Can Content Moderation Rely Solely on AI?
In recent years, there has been increasing pressure on social media platforms to restrict the spread of harmful content, including hate speech, rumors, and other forms of online harassment. Traditional media have always relied on human moderators to keep the peace, but the massive amounts of content being produced online render human moderation impossible. In response, an increasing number of social media sites rely on companies who utilize AI for providing content moderation services.
Penn State University carried out a study to determine if social media users have as much trust in artificial intelligence (AI) as they do in human editors to identify harmful or offensive posts.
Pros & Cons of AI and Human Content Moderation
You have to understand that content moderation services, both human and artificial, have their benefits and drawbacks. Human moderators are more accurate when determining whether or not content is unsafe, like provoking self-harm. However, people can’t keep up with the volume of content that’s being created and shared online. But the fact is that even though AI editors can quickly analyze content, many people still don’t trust their recommendations or worry that their access to certain information will be restricted.
Humans and AI: The Dynamic Duo of Content Moderation
Researchers found that combining human and AI moderators could be an effective strategy for creating a trustworthy AI-powered moderation system. One way to increase people’s faith in AI is to be more transparent about the fact that a machine or automated system is moderating. Researchers found that interactive transparency, in which users can provide input to the AIs, appears to further increase user trust.
In this study, participants’ trust in AI content editors was tested across 18 different conditions based on factors such as the moderation source and transparency. They recruited 676 individuals to use a system for content classification and test both static and dynamic transparency. Researchers examined how content was categorized, specifically whether it was marked as harmful or hateful (“flagged”) or not (“not flagged”). Content deemed “harmful” dealt with suicidal thoughts, while content deemed “hateful” included hate speech.
The researchers discovered, among other things, that users’ trust in an AI content moderator depends on whether the presence invokes advantageous features of machines, including their precision and objectiveness, or unfavorable attributes, like their inability to arrive at subjective judgments about specifics of human language. Allowing users to assist the AI system in determining the safety of content posted online may also increase their trust. According to the study, participants who supplemented the list of words suggested by AI categorized posts with their own terms trusted the AI editor as much as they would have trusted a human editor.
The Relevance of Automated Content Moderation
The benefits of automating the content review process extend beyond relieving human laborers of everyday tasks. If human moderators are used, they will be subjected to hours of violent and hateful material. To protect them, who are providing a public service by removing objectionable material, researchers argue that automated content moderation is a better choice.
The researchers suggest that future efforts could focus on educating the public so that they can better trust and utilize AI. They stress the need to involve the public so they can grasp the concept of artificial intelligence. Interactive transparency may be essential to comprehending AI. They also pose crucial questions, such as how to present AI in a way that properly balances and produces caution about its limitations. These are all excellent research questions.
Summary
While there may be moral questions about using AI to moderate content, it is of the utmost importance to shield human moderators from harmful materials. In the future, scientists should investigate methods for increasing the public’s confidence in and familiarity with AI. Last but not least, researchers can improve AI-powered content moderation systems by first gaining an understanding of the advantages and disadvantages of AI. You can get guidance from a professional AI company like the Opporture in North America for the best AI content moderation services.
