Assurance of Learning When Information Is Abundant

Article Icon Article
16 November 2026
Photo by iStock/Thapana Onphalai
In the AI era, students need more than foundational knowledge. They must learn to identify problems, select the right tools, and judge possible solutions.
  • Traditionally, students first learn a method and then how to apply it to a problem. But because AI can suggest detailed answers, today’s students should first determine what decision they need to make and what information they need.
  • Students should be asked to identify and work through the components of a problem before the instructor introduces them to the formulas commonly used to solve it.
  • Assessment is more accurate when faculty collect evidence at every point where students demonstrate a new capability, rather than rely solely on a final product.

 
In today’s technology-assisted learning and work environments, it is easier than ever for students and employees to retrieve knowledge and execute procedures. But that easy access to knowledge means it’s essential for individuals to exercise judgment about what information matters. As a result, both students and employers are demanding business education that trains graduates to make defensible decisions, not simply retrieve and act on information.

This shift is especially relevant in quantitative business courses. Students still must have a grounding in foundational knowledge. But they also need to be able to use formulas, methods, and terminology in the service of solving problems. For instance, before they ask AI to suggest possible solutions to a problem, they first need enough understanding to judge whether those suggestions are appropriate, credible, and useful.

However, the wide access to abundant information and tool-supported execution means that instructors can no longer assume that students are doing the critical thinking on their own. Therefore, schools must take steps to ensure that judgment work is visible during both learning and assessment.

A Shift in Teaching and Assessment

Traditionally, quantitative courses have emphasized method-first instruction and finished-product assessment. That approach made sense when it was reasonable to assume that, after repeated exposure to foundational knowledge, students would be able to apply it to final projects that instructors could assess. But that assumption no longer holds.

Method-first instruction assumes that a problem has already been framed. It also can signal that the most important part of a problem is determining the concrete steps needed to execute it. But today, students can use technology to both identify and execute the steps needed to solve a problem without understanding why those steps are required. They learn how to retrieve knowledge and execute a plan, but they don’t learn how to frame a problem and judge how to respond. Therefore, method-first instruction does not ask students to think early enough in the process.

Similarly, finished-product assessment looks too late at the evidence students have compiled. Even before generative AI, students could produce a polished submission by seeking outside help or imitating others’ work. Even so, instructors could generally treat a correct analysis or well-structured solution as a signal that the student had developed at least some of the intended capabilities. But when students can use generative AI, the quality of their final products tells instructors much less about what learners can independently recognize, decide, explain, or defend.

Method-first instruction does not ask students to think early enough in the process, and finished-product assessment looks too late at the evidence students have compiled.

The response is to move both teaching and assessment to earlier points in the process. Instruction should take a decision-first approach in which students have repeated opportunities to practice making decisions before they learn about methods. Similarly, assessment should move away from relying heavily on finished products and toward collecting evidence at the points where students are demonstrating capabilities.

In particular, schools should focus on helping students develop capabilities in five areas:

  • Decision identification. Can students clearly state the decision that they need to make?
  • Relevance filtering. Can they distinguish information that matters from information that is noise? Can they evaluate the credibility and usefulness of newly gathered information?
  • Identification of unknowns and next questions. Are they able to determine what else they must know before they move forward?
  • Tool selection. Are they prepared to select an appropriate method or tool and explain its role in the current situation?
  • Defense of process and conclusions. Can they explain their reasoning, evidence, assumptions, and limitations?

Decision-First Teaching

I frequently teach a business statistics course that requires students to work across multiple methods, interpret statistical outputs, and connect results to business decisions. I recently redesigned a multiweek hypothesis testing unit of this course around a decision-first approach to teaching and assessment.

In one typical assignment, students conduct a one-sample hypothesis test for a mean. But instead of beginning the lesson by introducing the test statistic, I start with a business question: A store’s recent data show that its average daily sales have been 75 USD above its target. Should we feel comfortable claiming that the store’s true average daily sales exceed the target?

This assignment requires students to decide what other information they might need and whether that new information strengthens or weakens the claim. For instance, what are the typical daily sales variations, and what was the sample size? A smaller standard deviation should make the 75 USD gap feel more meaningful; a larger sample size should make the pattern feel more stable.

Evidence accumulates at any point where judgment is visible: when students recognize decisions, identify relevant information, choose methods, interpret output, and defend conclusions.

In this approach, students must name the various components of the problem—gap, variability, and sample size—before they learn the actual formula. They begin to see what information matters and why.

I might vary the pieces of new information so that one factor strengthens the evidence while another weakens it. For instance, I might provide a problem that pairs lower variability with a smaller sample size. These less obvious situations push students to the limit of informal reasoning, where they require a formal way to balance competing factors.

Only then do I provide the test statistic, which quantifies the trade-offs students have already been reasoning through and helps them defend conclusions. The formula becomes a means to support judgment, not the main event.

Collecting Evidence Across the Process

During the same hypothesis testing unit, I can use a connected assessment sequence to collect evidence of student capabilities. The unit includes several related tests for means, proportions, independent samples, and paired samples. Rather than naming the test first, I make students work from the business situation toward the method.

As they complete the sequence, students first recognize what kind of decision they must make and what evidence would support it. Next, they informally judge the strength of their evidence; then they select and execute formal tests. Finally, they defend their process and conclusions under controlled conditions. As the table below shows, several assignments require students to use RStudio, coding software based on the R programming language.

Assessment Stage Student Task Capabilities Assessed Assessment Conditions Assurance Value
Recurring quizzes Students must answer mixed-scenario questions about hypotheses, test type, or decision structure. Decision identification; early method recognition. Working in class on paper, students complete individual quizzes that combine plausible distractors and some irrelevant information. Quizzes are graded. Results show whether students can recognize the structure of a problem as the set of possible methods grows.
Assignments to rank the strength of evidence Using summary information only, students must group eight business pilots by whether they show strong, medium, or weak evidence of potential. Relevance filtering; determining unknowns and next questions. Teams submit responses on paper in class; papers are graded. Answers demonstrate whether students can judge which information matters prior to formal testing and recognize when summary information is not sufficient for final judgment.
Practice with RStudio execution Assignments include both in-class work and homework. Tool selection; supported execution. In-class practice is ungraded and supported. Homework, which can be completed with use of AI, is graded. Results show whether students can connect the question, data structure, method, output, and conclusion while using AI support responsibly.
Midterm Students complete multiple long-form problems using written scenarios and provided RStudio output. Defense of process and conclusion; independent tool selection. Students take midterms in class on paper. Tests show whether students can independently connect the question, method, evidence, and conclusion.

Not every activity provides the same evidence for assurance of learning, and that is purposeful. The point is not simply to scaffold the learning within the unit, but to be intentional about which forms of judgment are visible within the assessments. Students need supported practice, discussion, individual checks, and controlled assessment. I use each assessment in the sequence for a different purpose. My goal is to obtain different kinds of evidence at different points in the learning process.

For example, the task for ranking the strength of evidence is currently team-based. I consider it a useful learning activity because students have to talk through which features matter, such as observed gaps, sample sizes, and variations. Through these sessions, students strengthen their judgment. They also discover that while their informal ranking can narrow the field, it cannot replace formal tests. While I am experimenting with how to grade individual contributions, I do not treat this exercise as strong evidence of individual performance.

I make a similar distinction with the assignments that involve RStudio execution, in which students read unlabeled problems, translate claims into hypotheses, choose appropriate tests, run analyses, interpret outputs, and draw conclusions. I give students permission to use AI for coding support as long as they disclose its usage. I want students to practice using AI for statistical reasoning and data analysis when such support is appropriate. However, because they complete homework outside of a controlled setting, I give the homework modest weight.

The controlled evidence comes on the midterm, when students work through long-form problems on paper using written scenarios. For these problems, I provide RStudio output because coding is not the primary learning outcome I am assessing, and software mechanics should not obscure the reasoning task. In some problems, I give the relevant output to keep the focus on interpretation and conclusions. In others, I require students to decide which output applies before they interpret it.

Program-Level Assurance

Across the sequence of assessments, evidence does not accumulate only at endpoints. It accumulates at any point where judgment is visible: when students recognize decisions, identify relevant information, choose methods, interpret output, and defend conclusions.

The same logic should apply beyond a single course. Assurance of learning should not depend on one polished artifact or one final performance point. Completed analyses and capstone projects are valuable, but they should only be components of a broader evidence system.

This evidence is more important than ever in learning and work environments shaped by abundant information and AI-supported production. Assurance should show not only what students can produce, but what they can recognize, question, select, and defend.

Authors
Anna Gloria Devlin
Clinical Professor of Decision Sciences and MIS, LeBow College of Business, Drexel University
The views expressed by contributors to AACSB Insights do not represent an official position of AACSB, unless clearly stated.
Sign up for AACSB's LINK email newsletter.
Our subscribers receive leading insights, news, and knowledge in business education.