What gaps does #TrackerBenci fill in current hate speech monitoring?

Question

Accepted Answer

#TrackerBenci Addresses Limitations in Current Hate Speech Monitoring Systems
TrackerBenci, an AI-powered programme developed by the KL-based think tank The Centre, aims to address significant gaps in existing hate speech monitoring. These systems often rely heavily on human moderators, a method that is becoming increasingly unreliable due to the sheer volume of online content and the challenges of culturally nuanced interpretation. The AI tool is designed to complement human efforts, offering a more scalable and consistent approach to identifying and understanding online hate speech trends within Malaysia.
Over-reliance on Human Moderators and Cultural Nuance Gaps

Current hate speech tracking systems frequently depend on human moderators to sift through vast amounts of online content. This approach faces limitations due to the sheer scale of data, making it unsustainable in the long term. Furthermore, a lack of investment in locally-trained and diverse moderators creates blind spots in understanding hate speech that is specific to certain cultural contexts within Malaysia, as noted by researcher Tham Jia Vern.
Scalability and Long-Term Reliability Challenges

The continuous and overwhelming flow of social media content poses a significant challenge for human-only moderation systems. #TrackerBenci is intended to provide a more scalable solution. By leveraging AI, it can process a larger volume of data than human moderators alone, thereby improving the efficiency and long-term reliability of hate speech detection efforts.
Complementary Hybrid Approach for Enhanced Detection
TrackerBenci is not designed to replace human moderators entirely but rather to work alongside them. The think tank proposes a hybrid approach where the AI assists in the initial detection and classification of harmful online content. This partnership can enhance the accuracy and comprehensiveness of monitoring, allowing human moderators to focus on more complex cases or provide further analysis.
Training and Development of the AI Model

The development of #TrackerBenci involved training the AI on approximately 20,000 tweets curated by a team of six researchers, including Tham Jia Vern and Nelleita Omar. This training process taught the AI to differentiate between "hateful" content, "snarky replies," and content that is "not hateful or irrelevant to the Malaysian context." The AI was further trained to identify the intensity of hateful tweets, categorising them as "insult" or "threat to marginalisation," and to recognise specific themes such as "violent threat," "increases intergroup tension," "sexism," and "perpetuates a negative stereotype."

Reading history

Hide →

This history is cleared when the page is closed.

What gaps does #TrackerBenci fill in current hate speech monitoring?

Loading

Was this article helpful?

Helpful

Not helpful

Report an issue

1

0

0

Share with friends