TY - GEN

T1 - Professional and User Confidence in Synthetic Content

AU - Нигматуллина, Камилла Ренатовна

AU - Касымов, Ренат Масгудович

N1 - Conference code: 18

PY - 2026

Y1 - 2026

N2 - The research is devoted to the problem of recognition and trust of synthetic content in modern media. Recent works demonstrate various approaches to measuring it, which traditionally include surveys, experiments, and simulations. There is evidence that labeling synthetic content affects people’s perceptions in different ways, including depending on the cultural context. Audience segments and professional journalists are being studied separately. So far, no attention has been paid to novice professionals who are most affected by new technologies. The study included interviews with 12 media managers and media experts on the topic of trust in synthetic content and labeling, as well as an experiment involving 143 students of the Bachelor of Journalism program, dedicated to recognizing labeled and unlabeled texts prepared by both humans and large language models. The results of the study confirm the hypotheses that 1) explicit disclosure of information that a news text was created using artificial intelligence reduces perceived trust among journalism students, with a stronger effect observed in high-risk information contexts; 2) Disclosure of information using artificial intelligence serves as a signal of transparency, which increases the perceived transparency of the content creation process, but does not increase the perceived trust in the results.; 3) journalism students demonstrate incorrect confidence in text recognition using artificial intelligence, so that subjective confidence in the identification of texts created using artificial intelligence exceeds objective classification accuracy; previous media experience increases confidence, but does not increase accuracy; 4) when establishing authorship, journalism students rely primarily on linguistic and stylistic cues at the surface level, rather than reliable indicators of the underlying content creation processes that lead to systematic attribution errors; 5) despite critical attitudes towards the use of artificial intelligence in journalism, journalism students do not always show less confidence in news texts. created with the help of artificial intelligence, when there are no clues about the disclosure of information.

AB - The research is devoted to the problem of recognition and trust of synthetic content in modern media. Recent works demonstrate various approaches to measuring it, which traditionally include surveys, experiments, and simulations. There is evidence that labeling synthetic content affects people’s perceptions in different ways, including depending on the cultural context. Audience segments and professional journalists are being studied separately. So far, no attention has been paid to novice professionals who are most affected by new technologies. The study included interviews with 12 media managers and media experts on the topic of trust in synthetic content and labeling, as well as an experiment involving 143 students of the Bachelor of Journalism program, dedicated to recognizing labeled and unlabeled texts prepared by both humans and large language models. The results of the study confirm the hypotheses that 1) explicit disclosure of information that a news text was created using artificial intelligence reduces perceived trust among journalism students, with a stronger effect observed in high-risk information contexts; 2) Disclosure of information using artificial intelligence serves as a signal of transparency, which increases the perceived transparency of the content creation process, but does not increase the perceived trust in the results.; 3) journalism students demonstrate incorrect confidence in text recognition using artificial intelligence, so that subjective confidence in the identification of texts created using artificial intelligence exceeds objective classification accuracy; previous media experience increases confidence, but does not increase accuracy; 4) when establishing authorship, journalism students rely primarily on linguistic and stylistic cues at the surface level, rather than reliable indicators of the underlying content creation processes that lead to systematic attribution errors; 5) despite critical attitudes towards the use of artificial intelligence in journalism, journalism students do not always show less confidence in news texts. created with the help of artificial intelligence, when there are no clues about the disclosure of information.

KW - Russia

KW - artificial intelligence

KW - credibility

KW - journalism

KW - mass media

KW - trust

UR - https://www.mendeley.com/catalogue/25bf33c5-b9ac-3865-869c-acdaae1f67d4/

U2 - 10.1007/978-3-032-31407-9_34

DO - 10.1007/978-3-032-31407-9_34

M3 - Conference contribution

SN - 978-3-032-31406-2

T3 - Lecture Notes in Computer Science

SP - 522

EP - 532

BT - Social Computing and Social Media

PB - Springer Nature

Y2 - 26 July 2026 through 31 July 2026

ER -