From Predictions to Decisions

A system might predict whether an email is spam, recommend which video you might want to watch next, or generate a response based on what is likely to come next. But what happens after an AI system produces an output?

Sometimes, that output simply gives us information to consider. Other times, it can influence what we see, what opportunities are presented to us, or what action happens next. In this topic, we'll follow AI outputs one step further and explore how predictions and recommendations can influence what happens next, including the human choices that shape how these systems are designed and used.

An AI system can make a prediction, assign a score, classify something into a category, or recommend an option. But these outputs don't necessarily determine what should happen next.

Imagine an AI-supported learning tool identifies patterns in a student's work and predicts that they are likely to experience difficulty with an upcoming concept. That prediction may provide useful information, but it doesn't explain why the student might struggle or determine what should be done in response.

How an AI output influences what happens next depends on how the system is designed and used. A prediction might simply be presented to an educator as additional information, used by the system to generate a recommendation, or trigger an action automatically.

From Prediction to Action

The same prediction can play very different roles depending on what happens after it is made. The example below follows one prediction through three possible pathways, from informing an educator to triggering an automated action.

The prediction is the same in each pathway, but what happens because of it is different. In the first pathway, the prediction provides information for the educator to consider. In the second, the system goes a step further by using the prediction to recommend an action, while the educator still decides what to do. In the third, the system acts on the prediction automatically, without educator review beforehand. The significance of that automation also depends on what action is being taken. Automatically assigning an additional practice activity is very different from using an AI prediction to determine a grade, placement, or another high-impact outcome. As the potential impact on a learner increases, meaningful human judgment becomes increasingly important.

When algorithms are used to inform or shape what happens next, we enter the world of algorithmic decision-making.

Algorithmic decision-making broadly refers to processes where algorithms are used to inform, support, or automate decisions. These processes are already part of many digital experiences. Search engines rank which results appear first, recommendation systems influence what appears in content feeds, and educational platforms may use predictions to recommend activities, personalize content, or provide feedback.

Sometimes these decisions may seem small, such as which video appears next. Others can have a greater influence on someone's experience or opportunities. In each case, it is worth looking at not only what the system predicts, but how that prediction shapes what someone sees, experiences, or does next.

Let's look at a few examples educators and students may encounter.

AI systems can make predictions and recommendations, but people make many of the choices that shape how those systems work. Building an AI system isn't always just the work of AI developers or data scientists.

When AI is developed for a particular field, domain experts can also contribute. A domain expert is someone with specialized knowledge or experience in the area where the AI system will be used. They can bring context about the people, practices, and situations that technical teams may not have. Research on machine learning development has found domain experts contributing to areas such as collecting and validating data, defining concepts, and evaluating system performance (Díaz & Smith, 2023).

In education, educators are the domain experts. Their knowledge of students, learning, and classroom practice can help shape decisions about how AI-supported tools are designed and how they should work in educational settings. The following study shows what involving educators in that process can look like.

This kind of collaboration brings different forms of expertise together. Technical teams understand how to build the system, while educators can bring context about teaching and learning that may not be visible from the data alone.

A recommendation, prediction, or generated response can feel immediate when it appears on a screen. But we now know there is much more behind that output. AI systems learn from data, identify patterns, make predictions, and use those predictions in ways shaped by how the system has been designed.

For educators, understanding this process offers a useful shift in perspective. Instead of asking only, “Is this AI output correct?”, we can begin asking richer questions. What information might have shaped this result? What is the system trying to predict or accomplish? What might it be missing? And how much influence should its output have over what happens next? These questions don't require educators to become AI developers or understand every technical detail. They are part of building the AI literacy needed to approach these systems with curiosity, context, and informed judgment.

Understanding how AI works also gives us a foundation for thinking more critically about its impact. When we understand how data, patterns, predictions, and human choices shape AI systems, we are better prepared to consider the effects these systems can have when they are used in the real world.

Now that you’ve looked more closely at how AI learns, predicts, and influences what happens next, it’s time to put those ideas together. Use this knowledge check to see how well you can connect the concepts from across the chapter.

What’s Next in Your Learning Journey?

In Chapter 3, we’ll build on what we’ve learned about data, patterns, predictions, and decisions by looking more closely at the risks and responsibilities that come with AI use. We’ll explore issues such as bias, inaccurate and manipulated content, privacy, academic integrity, and responsible decision-making, along with the safeguards that can help educators navigate these challenges thoughtfully.