PowerSchool serves over 60 million students across more than 90 countries and 18,000 customers, making student safety a top priority. As we launched PowerBuddy™, our AI assistant integrated into various educational platforms like Schoology Learning and Naviance CCLR, we encountered the challenge of implementing robust content filtering capable of distinguishing between appropriate academic discussions and harmful content within an educational setting. This post details how we built a custom solution using Amazon SageMaker AI to achieve high accuracy while minimizing false positives.
Introducing PowerBuddy: Personalized Learning with AI
PowerBuddy is designed as an AI assistant to offer personalized insights, increase student engagement, and provide support throughout their educational journey. It integrates seamlessly across several PowerSchool platforms, creating a consistent experience for students and their support networks. Notably, the suite includes PowerBuddy for Learning, acting as a virtual tutor; PowerBuddy for College and Career, providing career exploration guidance; and PowerBuddy for Community, simplifying access to district and school information.
Accessibility Features of PowerBuddy
Furthermore, PowerBuddy prioritizes accessibility by incorporating features like speech-to-text and text-to-speech functionality. Therefore, students with diverse learning needs can effectively interact with the AI assistant and benefit from its capabilities. These inclusive design choices demonstrate PowerSchool’s commitment to equitable access to educational resources.
The Critical Need for Specialized Content Filtering in Education
Protecting student safety is paramount at PowerSchool, especially given our extensive reach. National statistics reveal concerning trends: approximately 20% of students ages 12–17 experience bullying, and a significant number of high school students report suicidal ideation. With the increasing adoption of PowerBuddy in K-12 schools, effective content filtering becomes absolutely essential.
Limitations of Generic Content Filtering
Generic, off-the-shelf content filtering solutions often prove inadequate because they lack the necessary understanding of educational contexts. For example, discussions surrounding sensitive historical topics like World War II or the Holocaust should not be flagged as violent content; however, instances of cyberbullying or threats require immediate attention and intervention. We needed a more sophisticated approach—a system capable of discerning between legitimate academic inquiries and truly harmful material related to bullying, self-harm, hate speech, inappropriate sexual content, violence, or other unsuitable subjects.
Amazon SageMaker AI: PowerSchool’s Custom Content Filtering Solution
PowerSchool leveraged Amazon SageMaker AI to build a highly customized content filtering solution. The core of this system involves fine-tuning Llama 3.1 8B, a large language model (LLM), with datasets specifically tailored for educational environments. This approach allows the system to understand nuances and subtleties often missed by generic filters.
Key Components of the Solution
- Fine-Tuning with Educational Data: A significant effort was dedicated to fine-tuning the LLM using education-specific data, enabling it to accurately differentiate between acceptable academic discussions and harmful content.
- Scalable Architecture: SageMaker’s deployment architecture ensures that our content filtering system can handle a large volume of student interactions reliably.
- Rigorous Validation Process: We conducted extensive internal validations to assess the accuracy of the system and minimize false positives, ensuring legitimate educational conversations are not inadvertently blocked. Consequently, this process helps maintain a positive user experience while safeguarding students.
The combined power of specialized training data, Llama 3.1 8B, and SageMaker’s robust infrastructure has resulted in an exceptionally accurate and reliable content filtering system.
Looking Ahead: Continuous Improvement and Student Safety
PowerSchool remains committed to student safety, with ongoing improvements planned for the PowerBuddy AI assistant and its content filtering capabilities. By continually leveraging Amazon SageMaker AI, we strive to provide millions of students with a safer and more enriching learning experience, ensuring they can benefit from the power of AI while remaining protected.
Source: Read the original article here.
Discover more tech insights on ByteTrending.
Discover more from ByteTrending
Subscribe to get the latest posts sent to your email.
