User Generated Content Moderation: Challenges of Human Moderation

User Generated Content (UGC) means any kind of media (photos, videos, comments, and product reviews) that is generated by users other than the company owning or operating the platform.  

Inappropriate content categories might differ from one platform to another based on the target users. Content that can harm or is illegal must be blocked regardless of the website hosting it.

In the past ten years, billions of people have used the online space to gain a wide variety of information. So, sharing ideas and opinions with a wider audience has led to growth in businesses, raising money for charity, and has aided in bringing about a political change.

The UGC community has also enabled the searching and dispersion of content which can lead to harm including terror propaganda, abuse, child nudity exposure, among many. The ability to gain access to online content sharing has also led to creation of tons of trash known as digital garbage. Unsupervised UGC can result in harming brands and sensitive groups.

Challenges in Moderating UGC

Online content moderation becomes a challenging task if your app or system can be easily circumvented. The growth in online communities along with companies permitting more UGC than ever before on their website has resulted in users controlling the tone and image of the community. This impacts users in a negative as well as positive way.

User interaction definitely drives engagement, but companies must know that users will not prioritize company’s best interest at all times. They may breach it by posting inappropriate content giving rise to the inevitable need to monitor UGC. Hence, a number of companies are investing in filtering technology for safeguarding their online presence. Moreover, setting up filters is not that easy as users can easily navigate around automated filters.

A majority of content moderation is done after the content is posted. Hence, it’s essential to review the brand’s existing content once it has been published and make edits or deletions as required. UGC requires one to be proactive than reactive to ensure users have a great experience.

There are many challenges to moderating UGC, however, the primary one remains filter circumvention which includes;

1. Single words used both as profanity and proper term;

2. Embedded words where the proper term or proper name has profanity;

3. Non-dictionary words where dictionary words are deliberately misspelled by users;

4. Character replacement where some characters are interchanged or appear almost the same;

5. Swapping where the characters within a word are changed with other alphabetic characters which do not alter the word’s phonetic structure;

6. Collapsing where non-essential characters are taken out without altering the phonetic structure of the word;

7. Repetition where characters are repetitive within a word;

8. Abbreviation where an unwanted phrase is abbreviated by users to remain unfiltered;

9. Separators where unwanted words or phrases are divided up using spaces, periods, or letters;

10. Image processing where a word is dispersed across various lines.

Human Content Moderation versus AI Content Moderation

The rise of social media has made content moderation the biggest and most secretive operative functions in the industry. There are groups of moderators across the world scanning the Internet for violence, pornography, hate speech, and stacks of inappropriate or illegal content.

Human 1: Ethical: There are tons of evidence to reveal that people who are exposed regularly to disturbing and hurtful content undergo psychological harm. It has resulted in employees being diagnosed with PTSD.

AI: Addresses Ethical & Economic Issues: AI Content Moderation addresses both the ethical and economic sides of an issue.

Human 2: Costly and Delayed: For supporting real-time and live streaming moderation.

AI: Bulk Data Processing: The AI algorithm is capable of processing and categorizing a large amount of data in real-time.

Human 3: Exponential Growth in Content Volumes: Moderation teams perform at a slow pace, hence it’s time-consuming.

AI: Can Scale Up Quickly: It can process large volumes of data at much faster rate, hence saving time.

Human 4: Training People: Since manual moderations requires involving lots of people, it becomes tough to amend policies or incorporate all of a sudden a new kind of content that needs recognition.

AI: Moderation Without Context: The only area where AI Content Moderation falls short is where there is a gap in contextual understanding of a situation. For instance, a conflict that’s been shown on TV as a scoop can be seen as promoting violence on social media and hence moderated by AI accordingly.

Human: 5. Overlooking Inappropriate Content: In case of content that’s been flagged as inappropriate, the machine never overlooks it, however, in manual moderation, there’s a high probability of such mishaps occurring.

AI: Highly Precise: It is highly capable of recognizing visuals in any content category including pornography, weapons, fights, etc. in a precise manner.

Conclusion:

From the above, we can safely conclude that a hybrid approach involving humans and machines is the best to filter the massive amount of illegal content generated online on a daily basis. AI algorithms can process the bulk of the content and send out just a small proportion to human moderators thereby limiting the workload of hundreds of thousands of psychologically harming content moderator positions. Finally, it is very productive and easy to scale.