How to Enable AI Content Moderation in eDirectory

Introduction

Keeping your online community safe and compliant is a top priority, especially with the constant influx of user-generated content. Fortunately, eDirectory now offers a powerful solution: AI-powered content moderation. This feature, which leverages OpenAI's ChatGPT moderation endpoint, automatically filters out potentially inappropriate, offensive, or policy-violating content. It's a game-changer for maintaining the integrity of your directory. Let's dive into how you can enable and effectively use this essential tool.

Essential Requirements Before Activation

Before you can harness the power of AI content moderation in eDirectory, a foundational step involves configuring your OpenAI integration. This ensures that your eDirectory platform can communicate effectively with OpenAI's moderation services.

To begin, follow these crucial steps:

• Access the Admin Panel (Sitemgr): Log into your eDirectory administrator interface.
• Navigate to AI Integrations: Go to Settings > General Settings > AI Integrations.

• Enter Your OpenAI API Key: Provide your valid OpenAI API key in the designated field. This key acts as your credential for accessing OpenAI's services.

• Enable Content Moderation: Check the box labeled "Enable content moderation with AI". This activates the moderation functionality within eDirectory.

• Choose Your Moderation Confidence Level: This critical setting allows you to define the strictness of the AI's moderation. Select one of the three options based on your platform's needs:
  • Flag everything, even if AI isn't sure: This is the most rigorous option. All content, including submissions with a low confidence score (below 0.5) from the AI, will be flagged for a full manual review by administrators. Ideal for platforms requiring stringent monitoring.
  • Flag content if AI is somewhat or very sure: This offers a balanced approach. It only flags content that the AI deems likely to violate your guidelines (a moderate to high confidence level). This is generally recommended as it helps to reduce false positives while still catching most problematic content.
  • Only flag when AI is very sure it's a problem: This is the most permissive setting. Content will only be flagged if the AI has a high degree of confidence that it is problematic. While it allows for broader discussions, it might miss some subtle or borderline issues.

Once you have saved these settings, the eDirectory system will immediately begin applying content moderation automatically to new submissions.

Content Under AI Scrutiny: What Gets Moderated?

Upon activation, the AI content moderation feature diligently processes various types of user-submitted content. This content is automatically sent to the OpenAI moderation endpoint for evaluation:

• Leads/Messages:
  • Messages sent to Listings
  • Messages sent to Events
  • Messages sent to Classifieds
• Reviews: Reviews submitted on listings
• Community Forum: Questions and answers posted within the Community Forum plugin
• Lead Generation Forms: Submissions through Lead Gen Widgets
• Contact Forms: General contact form submissions

The Moderation Workflow: What Happens to Flagged Content?

The AI moderation process is designed to prevent problematic content from appearing on your live site, ensuring a clean and safe user experience:

• Automatic Hold: Any content flagged by the AI will not be published automatically. Instead, it is immediately held for manual review.

• Mandatory Administrator Review: Even if you have "automatic review publishing" enabled for general content, flagged submissions will only be published after a site administrator manually approves them.

The Administrator's Moderation Panel: Your Control Center

To facilitate efficient content management, eDirectory provides a dedicated moderation panel for administrators to review flagged content. This panel offers comprehensive details to aid in decision-making:

• Content Details: Displays the flagged message itself, along with relevant user data, the timestamp of submission, and an assigned "threat level" indicating the AI's confidence in the violation.
• Threat Categorization: Crucially, the panel provides a detailed breakdown of the specific violation types detected by the AI. These categories can include:
  • Hate speech
  • Harassment
  • Violence
  • Explicit sexual content
  • Self-harm
  • And other predefined categories by OpenAI.

Administrator Action: Approval or Rejection

Once content appears in the moderation panel, administrators have two clear options:

• Approval: If the content is reviewed and deemed appropriate, approving it will result in its instant publication on the site.
• Rejection: If the content is found to be truly inappropriate or violating policies, rejecting it will lead to its permanent deletion from the system.

Conclusion

Integrating AI-powered content moderation into your eDirectory is a powerful step towards a cleaner, more secure, and ultimately, more trustworthy online environment. By proactively identifying and holding potentially problematic content, you can significantly reduce manual oversight and ensure that your platform remains appropriate for all users. This not only protects your brand and fosters a positive user experience but also allows you to focus on growing your community with confidence.