Assessment of Learning, Assessment for Learning, and Assessment as Learning: What is the Difference?

Three terms appear in nearly every serious conversation about educational assessment, yet they are regularly used as interchangeable. Assessment of learning, assessment for learning, and assessment as learning describe fundamentally different relationships between evaluation and instruction. Treating them as the same thing produces assessment strategies that serve the wrong purpose at the wrong moment.

Assessment of Learning

Assessment of learning is the most familiar category. It measures what a student achieved after instruction ends: final exams, standardized tests, end-of-course grades. The primary audience is not the learner. The data flows to institutions, registrars, and accreditation bodies. Its purpose is certification, not correction.

This is not an argument against assessment of learning. Institutions need documented evidence of competence. The problem arises when it is the only category in use, which is the default in most university courses. A student who scores 58% on a final exam receives a grade, not a diagnostic. The information arrives too late to influence learning and at too coarse a resolution to identify what actually went wrong.

See how assessment of learning compares to formative alternatives for a fuller breakdown of the structural differences.

Assessment for Learning

Assessment for learning shifts the purpose from documentation to instruction. Evidence gathered during the learning process informs what the teacher does next, not what goes into a grade book. A 2021 scoping review in Studies in Educational Evaluation identifies two core requirements for AfL: the assessment must generate information that leads to improved performance, and it must engage the learner in actions to close the identified gap.

The evidence base here is substantial. Black and Wiliam's landmark 1998 review of 250 studies found that effective formative assessment produced achievement gains of 0.4 to 0.7 standard deviations, among the largest effects reported in educational research. Despite that, implementation in higher education has remained inconsistent, constrained by large cohorts, limited contact hours, and the practical difficulty of turning frequent assessment into instructional action before the course moves on.

Most AfL tools in use today, quizzes, polls, exit tickets, short responses, rely on retrieval. A student selects or recalls an answer. That answer may be correct without reflecting genuine understanding. Retrieval and comprehension are not the same measurement.

Assessment as Learning

Assessment as learning is the category least understood in practice. Lorna Earl introduced it formally in 2003 to describe a third function: assessment in which the student becomes the active evaluator of their own understanding. As Earl's foundational framework describes it, the student monitors what they actually know, identifies gaps in their own comprehension, and adjusts their learning accordingly. The metacognitive process is itself the assessment, not preparation for it.

Unlike AfL, which is primarily teacher-directed, assessment as learning places the diagnostic work inside the learner. The challenge is accuracy. Black and Wiliam note that student self-assessment becomes unreliable when learners lack a clear enough model of what genuine understanding looks like. Without that reference point, metacognitive monitoring tends to produce confident ignorance: a student who cannot identify what they do not understand cannot accurately self-assess around it.

Where All Three Fall Short

Assessment of learning arrives too late. Assessment for learning monitors progress without verifying depth. Assessment as learning depends on a metacognitive accuracy that most students are still developing. Each framework has its role, but none resolves the same core problem: none of them require a student to demonstrate understanding by producing an explanation.

The Feynman Technique sits outside these three categories, but it addresses the gap directly. If a student can teach a concept clearly, they understand it. If they cannot, the breakdown in their explanation reveals exactly where the gap is located. Explanation is not a proxy for understanding; it is evidence of it.

What assessment for learning requires in practice

Teaching as Measurement

Axiom Flow goes beyond the formative assessment platform model by making teaching the mechanism of measurement rather than a preparation for it. Atlas generates a configurable set of misconceptions from the learning material and maps each one to an exam question. Sam, Axiom Flow's AI student, is initialized with those misconceptions. The student's task is to teach Sam until all misconceptions are corrected. Sam updates its understanding only in response to the quality of the explanation it receives.

What Sam still misunderstands after the session is not a confidence score or a quiz result. It is a direct record of what the student could not articulate. Atlas then evaluates Sam's performance on a constrained exam, producing a conceptual mastery assessment grounded entirely in what the student was able to teach.

This is a direct operationalization of assessment for learning and assessment as learning together. The student is building understanding through explanation at the same moment they are being assessed. The three-category framework describes what assessment can do. Teaching-based assessment describes what it needs to do: create the conditions in which understanding either demonstrates itself or exposes its own limits.