How do we know a person has learned when they use AI?
The end of the final product as evidence
When any essay, report, presentation or piece of code can be generated or rapidly improved with AI, the final product stops being reliable evidence of learning. EDUCAUSE identifies a transition towards authentic assessment and process-based demonstrations of learning.
This will likely be one of the highest-demand spaces of the coming years: redesigning assessment to observe what AI cannot simulate in a sustained way.
What to assess now
Assessment in the AI era must observe and value:
- Reasoning and quality of decisions.
- Iterations and evolution of the work.
- Transfer to new contexts.
- Oral defence of knowledge.
- Individual contribution in collaborative work.
- Metacognitive reflection.
- Quality of AI use as a tool.
- Evidence of the process, not just the result.
From graduate profile to evidence
The skills-first logic is gaining relevance: demonstrable competencies increasingly outweigh traditional credentials in the relationship between education and employment. Universities will need observable graduate profiles, competency maps, longitudinal evidence, verifiable portfolios and systems to explain what a graduate can actually do.
I work the complete chain: graduate profile → competency progression → learning experiences → assessment → evidence → credentials connected to employment.
Metacognition and traceability
My experience in competency-based assessment, 360º assessment, metacognition and process traceability — developed in BeChallenge and SkillsTracker — makes it possible to design assessment systems that generate useful evidence for the student, faculty, accreditation and the employer.