Assessment Twins: A Protocol for AI-Vulnerable Summative Assessment

2025-10-03

AI Summary

Under the disruptive influence of generative AI (GenAI), higher education assessment faces escalating crises of authenticity and validity. This paper proposes the “Assessment Twins” framework: a systematic approach involving paired, multimodal, interdependent assessment tasks validated through cross-verification to ensure both authenticity of learning outcomes and pedagogical coherence. Innovatively grounded in Messick’s validity framework, it integrates content, structural, and consequential validity dimensions, thereby reducing reliance on proctoring technologies. The framework operationalizes a scalable workflow comprising validity mapping, learning-outcome alignment, complementary task pairing, and interdependent scoring, supported by a three-phase implementation model. Empirical evidence demonstrates that the framework effectively safeguards academic integrity and fosters deep learning; however, it also uncovers persistent challenges related to resource allocation and educational equity.

Technology Category

• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Generative Artificial Intelligence (GenAI) is reshaping higher education and raising pressing concerns about the integrity and validity of higher education assessment. While assessment redesign is increasingly seen as a necessity, there is a relative lack of literature detailing what such redesign may entail. In this paper, we introduce assessment twins as an accessible approach for redesigning assessment tasks to enhance validity. We use Messick's unified validity framework to systematically map the ways in which GenAI threaten content, structural, consequential, generalisability, and external validity. Following this, we define assessment twins as two deliberately linked components that address the same learning outcomes through different modes of evidence, scheduled closely together to allow for cross-verification and assurance of learning. We argue that the twin approach helps mitigate validity threats by triangulating evidence across complementary formats, such as pairing essays with oral defences, group discussions, or practical demonstrations. We highlight several advantages: preservation of established assessment formats, reduction of reliance on surveillance technologies, and flexible use across cohort sizes. To guide implementation, we propose a three-step design process: identifying vulnerabilities, aligning outcomes, selecting complementary tasks, and developing interdependent marking schemes. We also acknowledge the challenges, including resource intensity, equity concerns, and the need for empirical validation. Nonetheless, we contend that assessment twins represent a validity-focused response to GenAI that prioritises pedagogy while supporting meaningful student learning outcomes.

Problem

• Redesigning assessments to address GenAI threats to validity
• Mitigating validity threats through complementary evidence formats
• Developing paired assessment tasks for cross-verification of learning

Innovation

• Assessment twins pair complementary evidence modes
• Cross-verification through closely scheduled linked tasks
• Triangulating evidence mitigates GenAI validity threats