Insert your title here

Your Text Here

Using Generative AI in Assessments

A Paradigm Shift Framework with Examples

Introduction

Generative AI, can complicate traditional academic integrity, but it also presents a unique opportunity to redesign assessments. With a focus on fostering the critical thinking and durable understanding vital for student success, the following strategies and principles can guide your approach to innovative assessment solutions.

Strategies

Redesign assessment to emphasize the learning process. This includes acknowledging the iterative and often “messy” nature of learning.
Focus on developing higher-order thinking skills such as critical thinking, creativity, and synthesis, rather than solely on knowledge recall.

Key Principles

Prioritize process over product in evaluating student learning.
Move beyond simple knowledge checks towards assessing complex understanding and synthesis.
Equip students with pragmatic skills to effectively and ethically utilize generative AI for problem-solving and achieving learning outcomes.

Paradigm Shift Assessment Framework Examples

Core Challenge	Old Paradigm Focus	New Paradigm Focus	Application Example	Guiding Principle
Standard essays & research papers	A single, polished essay or report, graded primarily on the final product.	Submission of a portfolio showcasing the learning process.	The “Living” Research Paper: AI-generated literature reviews focus on a “Project Log” of AI use, annotated drafts, a final presentation defending unique contributions and reflections, and an AI tool disclosure statement, emphasizing the inquiry process.	Prioritize process over product. This makes visible the “messy” but essential journey of inquiry and discovery, which cannot be automated.
Take-home / in class exams requiring factual recall	Examinations focused on factual recall (e.g., proctored, closed-book multiple-choice or short-answer tests).	Submission of a reflective analysis demonstrating knowledge application.	The Adversarial AI Strategy Session: Students use AI as a research assistant to build a plan (e.g., marketing, policy). They then engage the AI, prompted to act as a skeptical stakeholder (e.g., a “CFO”), to defend their strategy in a simulated challenge. The submission is the annotated transcript of the session, where students explain how they countered critiques or adapted their plan.	Embrace authenticity. This assesses the application and defense of knowledge in a dynamic,real-world context. It transforms the assessment from a test of static knowledge into a simulation of professional practice and strategic thinking.
Basic code generation or literature summaries	Submission of a functional code snippet or a comprehensive literature summary for a grade. This paradigm views AI as a threat, leading to a focus on detection and prohibition.	Refactoring and Analysis Report; Critiqued Research Proposal	AI-Generated Code Refactor: Students use AI to generate a first draft, but are graded on their process of identifying flaws, implementing improvements, and justifying their changes in a detailed report. AI-Augmented Research Proposal: Students use AI for initial literature search but are assessed on their ability to design a rigorous experiment, critique the AI’s suggestions, and formulate a unique, testable hypothesis.	Cultivate AI literacy.This approach leverages AI as an assistive learning tool to be mastered, not a threat to be banned. The assessment targets human expertise, such as debugging, critical analysis, and strategic improvement, which are durable professional skills.
Reading responses, simple analyses, solutions to defined problems	A refined response essay or solution as a final product.	A critique demonstrating how students think.	Scholarly Deconstruction: Students create and/or critique an AI-generated passage, demonstrating their ability to identify bias, find inaccuracies, and build a nuanced argument with credible evidence.	Oversee AI substantively. This involves a detailed, line-by-line critique where students identify specific flaws and substantiate their corrections with evidence from course materials, thus overseeing the AI’s work.

Schedule an Appointment with a TLOS Consultant