Global tech giant TikTok has stepped up its efforts to ensure the integrity and safety of content shared on its platform, especially during election periods, as social media continues to play a critical role in shaping public opinion and political discourse.
At a recent Trust and Transparency Summit held at TikTok’s offices in Singapore, the company showcased its advanced content-moderation strategies aimed at preserving electoral credibility and reinforcing public trust in democratic processes.
TikTok employs a dual approach to content moderation, leveraging both artificial intelligence (AI) and human moderators.
The platform removes about 2-million pieces of content daily, with 77% identified by AI systems and the remaining 23% by human moderators.
These AI systems are capable of detecting harmful content within seconds, scanning videos for dangerous objects, alcohol, extremist symbols, and smoking.
The AI’s ability to detect harmful objects in videos is rooted in sophisticated computer-vision techniques and machine-learning models.
These models are trained on extensive data sets containing labelled images of various objects.
They also help identify and locate objects within video frames by extracting features such as shapes and textures.
Human intervention
For instance, the context in which an object appears — like a knife in a kitchen versus in public — can significantly influence the detection process.
However, despite these advancements, AI systems often fall short in contextual nuances, necessitating human intervention, the company said.
To remedy the shortfalls, TikTok said it has employed over 40 000 trust and safety human moderators, including experts native to each country where the app is available.
“Machines can detect things like objects and posture, but they struggle to pick up context, hence the need for human moderators.”
The company explained that human moderators play a crucial role in understanding and moderating context-specific content, especially during sensitive periods like elections.
During elections, social media can become a hotbed for misinformation and hate speech. TikTok’s content-moderation strategies are designed to mitigate these risks.
The company confirmed that the platform also operates around-the-clock moderation, using sensitive word lists to detect potentially harmful content in captions, hashtags, comments, and profile bios.
Accuracy of the content
Content on TikTok either gets removed for violating community guidelines or is immediately published.
However, the company has stated that even approved and published content can still be removed if reported by other users.
During election season, misinformation becomes a significant concern for the company.
To remedy this, TikTok has partnered with various government departments and fact-checking organisations, such as Africa Check, the International Fact-Checking Network, Politifact, and Thomson Reuters, among others, to ensure the accuracy of the content.
In South Africa alone, TikTok has implemented stringent content-moderation protocols.
The platform removes about 721 120 videos, achieving a proactive removal rate of 96.1% and a removal rate within 24 hours of 86.3%.
TikTok also collaborates closely with local law enforcement and regulatory bodies.
In South Africa, the platform has onboarded the Independent Electoral Commission and the South African Human Rights Commission to ensure compliance with national laws and policies.
Keeping harmful misinformation off the platform
The team at TikTok explained that law-enforcement teams within TikTok handle illegalities detected on the app; however, they often require court orders to access user information to balance enforcement with user privacy.
With regards to election integrity, TikTok uses the Safety Enforcement Tool, which provides a centralised, secure portal for submitting and reviewing takedown requests.
This system ensures that only verified agencies and the government can request content removal, the company said.
“Protecting the integrity of our platform, particularly around elections, is a responsibility we take seriously,” TikTok spokesperson Fortune Mgwili-Sibanda stated.
“We’re proud to be a place that brings people together; we work hard to keep harmful misinformation off our platform.”
