Beyond the Snapshot: How Might We Make “Assessment in the Service of Learning” Possible for All Learners, in the AI Era

For too long, assessment has functioned as an educational autopsy. Like an examination performed after the patient has expired, traditional testing provides an accurate post-mortem “verdict” that arrives too late to help the learner. It offers a static point-in-time picture—a snapshot of developed ability—but too often remains silent on the dynamic processes by which students actually learn and mature.

In the 1950s, I (Edmund W. Gordon) worked alongside Else Haeussermann. When others dismissed children with neurological impairments as ‘uneducable’, Haeussermann saw potential waiting to be unlocked. She did not use tests to sort these children; she used diagnostics and observations to find the specific conditions under which each child might succeed at a task. The lesson I took: assessment should be used not only to identify what is, but to imagine and cultivate what might become.

Today, as we stand on the precipice of the AI era, the technology just might have begun to catch up to Haeussermann’s vision. We increasingly possess the potential to move beyond the “autopsy” model and build engines of human potential.

From Snapshot to Dynamic Interaction

In 2013, the Gordon Commission on the Future of Assessment in Education challenged the artificial wall separating “testing” from “teaching”. Traditional standardized tests offer little direct utility in the classroom because they are designed to rank students rather than understanding their needs. We must shift from a “check engine” light flickering on (often months late) to a GPS dashboard designed to aid navigation. 

As I suggest in The Testing and Learning Revolution (2016), to cultivate a seedling you don’t measure the plant to judge it; you measure it to understand if it needs water, sunlight, or richer soil. Analogous approaches for learners include Dynamic Pedagogy—which entails integrated efforts in which assessment, curriculum, and instruction reinforce each other. 

Existence proofs show how this works. We see it in game-based systems where the “test” is invisible because it is intrinsic to the challenge, in learning platforms (such as Khanmigo) providing real-time feedback, and in portfolios, badges, and dynamic learning maps. We also see it in classroom and personalized assessment approaches and practical measurement for improvement. Ultimately, assessment should provoke the student to think differently and struggle productively. 

The Strength of Human Variation

Traditional testing misunderstands human variance. In its quest for comparability, current approaches too often treat variation of factors such as linguistic and cultural background, motivation, attention, memory, and processing as statistical noise to be minimized, rather than richness to be appreciated and leveraged. Understanding functional characteristics (including the processes of cognition) shifts our question from: “How smart is this learner?” to, “How is this learner smart?”.

The Pedagogical Troika

In the recently released Handbook for Assessment in the Service of Learning series (2025), we solidify this vision through the metaphor of the Troika—a three-legged stool consisting of Assessment, Teaching, and Learning. If you remove assessment—the feedback and insight leg—the entire structure suffers.

What does it mean to be an educated person in the AI era? Bereiter and Scardamalia recommend preparing learners to engage in lifelong learning, enabling them to gain new competencies while adapting to the accelerating pace of change. They recommend developing methods for assessing knowledge creation, working with abstractions, complex systems thinking, cognitive persistence, and collaborative responsibility. Assessment design must ensure that inferences serve the learner, not just the system, creating a learning environment that nurtures rather than sorts.

Why Now?

While cost once prevented personalized prescriptions (including learning plans) for every learner, AI-powered solutions have the potential to enable what I call ‘Pedagogical Analytics’ at scale, freeing teachers to focus on human connection and mentorship. We can use AI to power components of a GPS for learning—a system that offers learners step-by-step guidance. If we are bold, students will receive a useful map, not a definitive verdict. The technology increasingly within reach makes the practical imperative clear.

Suggested Reading: 

• Armour-Thomas, E., McCallister, C., Boykin, A. W., & Gordon, E. W. (Eds.). (2019). Human variance and assessment for learning. Third World Press. 
• Armour-Thomas, E., & Gordon, E. W. (2025). Principles of dynamic pedagogy: An integrative model of curriculum instruction and assessment for prospective and in-service teachers. Routledge.
• Gordon, E. W. (2020). Toward assessment in the service of learning. Educational Measurement: Issues and Practice, 39(3), 72–78. 
• Gordon, E. W. (2025). Series introduction: Toward assessment in the service of learning. In E. M. Tucker, E. Armour-Thomas, & E. W. Gordon (Eds.), Handbook for Assessment in the Service of Learning, Volume I: Foundations for Assessment in the Service of Learning. University of Massachusetts Amherst Libraries.
• Gordon, E. W., & Rajagopalan, K. (2016). The testing and learning revolution: The future of assessment in education. Palgrave Macmillan US. https://doi.org/10.1057/9781137519962
• The Gordon Commission on the Future of Assessment in Education. (2013). To assess, to teach, to learn: A vision for the future of assessment[Technical Report]. https://www.ets.org/Media/Research/pdf/gordon_ commission_technical_report.pdf

This blog series on Advancing AI, Measurement and Assessment System Innovation is curated by The Study Group, a non-profit organization. The Study Group exists to advance the best of artificial intelligence, assessment, and data practice, technology, and policy and uncover future design needs and opportunities for educational and workforce systems.