As educators who love technology, we can barely contain our enthusiasm for the potential applications of artificial intelligence (AI). But AI requires massive amounts of data, so before jumping on the AI bandwagon we need to:
- reflect on the kinds of data that would make teaching more effective and improve learning outcomes;
- consider the systems that will allow us to collect and manage the data; and
- create processes to share and analyze the data.
Most districts do not yet have the foundation to make the leap to AI (other than what is already embedded in the apps and programs they’re currently using). Schools still exhibit a lack of maturity around data collection that should make us cautious about AI. There are also algorithmic bias and equity issues that need to be resolved before we move to wide-scale AI adoption. For most districts, spending money on AI over the next three to five years would be money down the drain. The ecosystems to support AI implementation are simply not yet in place in most schools and districts.
5 essential questions to test your district’s AI readiness
Before moving to AI, districts need to systematically build their digital landscape to get the full benefit of the technology they’re already using. Let’s begin with some basic questions:
- What kinds of data can help us make decisions to improve learning outcomes?
- Which programs can help us collect valid data and manage it safely?
- Is “adaptive” content always beneficial or is it sometimes more important to let teachers or students decide what comes next?
- What kind of feedback is the most valuable for student growth?
- When is an intervention a positive action and when does it eliminate constructive struggle, which is at the heart of deeper learning?
(Next page: How to evaluate your district’s data readiness for AI)
Many educators are uncomfortable with the ways we quantify results and argue that we are not really measuring learning. Other educators believe students would achieve deeper learning if we focused on providing feedback instead of scores. We need to articulate these and other questions before building out a digital ecosystem.
Why is AI so attractive?
The argument for building AI capability is tempting, and possibly a good solution for problems such as closing performance gaps. If Amazon can “learn” who we are and make suggestions based on our interests and preferences, then wouldn’t AI help us deliver optimized learning environments? Not necessarily.
Most of us believe that well-implemented technology improves learning. The goal is not to have AI move to the center of learning but to let it do what it does best: “classify and predict” and help make decisions. In this scenario, AI would be one of the tools in the box, and teachers, students, and parents would be in charge. Let the machines recommend and the humans make the decisions.
How to make decisions about your districts’ data
We need to have robust conversations about how technology can support student-centered learning and student-teacher collaboration. It’s imperative for curriculum specialists to have honest conversations with IT before investing money in significant projects. These conversations will help districts effectively strategize their instructional technology investments so they purchase products for the right reasons.
AI requires massive amounts of data. Schools and districts have been collecting student data for some time now. But we should develop our data-literacy capacity before collecting any more data. What data do we want to collect and for what purposes? Where are we today, and what are the next steps?
- Get a handle on your apps. K-12 education has too many apps in use, many for the same course. You don’t want to send data to 15 different apps for statistics, for example. As you examine what you have and what data is being captured, it is a good opportunity to streamline the number of apps being used.
- Evaluate your teaching and learning platform. Does the platform provide student usage reports? Other activity metrics? Where is that data being captured and shared? Teachers need help to become data literate so they can use data to improve learning.
- Is your data usable? Much of the data collected is unusable as it can’t be integrated with other data systems. The data needs to get to one place where it can be mixed, matched, and cross referenced. We need dependable data transfer and data interoperability before we move on to more sophisticated AI.
- Can you obtain everything you need? There is data about individual characteristics, preferences, and choices that we can’t yet gather while keeping it private and safe. This kind of data should be based on objective recording of behavior, collected by machine sensors, and not by well-intentioned educators.
If we want AI to effectively predict outcomes and prescribe learning resources, we need an enormous bank of top-quality digital resources, a learning-object repository in which resources are properly meta-tagged against topics, standards, object types, format, reading levels, author, and language preferences. If your district does not have that, why purchase an AI engine?
Our focus today should be on building a stable digital ecosystem while we develop our data-literacy skills and knowledge. There are AI functions within some existing K-12 programs, including adaptive technologies, learning management systems, digital libraries, and data warehouses with role-based data dashboards. We are already using AI tools such as speech recognition, text-to-speech, flexible playlists, learning preferences data, sensors, human-machine dialogue, and facial recognition. But the kind of big data and the interoperability required to have AI at the base of a district’s instruction and learning systems is just not in place today.