly. The participant who reported not using AI tools is now explicitly acknowledged as a divergent or contrasting position within the journalist group, rather than being treated as an insignificant exception. The revised Results section retains and contextualises the quotation concerning the participant’s concerns about the style, authenticity, and stereotypical character of AI-generated text. At the same time, I have avoided interpreting this participant as representing a general opposition to AI among journalists. Rather, the finding is presented as evidence of variation in professional responses to GenAI, ranging from active use to cautious or non-use. The Discussion and Conclusions have also been revised to avoid statistical or population-level generalisations. The findings are now explicitly framed as qualitative, exploratory insights into the experiences of the ten participants, rather than as representative claims about journalists or teachers as populations.

Comment 5: Use of the term “education” in general

Reviewer comment:
The manuscript refers to education generally, although the empirical sample consists only of informatics teachers in Croatian primary schools. The reviewer notes that this limits generalisation.

Response:
I fully agree and have narrowed the terminology throughout the revised manuscript. I have clarified throughout the manuscript when statements refer specifically to the empirical participants and when they refer to the broader educational context. Claims derived from the empirical data are explicitly limited to informatics teachers in Croatian primary schools, and the Limitations and Future Research section now states that the findings cannot be generalised to teachers across subjects, educational levels, or national contexts. Considering that Reviewer also raised concerns about and problematized the use of the term „education in general,” I suggest that, in addition to the proposed revisions to the text, and provided that the editors and reviewers agree, I also change the title by removing the term “education,” as it is too broad. The new title would be:

„The Efficiency-Verification Paradox of Generative AI Use: Exploring the Tension Between Productivity and Accuracy“.

Comment 6: Discussion is repetitive and insufficiently interpretive

Reviewer comment:
The reviewer observes that the Discussion largely repeats the Results and recommends making it shorter, more precise, and more critically interpretive.

Response:
I agree with this criticism and have substantially revised the Discussion. The revised Discussion no longer primarily restates the empirical findings. Instead, it is organised around the conceptual interpretation of those findings. In particular, the revised Discussion focuses on AI as a form of professional transformation rather than simple replacement; the efficiency–verification paradox as the central interpretive contribution; verification as a shared responsibility connecting journalism and education; and the limitations and directions for future research. The revised Discussion also explicitly connects the empirical findings to the SCOT framework and the complementary concept of human oversight. This enables the findings to be interpreted in relation to existing scholarship rather than simply summarised. The revised Discussion is therefore more focused on what the findings mean, why the similarities and differences between the two professional groups matter, and what the findings contribute to the existing literature.

Comment 7: Figure 1 is not properly presented

Reviewer comment:
The reviewer states that Figure 1 is not adequately presented.

Response:
Thank you for pointing this out. Figure 1 has been removed, and the underlying conceptual relationship is now explained directly in the text.

Comment 8: Absence of limitations

Reviewer comment:
The reviewer notes that there was no limitations section.

Response:
I agree and have addressed this issue in the revised manuscript. There is now Section „Limitations and Future Research“ and now explicitly discusses the principal limitations of the study. These include the small purposive sample, the qualitative design, the combination of oral and written responses, the specific national context of Croatia, the restriction of the teacher sample to informatics teachers in primary schools, and the exclusion of other relevant stakeholders such as students, editors, media audiences, and policymakers. The revised manuscript also explicitly states that the findings are not intended for statistical generalisation.

Additional revisions made in response to the reviews

In addition to the specific points above, I have carefully reviewed the entire manuscript for consistency between the research questions, theoretical framework, methodology, results, discussion, conclusions, and references. The major conceptual revision is that the manuscript now consistently treats SCOT as the main theoretical lens, while human oversight functions as a complementary analytical concept. This distinction has also enabled me to formulate the efficiency–verification paradox more precisely. Rather than presenting the paradox simply as an empirical observation that AI can both save and consume time, the revised manuscript interprets it as a consequence of the way professional groups negotiate and stabilise the role of GenAI: routine cognitive tasks may be delegated to AI, but the resulting outputs generate additional requirements for verification, contextual evaluation, professional judgement, and accountability. I have also revised the Conclusions to ensure that claims remain proportionate to the qualitative and exploratory nature of the study and to the characteristics of the sample. I am grateful to both reviewers for their constructive comments. Their observations have helped me substantially improve the theoretical grounding, methodological transparency, empirical contextualisation, and analytical precision of the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

There are areas in the study that require further attention. In the section titled Theoretical Framework, there were no theories specified. It would be better to use the subheading 'Literature Review'. The omission of a section on Empirical Review needs to be addressed.

Details are found in the attached document.

Comments for author File: Comments.pdf

Author Response

Response to Reviewer 2

„The Efficiency-Verification Paradox of Generative AI Use: Exploring the Tension Between Productivity and Accuracy“.

A revised version of the paper is attached. Changes in the manuscript are highlighted in red.

I sincerely thank Reviewer 2 for recognising the relevance and originality of the comparative perspective and, in particular, for highlighting the contribution of connecting journalism and teaching as two professional groups involved in different stages of contemporary information practices.

I have carefully addressed all four specific recommendations.

Comment 1: Rename “Theoretical Framework” as “Literature Review”

Reviewer comment:
The reviewer suggests that the section currently titled “Theoretical Framework” should instead be titled “Literature Review.”

Response:
I respectfully followed the reviewer’s broader recommendation concerning the distinction between the literature review and theoretical framework, but I did not simply rename the theoretical section. The manuscript has now been reorganised so that the Literature Review and Theoretical Framework perform clearly different functions. The Literature Review establishes the empirical and conceptual state of research on GenAI in journalism, education, AI literacy, verification, and professional practice. The subsequent Theoretical Framework explicitly identifies and develops the theoretical perspective used to interpret the empirical material.

This restructuring addresses the reviewer’s concern while also retaining a clearly identifiable theoretical framework, as requested in Comment 2.

Comment 2: The article should be underpinned by specific theories

Reviewer comment:
The reviewer states that the article should be grounded in a specific theory or theories, suggesting Technology Determinism Theory, Gatekeeping Theory, and/or Social Construction of Technology (SCOT).

Response:
I fully agree that the original manuscript did not provide a sufficiently explicit theoretical foundation. I have therefore substantially rewritten the Theoretical Framework.

The revised manuscript adopts the Social Construction of Technology (SCOT) framework as the principal theoretical perspective. SCOT is particularly appropriate because the empirical findings concern how different professional groups assign meanings to GenAI, determine appropriate and inappropriate uses, negotiate the boundaries of automation, and retain particular tasks under human responsibility.

The revised theoretical framework draws on the foundational SCOT work of Pinch and Bijker, as well as subsequent developments concerning relevant social groups, interpretative flexibility, closure, and stabilization, including Humphreys (2005).

I deliberately selected SCOT rather than technological determinism because the findings do not suggest that the technological capabilities of GenAI automatically determine professional practice. Instead, journalists and teachers actively negotiate what GenAI should be used for, which tasks can be supported by AI, and which activities must remain under human control.

I have also introduced human oversight as a complementary analytical concept, rather than treating it as a second overarching theory. This concept is informed by work on meaningful human control and human control of AI systems, particularly Santoni de Sio and van den Hoven (2018) and Tsamados et al. (2025). This theoretical structure is now explicitly connected to the empirical findings in the Results and Discussion sections.

Comment 3: Lack of empirical studies

Reviewer comment:
The reviewer identifies the lack of empirical research as a major weakness and recommends incorporating local and international studies concerning AI use in specific professions, particularly journalism and teaching.

Response:
I agree and have substantially expanded and revised the Literature Review. The revised Literature Review now incorporates a broader range of empirical and profession-specific research concerning GenAI in journalism and education. In journalism, the revised manuscript engages with research concerning AI-supported news production, professional experiences, journalistic ethics, verification, credibility, and the changing organisation of newsroom work.

In education, the revised review incorporates research concerning teachers’ AI literacy, preparedness, professional development, responsible AI use, and the role of critical thinking in AI-mediated educational environments.I have also incorporated more recent studies, including research on journalists’ experiences with GenAI and research concerning AI literacy in teacher education. These additions enable the empirical findings of the present study to be positioned more clearly in relation to existing international research.

Importantly, the revised Literature Review now distinguishes between conceptual literature, theoretical literature, and empirical research, thereby addressing the reviewer’s concern that the original manuscript did not sufficiently demonstrate how the present study relates to previous empirical work.

Comment 4: Recommendations based on the findings

Reviewer comment:
The reviewer suggests including a recommendations section, while noting that such a section may not be part of the MDPI format.

Response:
I appreciate this suggestion. Rather than introducing a separate recommendations section, which may not be required by the journal’s format, I incorporated the recommendations into a dedicated subsection, ‘Implications for Professional Practice and AI Governance’ (Section 5.5). This section translates the empirical findings into context-specific implications for journalism and education. The revised section identifies implications for both professional contexts.

For journalism, the findings point to the importance of transparent organisational AI policies, clear verification procedures, editorial accountability, and professional training. For education, the findings indicate the need to strengthen AI literacy within broader digital and media literacy frameworks, provide teachers with opportunities for continuous professional development, and support pedagogical approaches that emphasise critical evaluation of AI-generated information. The revised manuscript also discusses implications for AI governance, particularly the importance of transparency, accountability, and meaningful human oversight.

I believe this approach addresses the reviewer’s recommendation while remaining compatible with the structure and conventions of the journal.

Additional revisions made in response to the reviews

I am grateful to both reviewers for their constructive comments. Their observations have helped me substantially improve the theoretical grounding, methodological transparency, empirical contextualisation, and analytical precision of the manuscript.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author has done a great job in revising the manuscript. All comments were addressed and, to my opinion, it is now in publishable form. Congratulations!

Reviewer 2 Report

Comments and Suggestions for Authors The revision is quite ok. Sometimes, to make a study more engaging, a different section (subhead) dedicated to empirical review is essential. Nonetheless, incorporating the same into the body of the literature review does no harm.