Emerging Standards for Data Products and AI
New industry analysis highlights the emergence of architectural standards for data products and AI interactions. For adaptive edtech, this points to a growing need to standardize how learner data is structured and used across systems to ensure interoperability and effective personalization.
- Bayesian Knowledge Tracing (BKT) and Deep Knowledge Tracing (DKT) are two prominent models for tracking a student's knowledge state in real-time. BKT, a long-standing method, models skill mastery as a latent variable, while the more recent DKT uses recurrent neural networks to predict future performance based on a sequence of student actions. - Reinforcement learning (RL) is used in adaptive learning to create personalized experiences by adjusting content and pace based on student performance. Intelligent Tutoring Systems (ITS) leverage RL to simulate one-on-one tutoring, adapting to individual learning styles and providing customized instruction. - Multi-armed bandit (MAB) algorithms, a form of reinforcement learning, are applied to content recommendation in edtech to balance exploring new material with exploiting known successful content. This approach helps address the "cold start" problem for new users by efficiently learning their preferences. - Automatic Speech Recognition (ASR) for young children presents unique challenges due to variations in pronunciation and speech patterns. Datasets like the UCLA JIBO Kids' Database, containing recordings of children aged 4-7, are crucial for training accurate ASR models for early literacy applications. On-device ASR is a key technology for privacy-compliant and responsive reading tutors as it keeps voice data local. - Designing AI for children requires a focus on safety, with guidelines recommending the use of platforms specifically built for kids that include parental controls, content filtering, and compliance with regulations like the Children's Online Privacy Protection Act (COPPA). Experts advocate for "non-negotiable" safety standards, such as preventing the generation of fake images of children and avoiding the creation of emotional dependency. - The IEEE has established standards for Adaptive Instructional Systems (AIS), such as IEEE 2247.4, which provides recommended practices for the ethically aligned design of AI in these systems. These standards aim to ensure transparency and interoperability in how AI is used in educational technology.
