But Will It Work?
- Jun 8
- 2 min read
A natural question follows any proposal to redesign education: how would we know if it is successful?
This is not a trivial question, and it deserves a serious answer. The challenge is that the outcomes we care about—adaptability, model-building, collaboration, and the ability to learn in unfamiliar situations—are not fully captured by the measures most commonly used today. Standardized tests and isolated performance tasks can tell us something, but they do not tell us everything.
If the goal is to produce adaptive learners, then the evidence we look for must align with that goal. We should think like a gardener shaping the garden to give her plants the best chance possible to bloom.
Expanding What Counts as Evidence
Rather than relying on a single metric, evaluation should draw from multiple layers of evidence.
1. Conditions Within the System
First, we ask whether the system is actually creating the conditions required for adaptive learning:
Are students given protected time for deep engagement?
Do they experience repeated cycles of feedback and revision?
Is learning social, with visible collaboration and shared problem solving?
Are students required to make their thinking explicit and revise it over time?
If these conditions are not present, the system cannot reasonably be expected to produce adaptive learners.
2. Student Behavior and Learning Patterns
Next, we look at how students behave within the system:
Can students identify gaps in their own understanding without prompting?
Do they know how to find and evaluate information to address those gaps?
Do they improve their performance over repeated cycles of work?
How do they approach unfamiliar problems?
These behaviors provide evidence that students are developing the underlying capacities associated with adaptability.
3. Performance in Novel Situations
Finally, we examine how students perform when they encounter tasks they have not been explicitly trained for.
Can they organize themselves and make progress without a clear template?
Do they collaborate effectively with others?
Can they build and refine models in real time?
If we care about how learners perform in unfamiliar situations, then we must measure them in unfamiliar situations.
A Different Kind of Measurement
This approach does not produce a single number that fully captures success. Instead, it builds a body of evidence over time. This is a tradeoff. Simpler systems are easier to measure, but they often measure what is easiest to quantify rather than what matters most.
More complex goals require more nuanced forms of evaluation. Here is an idea for one possible path to deeper evaluation.
An Honest Position
At present, there is no single, scalable metric that captures adaptive learning perfectly. That does not mean it cannot be evaluated. It means that evaluation must be approached differently.
The goal is not to eliminate measurement, but to align it with what we are actually trying to develop. If we expand what counts as evidence, we can begin to see whether a system is producing learners who can make progress in situations where the answers are not already known.
