Document is current

Article History

Received: 15 August 2025

Accepted: 19 February 2026

First Online: 7 April 2026

Declarations

:

: On behalf of all authors, the corresponding author states that there is no Conflict of interest.

: This research addresses a pressing societal issue-hate speech on social media-which can significantly impact individual well-being and social harmony. By developing automated detection systems, particularly in low-resource and culturally specific contexts such as Indonesian multimodal content, this study contributes to advancing responsible AI applications that promote healthier online environments. All data utilized in this study were obtained from publicly accessible sources, specifically Twitter (X), and conform to the platform’s Developer Agreement and Policy. The dataset only includes tweet IDs to ensure adherence to data-sharing ethics and privacy regulations. Although the tweets were annotated for hate speech content, we make no claims or assumptions regarding the intent or identity of the original authors. The annotation process was conducted by a team of three trained annotators with prior experience in analyzing social media content. Annotators were fully briefed on the sensitive nature of the task and were explicitly informed that the content might contain harmful or offensive material. They were encouraged to pause or discontinue their participation if the labeling process became emotionally distressing. All annotators received appropriate financial compensation for their contributions. This work promotes transparency and accountability by documenting its methodology and dataset construction in detail. Furthermore, the study emphasizes the importance of cultural sensitivity when analyzing hate speech in multilingual and multimodal settings, particularly in underrepresented regions. No human subjects were directly involved in the study beyond publicly available social media data, and no personal identifiable information (PII) is retained or disclosed.