01

Begin with the claim

Before writing a question, define what successful performance would mean. Explaining a concept, solving a novel problem, writing working code and performing a supervised task are not interchangeable outcomes. An assessment should make the intended claim as specific as possible.

That specificity also protects the learner. Someone should not receive a sweeping label from a narrow task. A result can be useful without pretending to describe every aspect of a person’s ability.

02

Make assistance rules visible

A guided lesson can invite hints and examples. An independent examination may limit them. The interface should tell the learner which state they are in and what support is allowed. It should also preserve enough information for the result to be interpreted later.

Unexpected behavior may warrant review, but an automated signal is not by itself proof of misconduct. Any serious conclusion needs an appropriate process and a chance to correct an error.

03

Interpretation needs context

Accuracy, reasoning, application and consistency can matter differently across skills. Time may be relevant for a safety-critical response and much less relevant for a careful design task. The standard and any comparison group should be defined before results are used to rank people.

A benchmark without adequate evidence can look precise while being misleading. FLOOR9’s plan therefore treats broad rankings as a later possibility requiring a defensible method, not as an automatic output of the first assessments.

04

Connect the result to the next step

An assessment can guide further practice. It can also become one source of evidence in a broader record when its method and issuer are clear. Practical work may add another perspective.

The first FLOOR9 pilot should test whether this sequence helps students learn and gives educators useful, fair information. A credible assessment is not merely a score; it is an interpretable decision point in a larger learning journey.