A Tale of Two Assessments: Navigating Generative AI Integration in Authentic Assessment Design



Author Information

Ramalakshmi Vaidhiyanathan, Cardiff University, United Kingdom

Abstract

The study presents a comparative analysis of two approaches to authentic assessment design, highlighting how scaffolding and generative AI integration influence student learning in computing education. In the first approach, students engaged in structured coursework supported by predefined requirements and starter code. A custom AI chatbot was provided to enhance comprehension and guide processes while deliberately restricting code generation to uphold academic integrity. This structured design fostered strong student engagement, reduced instructor intervention, and supported consistent understanding of expectations. Although some evidence suggested limited AI-assisted code generation, the controlled scope mitigated the risk of overreliance on automated solutions. In contrast, the second approach adopted an open-ended use of generative AI, permitting students to employ AI tools as virtual clients to create requirement specifications and to use AI-generated code with declaration. This high-autonomy model unintentionally elevated student anxiety and increased requests for instructor support. Students faced difficulties determining appropriate requirements to demonstrate learning outcomes and struggled to connect AI outputs to underlying design principles. Instead of empowering learners, unrestricted AI use resulted in decision paralysis and conceptual confusion. Together, these findings underscore that autonomy without adequate scaffolding can overwhelm learners engaging with emerging technologies. Effective AI integration requires calibrated structure that supports critical thinking, along with explicit instruction on evaluating AI outputs against learning objectives as a core digital literacy competency.


Paper Information

Conference: ECE2026
Stream: Assessment Theories & Methodologies

This paper is part of the ECE2026 Conference Proceedings (View)
Full Paper
View / Download the full paper in a new tab/window


Comments & Feedback

Place a comment using your LinkedIn profile

Comments

Share on activity feed

Powered by WP LinkPress

Share this Research

Posted by James Alexander Gordon