AI to Learn 2.0: A Deliverable-Oriented Governance Framework and Maturity Rubric for Opaque AI in Learning-Intensive Domains

Seine A. Shintani

cs.AIMar 16, 2026

AI to Learn 2.0: A Deliverable-Oriented Governance Framework and Maturity Rubric for Opaque AI in Learning-Intensive Domains

Authors: Seine A. Shintani

Organizations: Department of Biomedical Sciences, College of Life and Health Sciences, Chubu University, Kasugai, Aichi 487-8501, Japan · Center for Mathematical Science and Artificial Intelligence, Chubu University, Kasugai, Aichi 487-8501, Japan · Institute for Advanced Research, Nagoya University, Nagoya, Aichi 464-8601, Japan

Abstract

Generative AI is entering research, education, and professional work faster than current governance frameworks can specify how AI-assisted outputs should be judged in learning-intensive settings. The central problem is proxy failure: a polished artifact can be useful while no longer serving as credible evidence of the human understanding, judgment, or transfer ability that the work is supposed to cultivate or certify. This paper proposes AI to Learn 2.0, a deliverable-oriented governance framework for AI-assisted work. Rather than claiming element-wise novelty, it reorganizes adjacent ideas around the final deliverable package, distinguishes artifact residual from capability residual, and operationalizes the result through a five-part package, a seven-dimension maturity rubric, gate thresholds on critical dimensions, and a companion capability-evidence ladder. AI to Learn 2.0 allows opaque AI during exploration, drafting, hypothesis generation, and workflow design, but requires that the released deliverable be usable, auditable, transferable, and justifiable without the original large language model or cloud API. In learning-intensive contexts, it additionally requires context-appropriate human-attributable evidence of explanation or transfer. Worked scoring across contrastive cases, including coursework substitution, a symbolic-regression governance contrast, teacher-audited national-exam practice forms, and a self-hosted lecture-to-quiz pipeline with deterministic quality control, shows how the framework separates polished substitution workflows from bounded, auditable, and handoff-ready AI-assisted workflows. AI to Learn 2.0 is proposed as a governance instrument for structured third-party review where capability preservation, accountability, and validity boundaries matter.

Figures & tables

Approach	Unit of analysis	Residual focus after AI use	Black-box-free final deliverable required?	Explicit validity-domain / failure rule required?	Designed for structured review?	Capability preservation explicit?
Post-hoc XAI	Model prediction / deployed model	An explanation is attached to a black-box prediction or model; the black box remains central.	No	Usually no	Limited	No
Interpretable ML	Model	An inherently interpretable model remains in deployment.	Sometimes, at the model level	Not inherently	Limited	No
Human-in-the-loop / Interactive ML	Learning or decision process	Human intervention points remain in the process or lifecycle.	No	Not inherently	Usually no narrow compliance instrument	Implicit, not explicit
Documentation / model reporting	Model / dataset	Documentation remains with the model or dataset, including intended use and limitations.	No	Partial	Partial documentation audit	No
Responsible AI / AI ethics	Organization / policy / principle set	High-level principles, commitments, and impact considerations remain.	No	Usually high-level	Usually no deliverable-side rubric	Sometimes indirect
Official governance frameworks	Organization / system lifecycle / professional capability	Risk-management processes, governance artifacts, and competencies remain.	No	Encouraged, but not as a deliverable-sufficiency test	Partial	Partial

Table 1: Conceptual crosswalk between AI to Learn 2.0 and adjacent approaches. The rows correspond to the literatures discussed in this section: post-hoc explanation [ 20 , 21 ] , interpretable machine learning [ 22 ] , human-in-the-loop and interactive machine learning [ 25 ] , documentation/model reporting [ 23 , 24 ] , broad governance and policy frameworks [ 4 , 5 , 2 , 3 , 7 ] , AI-aware assessment design and learning assurance [ 12 , 13 , 14 , 15 , 17 , 18 , 19 ] , and sustainability-oriented work [ 26 ] .

Symbol	Element	Minimum contents
$A$	Distilled artifact	The black-box-free artifact intended for routine use: for example, an audited codebase, explicit equation, rule set, checklist, lesson material, form system, or documented procedure.
$P$	Provenance / audit record	Which AI tools were used, for what purpose, what classes of data were exposed, what humans accepted or rejected, and what verification was performed.
$V$	Validity-domain statement	Intended use, assumptions, regime of applicability, out-of-scope cases, and known limitations.
$F$	Failure / escalation rule	Known failure modes, warning signs, uncertainty triggers, withdrawal criteria, and when to escalate to human review.
$R$	Operational resource note	Runtime dependencies, cloud or local execution assumptions, access requirements, cost or energy implications, and low-resource fallback if relevant.

Table 2: Core elements of the AI to Learn deliverable package.

Symbol	Dimension	Main linkage to the formal definition	Gate?	Guiding question
$m_{1}$	Residual opacity in the final deliverable	Artifact residual $ArtRes$	Yes	Does routine use still depend on the original opaque AI component?
$m_{2}$	Human sovereignty and release authority	Human sovereignty $Sov$	Yes	Who has the authority and responsibility to accept, reject, revise, and release the deliverable?
$m_{3}$	Auditability and stabilization trace	Provenance / audit record $P$ , package completeness	No	Can a competent third party reconstruct how the artifact was stabilized and checked?
$m_{4}$	Validity-domain and failure awareness	Validity-domain statement $V$ , failure/escalation rule $F$	Yes	Does the package say where the artifact applies, where it fails, and when to escalate?
$m_{5}$	Information protection	Information protection $Prot$	Yes	What information reached opaque AI, and were appropriate protections used?
$m_{6}$	Energy and resource proportionality	Operational resource note $R$	No	Did opaque AI remain as a heavyweight runtime dependency, or was it distilled away?

Table 3: Overview of the AI to Learn 2.0 maturity rubric.

Case	$m_{1}$	$m_{2}$	$m_{3}$	$m_{4}$	$m_{5}$	$m_{6}$	$m_{7}$	$S$	Gate	$C$	$\tau$	Core	AI to Learn 2.0
C1. Chatbot-authored take-home report submitted as-is	1	1	0	0	1	2	1	6	0	0	3	0	0
C2. LLM-assisted report plus structured oral defense	3	3	2	2	2	3	2	17	0	3	3	0	0
C3. Symbolic regression package without validity-domain / failure analysis	4	3	3	1	4	4	3	22	0	–	–	0	0
C4. Symbolic regression package with validity-domain statement and failure-oriented tests	4	4	4	4	4	4	3	27	1	–	–	1	1
C5. Teacher-audited national-exam practice forms (tool only)	4	4	3	4	4	4	4	27	1	–	–	1	1
C6. Course workflow combining practice forms with no-AI mastery checkpoints	4	4	3	4	4	4	4	27	1	3	3	1	1

Table 4: Worked-example case scoring under the AI to Learn 2.0 rubric.

C

and

\tau

apply only to learning-intensive contexts. ‘Core’ denotes the author-judged satisfaction of

Core(w)=1

against the formal definitions; it is not derived from the total score alone.

Figure 1: Heatmap of the worked-example scores across the seven AI to Learn 2.0 maturity dimensions. High scores in one region do not guarantee compliance if critical gate dimensions remain below threshold.

Figure 2: Descriptive total scores $S(w)$ for the worked examples. The dashed horizontal lines indicate the interpretation bands proposed in Section 6. The annotations remind the reader that gate satisfaction and, where relevant, capability residual remain decisive.

AI Governance

Artificial intelligence governance studies how laws, institutions, and technical controls shape AI development, deployment, accountability, and social effects.

See all 379 papers

Dynamic Evidence Collection Ecosystem for Assessment Integrity and Authentic Competence
Aug 17, 2026Rajan Kadel, Bellal Hossain, Samar Shailendra +1Academic Integrity Educational Assessment
NAPS, Melbourne, Australia · NAPS, Sydney, Australia · MIT, Melbourne, Australia
Explainability and Certification of AI-Generated Educational Assessments
Mar 18, 2026Antoun Yaacoub, Zainab Assaghir, Anuradha KarEducational Assessment Generative AI in Education
aivancity School for Technology, Business & Society, 151 Bd Maxime Gorki, 94800 Villejuif, France · Faculty of Science, Lebanese University, Hadath Campus, Hadath, Lebanon
AI Misuse in Education Is a Measurement Problem: Toward a Learning Visibility Framework
Mar 8, 2026Eduardo Davalos, Yike ZhangAcademic Integrity Generative AI in Education
Trinity University, San Antonio TX, 78209, USA · St. Mary’s University, San Antonio TX, 78228, USA

Aug 17, 2026J/K move · Enter open · S save

Dynamic Evidence Collection Ecosystem for Assessment Integrity and Authentic Competence

Rajan Kadel, Bellal Hossain, Samar Shailendra, Bushra Naeem

NAPS, Melbourne, Australia · NAPS, Sydney, Australia · MIT, Melbourne, Australia

Academic Integrity Educational Assessment Generative AI in Education

Open paper AskPDF arXiv

Generative Artificial Intelligence (GenAI) can produce high-quality essays, code, and design artefacts, challenging the validity of conventional assessments that rely on single-point submissions and product-only grading. This paper proposes a design framework called "Dynamic Evidence Collection Ecosystem" that shifts assessment toward continuous, authentic, multi-source evidence of student learning over time. The framework collects process evidence through iterative artefacts, design logs, activity rounds, self-reflection, and peer collaboration, supported by an AI-enabled layer for learning analytics, formative feedback, and transparency. The approach is grounded in recent assessment-redesign scholarship in AI-rich contexts and aligned with contemporary views of authenticity in assessment. This paper builds on the hypothesis that academic integrity is strengthened when it is treated as an assessment design rather than as an AI detection problem. The tools have limitations and risks of use that carry academic penalties. This paper presents an implementation scenario to support institutional adoption.

Abstract

AI Governance

Explore similar work

Dynamic Evidence Collection Ecosystem for Assessment Integrity and Authentic Competence

Explainability and Certification of AI-Generated Educational Assessments

AI Misuse in Education Is a Measurement Problem: Toward a Learning Visibility Framework