91传媒

By AJ Gutierrez, Alaina Boyle Ph.D., and Elizabeth Sykes

For years, 91传媒 has been trying to answer a fundamental question. What would schools do differently if they could see students more fully?

Mathematica found that when schools broaden the information they use to understand student potential, they identify substantially more students who may be ready for advanced coursework. Its second study found strong evidence that 91传媒 partnership increased AP enrollment.

Our work with the Center for Trust and Transformation showed something equally important. Relationships matter, but having a trusted adult does not automatically mean that relationship gets activated to connect a student to an opportunity.

Together, those findings have pushed us toward the next stage of our research journey. The question is no longer simply whether we can identify students who may be overlooked. Increasingly, we are asking something harder. Can we turn student insight into evidence that helps an adult know what to do next? This is the question that we are now exploring with the

Looking for stronger signals

The preliminary analysis also reinforces something that has been emerging across our research for several years. Student experience and relationships appear to provide useful information beyond academic performance alone. A transcript can tell us whether a student has demonstrated academic readiness. It may tell us much less about whether that student believes AP is for them, understands how to enroll, sees themselves belonging in the classroom, or has someone encouraging them to take the next step. Two students can have nearly identical academic profiles and still need completely different things from their school. One may simply need information. Another may need encouragement. Another may be concerned about workload. Another may never have considered the opportunity at all.

If we can understand those differences, we can begin moving from identifying students to understanding how to engage them.

Prediction is not the destination

There is enormous interest in predictive analytics and AI in education right now. Much of the conversation focuses on whether we can build better models to predict which students will enroll in advanced coursework, become chronically absent, need tutoring, or experience some other outcome. Those can be useful questions. But a prediction does not change a student’s trajectory. A model could identify exactly the right student and still have no impact if no one knows what to do with that information. That is why the next stage of our Learning Collider partnership is deliberately focused on something larger than prediction. We want to study how evidence gets used.

The proposed research will compare the existing 91传媒 student insight process with a more structured approach that uses an AI-assisted synthesis of academic information, student aspirations, assets, barriers, and trusted relationships to suggest student specific outreach or support. Just as importantly, the process assigns responsibility for follow up and asks staff to document whether the action actually occurred. The adult still makes the decision. They can adopt the recommendation, adapt it based on what they know about the student, or decline it altogether. In fact, disagreement with a recommendation may sometimes represent better judgment because the educator has context that the model does not. The goal is not to replace professional judgment. It is to understand how better evidence and professional judgment can work together.

What happens between insight and outcome

This is what makes the next stage of the research particularly exciting to us. We want to understand what happens between insight and outcome. Does the educator understand the evidence? Does it change how they think about the student? Does someone take responsibility for reaching out? Does the conversation actually happen? Is the person having that conversation someone the student trusts? Does the outreach address the barrier that was standing in the way? And ultimately, does the student enroll and succeed? The proposed study will examine those steps while also exploring why educators adopt, adapt, or decline suggested actions and what organizational conditions make effective follow through more likely. 

The research is expected to move through validation and collaborative piloting before a larger cluster randomized evaluation across 91传媒 partner schools. The proposed trial would compare the existing 91传媒 process with the structured evidence use approach and examine outcomes including evidence comprehension, completed outreach, AP enrollment, persistence, and performance. That distinction matters because the intervention we ultimately want to understand is not an AI summary. It is a better process for helping people use evidence.

Making better use of what schools already have

There is a bigger idea underneath all of this. Schools already invest heavily in programs and people. AP exists. Dual enrollment exists. Career pathways exist. Teachers, counselors, coaches, advisors, and school leaders are already building relationships with students every day. We are not talking about redesigning schools from the ground up. The opportunity may be to make better use of what is already there.

Imagine knowing that a student is academically ready for dual enrollment, that they have expressed an interest in health care, that they are worried about the cost of college, and that they identify a particular teacher as someone they trust. Then imagine giving that teacher the information and guidance needed to have a thoughtful conversation with that student. That is fundamentally different from putting another dashboard in front of an educator. It is about helping the right adult understand the right student and take the right action at the right time.

While advanced coursework gives us a powerful place to test this idea, the implications could extend much further. The same evidence use process could eventually help schools think differently about chronic absenteeism, tutoring, career pathways, and other consequential decisions about students. We still have a lot to learn. The predictive analysis is preliminary, and a strong model does not tell us whether adults will use the evidence effectively or whether students will benefit. That is precisely why we are doing the next study.

The research journey started with a question about who schools were overlooking. It is increasingly becoming a question about what we can do once we see them. Identifying the student is important. But the real opportunity begins when someone knows how to act.

There is much for us to learn. Here is the preliminary report from the Learning Collider.