Evaluación auténtica e inteligencia artificial generativa en la formación docente
Year: 2026
Type: article
Abstract: Generative artificial intelligence weakens the traditional inference between a submitted product and the learning achieved, requiring higher education to reconsider assessment validity, authenticity, and fairness. This article analyses the challenges that generative artificial intelligence poses to authentic assessment and derives design criteria relevant to teacher education. A critical integrative review was conducted. Preliminary identification relied on Elicit through four search clusters in Spanish and English and was complemented by reference chaining and verification of DOIs and publishers' webpages. The analytical corpus comprised 51 documents: 37 publications from 2022 to June 2026 and 14 earlier foundational contributions. Information was organised in a thematic coding matrix comparing assessment problems, design strategies, teacher competencies, and ethical risks. The synthesis identified five categories: crisis of assessment evidence, authenticity and task redesign, process traceability, teachers' professional judgement, and algorithmic equity. Based on these categories, a framework of six interdependent design moves is proposed: contextualise, disclose sequence, defend, feed back, and care. Authentic assessment remains relevant when it shifts emphasis from controlling the final product to pedagogical interpretation of the learning process, incorporates multiple forms of evidence, and protects agency, privacy, and equity. The review also maps limitations and priorities for empirical validation in teacher education.
Author: Dionelio Jesús Moreno Villalobos
Institution: Universidad EAN
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: diamond
OpenAlex ID: W7225335849