Claim Missing Document
Check
Articles

Found 1 Documents
Search

HYBRID TRAINING IN STYLISTIC DETECTION OF HOAX AND HATE SPEECH LANGUAGE: KARANG TARUNA TUNAS MUDA KLAIGIT Yeni Witdianti; Nursalim Nursalim; Ismail Marzuki; Kumaidi Kumaidi
IJCDE (Indonesian Journal of Community Diversity and Engagement) Vol. 7 No. 1 (2026): Vol. 7 No. 1 , 2026
Publisher : LEMBAGA PENELITIAN, PENGABDIAN PADA MASYARAKAT, PENINGKATAN AKTIVITAS INSTRUKSIONAL, PENINGKATAN DAN PENJAMINAN MUTU UNIVERSITAS ISLAM MAJAPAHIT

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36815/pengabdian.v7i1.4788

Abstract

The circulation of hoaxes and hate speech in Indonesia has shifted from fabricated factual claims towards affectively framed content that fact-checking alone cannot intercept. Existing community engagement programmes on digital literacy predominantly train verification procedures, leaving participants unable to detect provocative intent in texts that contain no verifiable proposition. This programme addressed that gap by training members of a youth organisation to recognise the linguistic markers through which provocative texts operate, using a stylistic analytical framework. The programme was delivered in a hybrid mode across five stages: asynchronous online initial assessment, two days of face-to-face core training, two synchronous online deepening sessions with a guest lecture on critical discourse analysis, three to four weeks of asynchronous mentoring in counter-narrative production, and a face-to-face showcase and evaluation. Participants were 20 members of Karang Taruna Tunas Muda Klaigit. Data were collected through a pre-test and post-test comprising ten texts, participant worksheets, and the counter-narrative products, then analysed using the normalised gain formula, the Wilcoxon signed-rank test, and qualitative description. The mean score rose from 21.00 to 40.85 on a 100-point scale. Every participant improved, and the increase was statistically significant (W = 0; z = -3.920; p < 0.001) with a large effect size (r = 0.877). The normalised gain of 0.251 nevertheless falls in the low category, indicating that participants improved consistently but remained far from mastery. Disaggregation by aspect reveals an instructive pattern: classifying a text as provocative gained most (g = 0.392), explaining its persuasive function followed (g = 0.274), while naming the linguistic marker itself gained least (g = 0.176). Participants therefore acquired sharpened intuition more readily than analytical vocabulary. The programme demonstrates that stylistic competence, conventionally confined to literary analysis, can be reformulated as an accessible early-detection instrument for lay citizens, but that a single training cycle is insufficient for the analytical layer. Sustained impact requires the cadre scheme, continued online mentoring, and denser repeated practice on each marker.