Online Hate Speech Expressions in Local Languages in Indonesia

"PR2Media understands the strategic importance of content moderation by digital platforms, and for this reason, this process must be equipped with expressions in local languages..."



This report, published by PR2Media, shares findings of research to identify words and phrases frequently used by social media users in three Indonesian provinces to express hate speech based on ethnicity and religion. It is hoped that, as a result, social platforms will be better able to deal with hate speech content that has the potential to trigger social conflict in the languages spoken in those provinces. The report forms part of public outreach and advocacy efforts to promote better content moderation policies in Indonesia.



As explained by PR2Media, "As of January 2023, there were 167 million Indonesian social media users in a country with over 600 local languages dispersed across over 13,000 islands. With a wide variety of languages and dialects, social media platforms face huge challenges in moderating illegal and harmful content in these local languages.



A survey of 1,500 social media users from 38 provinces in Indonesia by PR2Media (2023) showed that the type of illegal content most frequently encountered by respondents was hate speech (67%), followed by disinformation (66%), digital scams (60%), defamation (43%), and pornography (40%)."



This research aims to respond to this phenomenon with the understanding that language is a strategic tool for creating social harmony, especially ahead of Indonesia's 2024 election. The research is based around nine types of "dysphemism" in expressing hate speech identified by Darmawan and Muhaimi (2019). Dysphemia is the use of words or phrases to coarsen the meaning of lexical units so they appear negative to readers and listeners. The nine types are: 1) synecdoche - used to describe something or someone as a whole; 2) dysphemistic epithets - the use of animal names; 3) euphemistic dysphemism - making an attack seem like a compliment by avoiding the use of harsh words; 4) dysphemistic euphemism - a mockery between close friends or family members without any animosity; 5) "-ist" dysphemism - targeted at a particular ethnicity; 6) homosexual dysphemism - used for attacks related to homosexuality; 7) name dysphemism - calling someone by their casual nickname without a proper title; 8) nonverbal dysphemism -  offending someone with gestures; and 9) cross-cultural dysphemism - using slang terms that have a certain meaning in one culture but another meaning in another culture.



Data collection was carried out in West Java, Central Kalimantan, and Papua through focus group discussions (FGDs) where participants were asked to list local words and phrases used frequently by social media users in their provinces prior to the FGDs. With this list of expressions, the researchers sought to learn more about their usage through FGDs. Two main questions were discussed: 1) What words and phrases are frequently used by social media users in the three provinces to express hate speech based on ethnicity and religion? 2) How are the words and phrases used in the local context?



The report explores the findings from each province, with each discussion ending with a list of the identified words and phrases. The study found that three types of dysphemism were most frequently expressed by FGD participants in the three provinces - namely, synecdoche (describing something or someone as a whole), dysphemistic epithets (using animal names), and "-ist" dysphemism (targeting specific ethnicities or tribes). It also found that online hate speech expressions in the provinces are commonly conveyed in Indonesian, local languages, or a combination of Indonesian and local languages.



In general, FGD participants expressed concern about the various forms of ethnic and religious-based hate speech they encountered on social media, as well as hate speech targeting specific genders and professions. They expressed hope that social media platforms will improve the detection of these expressions as they emerge, particularly those conveyed in local languages and in combinations of Indonesian and local languages.



Another finding from the study is that latent tensions or potential hostility between ethnicities and religions in the three provinces are typically related to socioeconomic and political issues, such as unequal distribution of economic resources, political aspirations, and social status between natives and migrants, which also happen to follow certain ethnic and religious lines.



The researchers believe that the words, phrases, and sentences in local languages listed by researchers for each province can be used as data by social media platforms and the government to better moderate or regulate content based on the needs of the people in that province. These findings can also be used as data by other researchers for future research, adding to the body of knowledge about hate speech on social media in Indonesian local languages.



Click here for the Indonesian version of this 56-page report in PDF format.