These principles guide the framework. The Five Steps offer a flexible way to put them into practice; cases show the reasoning and its limits.
- 01
Fluent performance is not sufficient evidence of learning.
Artificial intelligence can produce accomplished intellectual products. Fluency may demonstrate performance, but it does not by itself establish whose judgment produced the work or what human capability changed.
- 02
Learning is a change in human capability.
Education is not complete when an acceptable product appears. Learning requires growth in what a person can understand, judge, explain, revise, or do with increasing independence.
- 03
The submitted text is evidence, not proof.
A paper, solution, presentation, or project is one source of evidence. Green Pen Framework does not assume the reliability of the finished text; it asks for choices, revisions, and explanations that can be examined alongside the conditions of assistance. These observations inform judgment without guaranteeing independent learning.
- 04
Assessment connects visible action with careful interpretation.
Assessment should reveal meaningful decisions as work develops. A repair, an explanation, and a new application support different inferences. Faculty must consider what each observation establishes, what assistance was available, and what remains uncertain.
- 05
Faculty require evaluative sovereignty.
Faculty need recognized authority to decide what matters in a particular case, explain their reasons, and reconsider in light of evidence. Rubrics, metrics, templates, policies, and algorithms may organize evaluation; they cannot eliminate the need for accountable human discretion.
- 06
Student agency requires consequential intellectual action.
Students need room to discover a question or conclusion that the assignment did not anticipate. Rubrics should make standards for evidence and reasoning clear without prescribing the intellectual destination. Faculty must be able to recognize a promising departure, examine its grounds, and challenge criteria that penalize it.
- 07
AI capability does not confer educational authority.
A system may be capable of generating feedback, assigning a score, or recommending a decision. Capability alone cannot determine whether that delegation is educationally legitimate, who authorized it, or who remains answerable for the result.
- 08
Judgment does not scale.
Institutions can focus judgment, sustain it, and learn from its signals. They cannot reproduce situated human judgment simply by accelerating production, standardizing its appearance, or automating its language.
- 09
Private faculty labor cannot indefinitely repair public design failures.
When the same problem recurs, the evidence may point beyond individual teaching to assignment architecture, curricular sequence, assessment design, or policy. Sustainable practice requires institutions to address failures at the level where they are produced.
- 10
Human agency must be designed into educational systems.
Institutions preserve human agency by assigning meaningful authority, consequential action, and visible responsibility to faculty and students. Agency must be built into practice, curriculum, assessment, and policy—not merely affirmed in institutional language.