01 · The validation gap
A good result leaves questions open
The quality of a product and the competency of the person presenting it are related but distinct matters. Recognizing learning requires examining what a person understands, which decisions they can justify and under what conditions they can act.
The paper argues that this gap is historical. Collaborative work, advice and other forms of support already required interpretation of individual contributions. Artificial intelligence expands the scale and accessibility of assistance, making the problem more visible.
Learning with support can be valuable. Assessment needs to examine whether that support develops capabilities, and what evidence distinguishes this from a better result achieved while assistance is available.
Validating competencies: unfinished work that predates AI
School, university and professional education share a question: against what standard do we establish that a person is competent? The same grade at two institutions does not guarantee equivalent capabilities. Assessor severity, the criteria used and biases can influence grades. Passing and earning a qualification require interpreting which capabilities were assessed and what evidence supported the decision.
In La Educación y la Entropía, first published in 2013, Gustavo Gómez Braun discusses subjective grading, differences in expectations and the need to demonstrate educational quality through evidence [1]. That same year, the OECD documented biases in school marks. The debate has concrete precedents.
AI offers new opportunities to address this work: expanding practice and feedback, supporting the application of shared criteria, documenting assessments and making review easier. These opportunities require checking results, detecting errors and improving procedures. Quality should be examined in human, automated and mixed systems, including ACVIS.
References such as ISO 21001 help organize educational quality management. Management system certification alone does not demonstrate each learner’s competency. The debate about AI can encourage a shared expectation across education: substantiating what is recognized as learning and why.
02 · Four dimensions
What to distinguish when examining a piece of work
- Product quality
- How accurate, relevant and useful the submitted result is.
- Personal contribution
- What the person did and how tools, collaborators or other support contributed.
- Personal competency
- What the person can understand, justify and apply within the scope assessed.
- Responsibility
- Who reviews decisions, answers for their consequences and acknowledges the result's limitations.
These dimensions need complementary evidence. A final file may inform judgments about product quality while leaving attribution open. An explanation adds evidence of reasoning; a new situation allows examination of application.
03 · Learning and performance
Assisted performance, learning and transfer
Assisted performance shows what a person achieves with particular resources. Claims about learning also require attention to what they understand and retain. Transfer raises another question: can they adapt that knowledge when the problem or context changes?
The paper distinguishes four gaps: the learning gap, concerning understanding and retention; the performance gap, when observed results do not hold under other conditions; the attribution gap, concerning personal contribution; and the validation gap, concerning what an institution can responsibly claim.
Competency must include its conditions
An assessment may examine independent performance, performance with permitted AI support, or the ability to transfer learning. Each purpose needs explicit conditions and criteria. Removing all support is not a universal rule: it depends on the competency being observed.
The research discussed in the PDF helps frame these questions. This synthesis does not reproduce its figures or turn findings from particular settings into a general estimate of the effects of AI or ACVIS.
04 · People and institutions
Two needs that must meet
The person needs to demonstrate
They may have learned at work, independently or with AI support, without that journey being easy to recognize. They need an opportunity to show concrete capabilities and understand the criteria used.
The institution needs to substantiate
It must decide what it recognizes, within which scope and why. It needs sufficient evidence, clear conditions and a process that allows the conclusion to be reviewed.
The paper connects these needs to an aspiration for inclusion: expanding ways to demonstrate learning developed in different contexts. Access to learning resources alone does not guarantee recognition or opportunities.
05 · From evidence to a decision
A conclusion that can be explained and reviewed
Define what is to be demonstrated
Specify the competency, assessment purpose, criteria and permitted support.
Gather relevant evidence
Combine a performance or product with an explanation of decisions, error review and application to a variation, according to the objective.
Connect evidence to criteria
Record which observation supports each conclusion, which conditions influenced it and which questions remain open.
Make a decision with a defined scope
Explain what has been demonstrated, what evidence is missing and how the process can be reviewed or continued.
Traceability makes that relationship reconstructable. It does not automatically make a decision true: sound criteria, relevant evidence and review remain necessary.
When the evidence is insufficient
The conclusion must be limited to what was observed. Insufficient evidence does not prove that the person did not learn; it may indicate the need for another activity or an additional observation.
06 · ACVIS: contribution and limits
An educational platform organized around demonstration
ACVIS organizes its educational approach around what a person seeks to develop and demonstrate. It connects learning, conditions, assessment, evidence and a documented decision for both personal and institutional pathways.
The original paper uses the idea of trust infrastructure. In this adaptation, that idea sits within ACVIS's educational identity: learning, practice and improvement are part of the journey towards a well-founded demonstration.
The functioning prototype supports examination of its operation in internal test cases. This does not demonstrate general educational effectiveness, institutional adoption or social impact. Those conclusions require pilots and external evidence.
The framework proposes transparency, privacy, proportionate controls, fairness, human supervision and the possibility of review. A reference to a standard expresses a direction; it does not establish ISO conformity or certification.
07 · Impact as a horizon
Learning and being able to demonstrate it
The horizon is for people with diverse learning histories to develop capabilities and show them in ways that those responsible for recognition can understand. This could support education, employment and mobility, provided there are appropriate procedures and effective recognition.
That impact is a possibility to evaluate. Substantiating it requires documented outcomes, institutional participation and follow-up with people. Educational ambition must advance alongside evidence that allows it to be assessed.
The connection with SERAS supplies the human purpose: learning with artificial intelligence should make us more capable of understanding, deciding and creating for ourselves.
Source document and editorial approach
From Knowledge to Evidence: The validation gap and the new trust infrastructure in the age of artificial intelligence
Gustavo Marcos Gómez Braun · Institutional impact paper · August 2026
This page is an ACVIS editorial synthesis based on the original document by Gustavo Marcos Gómez Braun. It selects and reorganizes its central ideas for web reading; it is not a complete transcription and does not replace the PDF. Clarifications about educational identity and current scope belong to this adaptation.
Open the original English PDF →
The passage on assessment before AI is an editorial addition from September 2026. It connects the book and the following references to the current discussion; it is not part of the original PDF linked above.
References for this addition
- [1] Gustavo Gómez Braun. La Educación y la Entropía (Education and Entropy). First edition: August 2013. Printed pages 14, 53–54 and 94–96 of the copy consulted. The author and his books.
- OECD (2013). Grade Expectations, PISA in Focus, No. 26. Research on school marks.
- ISO 21001:2025. Management systems for educational organizations. A management reference; this does not imply ACVIS certification.