What is the central trade-off between improvement and inequality?
The core tension is that AI assessment redesign can make learning more personalized, authentic, and aligned with real-world skills—but it also requires significant resources, training, and infrastructure that are not evenly distributed. Faculty members in one study explicitly cited 'equity and accessibility concerns' as a major challenge when redesigning assessments for the generative AI era [1]. Similarly, a survey of 285 educators, students, and ICT personnel in Nigerian technical education found that while AI tools like adaptive testing and learning analytics were widely recognized as beneficial, barriers such as 'poor infrastructure, limited training, high implementation costs, and resistance to change' were seen as major obstacles [2]. This means that without deliberate investment, AI assessment redesign could widen the gap between well-funded institutions and those with fewer resources.
Where do the studies agree that redesign can work without widening gaps?
The evidence also shows that AI can enhance authentic assessment and experiential learning, which are inherently more equitable because they focus on real-world application rather than rote memorization. One study demonstrated how generative AI tools can be used as 'agents-to-support experiential learning for authentic assessment,' moving beyond simple question-answering to cultivate higher-order skills [3]. This approach can benefit a wider range of learners, including those who struggle with traditional testing formats, potentially narrowing achievement gaps rather than widening them.
Where do the studies show that inequality could still grow?
The studies also differ in scope: the AIAS framework [4] and the human-centered model [5] are theoretical or based on expert feedback, while the Nigerian study [2] provides concrete survey data showing the gap between awareness and actual use. This doesn't mean the frameworks are wrong, but it does mean that their success depends on implementation context. The most cautious takeaway is that AI assessment redesign can improve outcomes without widening inequality, but only if accompanied by deliberate investment in equity—training, infrastructure, and inclusive design—as emphasized across all five studies.
About These Sources
This answer is built on 5 peer-reviewed studies — published from 2024 to 2025, 5 from 2024 or later, 1 in Q1 journals, collectively cited 186 times — selected as the most relevant from 5 studies that passed quality screening, drawn from 62 papers retrieved from a database of over 500 million.
Sources used in this answer
Redesigning Assessments for AI-Enhanced Learning: A Framework for Educators in the Generative AI Era
In a qualitative study of 61 faculty members, key motivations for redesigning assessments included maintaining academic integrity and preparing students for AI-augmented careers, but equity and accessibility concerns were flagged as major challenges, leading to a proposed 'Against, Avoid, Adopt, and Explore' framework.
Leveraging Artificial Intelligence to Redesign TVET Assessment Systems for Enhancing Creativity and Innovation in Technical Education
A survey of 285 educators, students, and ICT personnel in Nigerian TVET found that while 57.9% were aware of AI in education, confidence in using AI tools was only moderate, and barriers like poor infrastructure, limited training, and high costs threaten to widen inequality if not addressed.
Using Generative Artificial Intelligence Tools to Explain and Enhance Experiential Learning for Authentic Assessment
Using thing ethnography, this study showed how generative AI tools can be integrated with authentic assessment and experiential learning to cultivate higher-order skills, moving beyond simple question-answering to support deeper learning.
Reimagining the Artificial Intelligence Assessment Scale: A refined framework for educational assessment
The AI Assessment Scale (AIAS), adopted by hundreds of institutions and translated into 30 languages, was updated to be non-hierarchical and inclusive, adding an 'AI exploration' level to prepare students for an AI-augmented world while strengthening assessment validity.
MODELS OF INTEGRATING ARTIFICIAL INTELLIGENCE INTO ASSESSMENT IN HIGHER EDUCATION
This study proposed a human-centered, AI-integrated assessment model with five components—digital platforms, AI tools, human oversight, stakeholder capabilities, and ethical governance—designed to keep faculty central in decision-making and ensure alignment with pedagogical goals.