I keep hearing about AI. Hundreds of articles explain that we are on the verge of something big. Soon, AI will take care of mundane tasks and help your students grow.

But there’s a catch. You can’t build AI from scratch, you need to train it with data. Data about your students: their grades, demographics, and so on.

Learn about the K-12 industry standards in data collection, and get one step ahead of your competitors.

We’re discussing predictive AI, not generative AI.

There are 2 main types of AI in education:

Generative AI doesn’t need student data. This is the most hyped AI today. It’s ChatGPT that creates text, or Midjourney that generates images. Hundreds of millions of people use these models daily, for fun or work.

To generate content, you don’t need to know much about students. Students can open ChatGPT and start asking questions. ChatGPT doesn’t care about student grades or attendance.

Predictive AI gets trained with specific data. These models analyze student data and build predictions.

Predictive models are not consumer-grade, they are less universal. Instead such models are specific to their task. As a result, we don’t hear about predictive AI as much.

In education, predictive AI can be very effective. With it, you can optimize operations, locate at-risk students, or work out the school budget.

You can learn more about AI in K-12, and how it all works but here’s the point. Predictive models are very efficient, but they need specific data. Without good data, AI will have nothing to work with. Let’s see what kinds of data we can have.

Collect basic K-12 data with OneRoster

All the basic data is stored in the School Information System (SIS). Here are some examples of it:

Enrollments

Grades

Attendance

Rosters

Schedules

Demographic data

Parent data

Teacher and staff records

And much more.

You’ll need OneRoster to transfer that data from the SIS to your AI, and vice versa.

Here’s what OneRoster provides:

Standard for organizing school data. By default, each application may store data differently. Your SIS would keep student data in CSV file format, while the other app uses JSON. Or store grades in letters (A, B, C), while the other keeps numerical grades (10, 50, 90). The list goes on.

Now, how can you optimize that data transfer? When you’re getting an AI, you have two options:

Hire developers to manually integrate software. It can take a year to fully integrate two applications.

Get OneRoster applications. Most modern SIS support OneRoster. So integrating data through OneRoster will be much faster.

Real-time data updates. OneRoster supports data sharing through REST API. This way, your AI can analyze up-to-date information. If you don’t need real-time updates, OneRoster also supports CSV file sharing.

Security standards. OneRoster ensures data transfer via OAuth 2. With OAuth 2, the SIS can grant access to your AI without sharing passwords. Think of logging into Spotify with your Google account.

Most school districts prefer software with OneRoster. If you want to attract more potential customers, ensure OneRoster compliance.

Even if you’re building a custom AI application for inner use, OneRoster is still a great option. Most modern SIS support OneRoster, so it’s very efficient for integrating data.

Track student actions with xAPI

xAPI (Tin Can) is a protocol for tracking learning experiences. If you want to analyze learner actions, you’ll need xAPI.

With it, you can track any user actions:

When do students pause lectures?

Which quiz questions get the lowest scores?

What decisions do users make in learning games?

Data like this is much harder to analyze. But AI can find correlations where the human mind is helpless. If you leverage xAPI, you’ll get very powerful AI models with very implicit data. These models can predict SAT scores based on how often a student is late.

Teachers can have a better picture of student engagement, even during online classes. They can’t see if a student is scrol