The Role of Authentic Assessments and UDL

Jason Blomquist; Liza Long; Joel Gladd

13 The Role of Authentic Assessments and UDL

How This Chapter Fits in Part 2

The preceding three chapters clarified what learning should be visible, where it should be secured, and how AI use should be communicated. This chapter turns to the task itself. Authentic assessment asks students to use knowledge in meaningful contexts. UDL helps remove unnecessary barriers so students have supported ways to show what they know.

The next chapter, Why Process-Oriented Assessment Works Well in the AI Era, focuses on process evidence: drafts, checkpoints, AI transparency, and reflection that show how students arrived at the final product.

Beyond AI-Proofing

The preceding chapters covered learning outcomes, course maps, and assignment-level AI expectations. Those tools help you communicate what students may do and what evidence they need to provide. They also point back to a deeper design question. What makes students want to engage authentically in the first place?

Jason Gulya names the underlying dynamic that makes this question so pressing. When students experience education as a transaction (do the work, get the grade), AI becomes the most efficient way to complete that transaction. Transactional education reduces coursework to a set of deliverables, and AI can produce deliverables faster than students can. The response is to make the work worth doing on its own terms.

Recent research gives this a sharper edge. In a 2026 Wharton working paper, Steven Shaw and Gideon Nave define cognitive surrender as adopting AI output with minimal scrutiny. Participants often followed faulty AI advice and became more confident even after errors. A separate 2026 preprint by Grace Liu and colleagues found that brief AI assistance improved immediate performance but reduced persistence and unaided performance across reasoning and reading tasks. Nick Potkalitsky’s introduction to Terry Underwood’s essay explores possible risks to metacognition and self-efficacy, while Underwood notes that these longer-term risks still need more research. When students outsource the thinking an assignment is meant to develop, they may complete the product without developing the intended skill.

Corbin et al. (2025) draw a useful distinction here between discursive and structural changes to assessment. A discursive change adds language to a syllabus or assignment prompt. “Do not use AI for this assignment.” A structural change redesigns the task itself so that the design limits inappropriate AI use or transforms it into part of the learning process. Their argument is direct. Discursive approaches rest on three shaky assumptions: that students understand what’s allowed, that they’ll voluntarily comply, and that faculty can verify compliance. Structural changes don’t depend on any of those assumptions.

This chapter looks at two approaches that make structural changes possible: authentic assessment and Universal Design for Learning (UDL). These approaches predate generative AI by decades. They have become more relevant now because they can give the work a clearer purpose and reduce unnecessary barriers, although neither approach guarantees that students will avoid inappropriate AI use.

What Is Authentic Assessment?

Authentic assessment asks students to apply knowledge and skills to real-world contexts. The task often resembles something a professional or practitioner would actually do, where the work product has value beyond the gradebook. Villarroel et al. (2018) identify three recurring dimensions of authentic assessment: realism, cognitive challenge, and opportunities for students to judge the quality of work.

Disciplinary authenticity concerns the kind of judgment the task requires. Stolpe, Larsson, and Johansson Falck (2026) argue that critical engagement with AI depends on a field’s language, reasoning practices, and standards for knowledge. A realistic product can still miss that learning when students complete it without using those practices. During redesign, faculty should identify the field-specific decisions that make the product credible: how evidence is selected, how uncertainty is handled, which conventions guide the work, and what the student must be able to defend.

Consider the difference.

Traditional vs. Authentic Assessment Examples
Traditional Assessment	Authentic Assessment
Multiple choice exam on marketing principles	Create a marketing plan for a local nonprofit
Essay analyzing a historical event	Write a grant proposal for a community history project
Problem set with textbook equations	Design a solution for an engineering firm’s current challenge
Short answer quiz on research methods	Conduct and present original research on a campus issue
Written exam on patient assessment procedures	Conduct a simulated patient assessment with observed clinical skills check-off
Multiple choice quiz on welding safety standards	Complete a welding project evaluated against industry certification criteria

The left column primarily asks students to recognize or recall information. The right column primarily asks whether they can use that information to accomplish something meaningful. Authentic assessment often takes the form of project-based work, connecting to an established pedagogical tradition many faculty already practice.

Why Authentic Assessment Supports Motivation

A systematic review of 26 higher-education studies found that authentic assessment can support student engagement, satisfaction, and effort. Students may engage more genuinely when the task connects to their interests. They may invest more when the audience extends beyond the instructor and think harder when the challenge requires real judgment. These conditions can shift how students relate to the work, although no single format motivates every student.

A student designing a marketing plan for their own side project invests differently than one answering abstract questions about marketing theory. That personal relevance deepens when the work is meant for someone who will actually use it. Presenting recommendations to a community partner changes how students approach preparation. Open-ended problems with multiple valid approaches also engage different cognitive processes than questions with single correct answers. Students must make decisions, weigh tradeoffs, and defend their reasoning.

Leon Furze identifies two student mindsets that make this especially relevant. Some students adopt a “good enough” approach, using AI to produce work that meets minimum standards with minimal effort. Others develop a “better than me” belief, assuming AI can write more effectively than they can, which erodes self-efficacy. Transactional contexts support these patterns because the task feels like a hoop to jump through. The “good enough” mindset struggles when the task demands personal investment and local knowledge. The “better than me” belief erodes when students discover they bring irreplaceable judgment and context to the work.

Anna Mills positions intrinsic motivation as the first layer of defense against AI misuse. Her approach layers multiple strategies, starting with framing the value of the work itself, then adding student choice, peer review, and varied formats. Mills’s design argument is that students who care about the outcome may be less inclined to outsource the thinking that produces it.

The AI Dimension

Authentic assessments are sometimes described as “AI-resistant.” The reasoning is that they involve personal experience, local context, or original thinking that AI cannot replicate. This framing is partially accurate but can be misleading.

Jason Gulya tested this assumption by asking an AI agent to complete an exercise designed to be AI-resistant (a chatbot interaction with transcript grading). The agent handled it. As AI capabilities expand, especially with agentic tools, the bar for what counts as “AI-resistant” keeps rising. Treating it as an arms race is a losing strategy.

What seems clear is that authentic assessments can make AI less useful as a substitute and more useful as a tool. A student can prompt ChatGPT to “create my marketing plan for the nonprofit I’m actually working with” and receive a plausible draft. The final work still depends on evidence from the organization, verified local context, and decisions the student can explain and defend. AI can help brainstorm approaches, research comparable organizations, or refine a draft. The authenticity comes from the student’s accountable judgment, especially where local knowledge must be gathered and verified rather than merely generated.

Rethinking Bloom’s Taxonomy for the AI Era introduced Tina Austin’s UnBlooms Framework, which treats AI-era learning as a recursive process of questioning, generating, critiquing, revising, reflecting, and deciding. Students can now produce polished outputs that look like recall, comprehension, application, analysis, or creation without doing the thinking those categories are supposed to name. Authentic assessments can foreground contextual judgment and evaluation that students must exercise and defend, even when lower-level cognitive tasks are automated.

The AI Assessment Scale (AIAS), originally developed by Perkins, Furze, Roe, and MacVaugh and refined in 2025 by Perkins, Roe, and Furze, offers a practical tool for specifying what level of AI use is appropriate for a given assessment. The AIAS was introduced in AI Policy Frameworks and the AI Assessment Scale as part of AI policy frameworks. In this context, it serves a different function. When you redesign an assessment for authenticity, the AIAS helps you communicate to students exactly where AI fits and where it does not. As Perkins, Roe, and Furze note, the scale works best as a conversation starter and assessment redesign framework, not as a policing tool.

Universal Design for Learning (UDL)

Universal Design for Learning is a framework for designing instruction that accommodates individual differences from the start. It was originally developed to support learners with disabilities. UDL now treats learner variability as a starting condition and aims to reduce barriers for learners with and without disabilities by providing multiple pathways to engagement, understanding, and expression.

CAST released UDL Guidelines version 3.0 in July 2024, updating the framework that had been in version 2.2 since 2018. The 3.0 update shifts the core objective from developing “expert learners” to cultivating learner agency. It places greater emphasis on learner identity as part of variability, adds attention to joy and play in learning, and moves toward learner-centered language throughout. The three foundational principles remain.

UDL 3.0 Core Principles
Principle	Focus	Example Strategies
Multiple Means of Engagement	The “why” of learning. Motivating students and sustaining effort.	Offer choice in topics or formats. Connect to students’ interests and identities. Provide options for self-regulation.
Multiple Means of Representation	The “what” of learning. Presenting information in varied ways.	Use text, audio, video, and visuals. Clarify vocabulary. Highlight patterns and relationships.
Multiple Means of Action and Expression	The “how” of learning. Allowing varied ways to demonstrate knowledge.	Accept written, oral, or multimedia submissions when the format is not itself the learning outcome. Provide scaffolds and tools. Allow drafts and revision.

UDL and Authentic Assessment Together

UDL and authentic assessment complement each other. An authentic assessment asks students to do something real. UDL ensures multiple pathways exist for students to accomplish that task successfully.

Consider a final project where students create an educational resource for a community partner. Authentic assessment principles shape the task. It serves a real audience, connects to students’ disciplines, and requires professional-level work. UDL principles shape how students complete it. They might choose video, infographic, written guide, or podcast format when that format is not itself one of the competencies being assessed. As CAST’s assessment guidance explains, flexible options should preserve the knowledge or skill the assessment is intended to measure. Students also receive scaffolded checkpoints and clear criteria across the available formats.

AI tools can also help faculty implement UDL in practice. Lance Eaton notes that generative AI can assist with creating alternative formats, adapting reading levels, generating transcripts, and building scaffolding materials. After faculty review for accuracy, accessibility, and equivalence, these uses can support selected UDL practices that previously required significant time investment. AI doesn’t replace the design thinking behind UDL, but it can reduce the practical barriers to implementing it.

This combination addresses two design concerns. Authentic tasks can give the work a clearer purpose, while UDL options can reduce unnecessary barriers and provide needed scaffolds. Faculty still need explicit AI expectations and valid evidence of learning because task meaning and support do not prevent inappropriate use on their own.

Concrete Examples for Inspiration

A few examples make the pattern easier to see. Each one keeps the learning goal stable while changing the audience, product, support structure, or route through the work.

Authentic Assessment Examples: Student Products
Discipline	What Students Produce
Humanities: public history	A virtual museum label and audio guide for a historical object. Students analyze material culture, explain its context, and present it for viewers in a digital exhibit.
Humanities: primary-source learning activity	A digital learning activity (PDF) built around primary sources. Students connect a current issue to its historical roots and design an activity that classmates can work through.
STEM: environmental education	A presentation-based exam tied to course learning objectives. Students prepare a video presentation that explains environmental education concepts and applications.
STEM: biology or ecology	Two connected science communication products (PDF). Students first write a public-facing article or blog based on a recent research paper. Later, they reflect on feedback and choose a lesson plan for children, a public engagement talk, or a natural-history documentary storyboard based on a module topic.

Authentic Assessment Examples: Design Moves and AI Roles
Discipline	Authentic and UDL Moves	Appropriate AI Role
Humanities: public history	The task resembles public-history and museum work. Students choose an object, interpret evidence, work with models from museum professionals, and use text, image, and audio to reach an audience beyond the instructor.	AI can help students generate background questions, test whether the label is clear for a general audience, or revise the audio script. The historical interpretation, source selection, and ethical framing need to come from the student.
Humanities: primary-source learning activity	Students practice historical inquiry for a peer audience. The design includes topic choice, primary-source work, formative feedback, digital composition, and learning from classmates’ projects.	AI can help with brainstorming activity formats or checking instructions for clarity. Students still need to verify historical claims, interpret the primary source, and decide how peers should encounter the material.
STEM: environmental education	The assessment keeps the content goals of an exam while asking students to communicate, organize, and apply the concepts. This is primarily an authenticity move. Detailed criteria, preparation time, and visual supports reduce some barriers, although every student uses the same video format.	AI can help students outline the presentation, rehearse explanations, or identify places where an audience might need more context. The final evidence of learning is the student’s own explanation and application.
STEM: biology or ecology	Students translate specialized knowledge for non-specialist audiences. The assessment offers choice of product, templates, examples, workshops, and feedback while keeping the outcome focused on accurate science communication.	AI can test readability, suggest analogies, or help students compare explanations for different audiences. Students remain responsible for scientific accuracy, audience fit, and decisions about what to simplify.

These examples show that authentic assessment does not require the same structure in every discipline. Sometimes the authentic move is a real audience. Sometimes it is a professional genre. Sometimes it is an applied performance, a public explanation, or a decision made under realistic constraints. UDL enters through the supports that make the task possible: models, checkpoints, feedback, multiple formats, and clear criteria.

For additional examples, browse the University of Central Florida’s Teaching Repository of AI-Infused Learning. Its collection includes AI-infused activities from the humanities, sciences, health fields, and professional programs. Use an entry as a starting point, then identify what makes the task authentic and whether its formats, supports, and expectations fit the learning outcome for your students.

Redesigning with Authenticity and UDL

The starting point is the assessment you already have. Look at it with fresh eyes and ask what would need to change to make it more authentic and more aligned with UDL.

Faculty Spotlight: Marcus Hickens, Business

In a Principles of Marketing course in the same business program, Marcus faced a different challenge. For years, his midterm has been a 100-point exam with short answer and essay questions about marketing concepts. Students define terms, explain theories, and analyze hypothetical scenarios. The exam works fine as a knowledge check. But Marcus has noticed that students rarely engage deeply with it. They study the night before, write what they remember, and move on.

He decides to run the exam through the authenticity and UDL questions from this chapter. The results are clear. The exam tests recall. It offers no format flexibility. The exam has no real audience. A student could pass it by memorizing definitions without ever applying marketing thinking to an actual problem. With generative AI, a student could produce passable essay answers in minutes. In Corbin et al.’s terms, Marcus realizes that any AI restrictions he adds to the exam would be discursive changes. A policy statement would ask for compliance; the task would still reward fast answer production.

Marcus redesigns the midterm as a marketing plan for a real local business. Students research an actual company, identify a problem or opportunity, and propose a strategy. They can deliver the plan as a written document, a pitch deck, or a video presentation. He adds scaffolded checkpoints, with a business selection proposal in week two, a situational analysis draft in week four, and a peer pitch session in week six before the final submission. This is a structural change. The task requires judgment, context, and direct engagement, which gives Marcus more evidence to review than a one-time generic response even though AI can still assist with parts of the work.

To generate ideas for the scaffolding, Marcus types a prompt into an AI chatbot describing his course outcomes and the new assignment structure. The AI suggests several checkpoint options, including a peer review stage and a presentation to the business owner. Marcus likes the peer review idea. But he knows that arranging meetings with business owners for a class of 35 is logistically difficult. So he scales that suggestion to a peer pitch session, where students present to classmates who role-play as stakeholders.

Using the AIAS, Marcus specifies what level of AI use is appropriate at each stage. For the situational analysis, students can use AI to research industry trends and comparable organizations. For the final strategy recommendation, the judgment and rationale must be their own. Making these expectations explicit helps students understand where AI is a legitimate tool and where the learning depends on their own thinking.

The redesigned midterm assesses some of the original exam’s application and analysis outcomes, along with research and communication skills. If unaided knowledge of definitions or theory remains essential, Marcus will need a separate secured knowledge check. Students who use AI to brainstorm approaches or refine their drafts are using it as a tool, and the authenticity of the work comes from their direct engagement with a real business and real constraints.

Evaluating Your Current Assignment

Start by asking whether the assignment resembles something a practitioner would actually do. If a student can’t see how the work matters beyond the course, the task may be too abstract. Consider whether it requires genuine judgment and decision-making or mainly rewards recall and reproduction. Ask whether the finished work has a real or realistic audience. Assignments that lack these qualities can feel like exercises, and generic exercises can be easier to outsource.

Then look at the assignment through a UDL lens. Do students have meaningful choices in how they approach or present their work? Are there scaffolds or checkpoints that support students at different levels of preparation? Could students with different strengths succeed equally well? Are the expectations clear enough that students know what quality looks like before they begin? Mills recommends layering strategies like student choice, peer review, video assignments, and social annotation. These practices can support choice, feedback, and participation when they align with the learning outcome. The cited source does not show that they reduce AI misuse.

Assignments that fall short on both dimensions are strong candidates for redesign. Apply the structural vs. discursive distinction. If your only defense against AI misuse is an instruction on the assignment sheet, that’s a discursive change. If the task design itself makes AI less viable as a substitute, that’s a structural one.

A Framework for Revision

Furze’s five principles for rethinking assessment (introduced in AI Policy Frameworks and the AI Assessment Scale) provide a useful starting framework here, with validity and authenticity as the most directly relevant. Does the assessment generate trustworthy evidence of learning? And does the task mirror something a practitioner would actually do?

One approach, adapted from the Idaho AI Catalyst initiative, uses AI itself to assist with redesign. The following prompt template can help you generate revision options.

I’ve uploaded my course outcomes, the UnBlooms framework from Rethinking Bloom’s Taxonomy for the AI Era, a presentation on authentic assessment, and my current summative assignment instructions. Propose three different ways to revise the assignment instructions so it works as a process-oriented and authentic assignment, and make sure it also aligns with UDL recommendations. The updated instructions should provide opportunities for using and/or evaluating GenAI outputs.

This prompt generates starting points. Use the AI’s suggestions as raw material. Your judgment determines which ideas fit your context, your students, and your learning outcomes. As Rethinking Bloom’s Taxonomy for the AI Era explained, Bloom’s verbs need evidence behind them. Keep that in mind as you evaluate which levels of thinking your assessment actually measures.

What Changes in Practice

When assignments become more authentic and UDL-aligned, the work usually becomes staged and iterative. A single high-stakes submission becomes a scaffolded process with intermediate checkpoints and feedback loops. Students build their work over time, which makes last-minute generation a poorer fit for the full process. Marc Watkins captures this idea well. Some learning outcomes need to be grown through sustained engagement, not measured through one-time assessments and certainly not automated. The format opens up too. One student might write a detailed report while another creates an annotated video. The learning outcome remains constant. The demonstration varies.

The audience shifts as well. Work products might be shared with community partners, presented to peers, or added to professional portfolios. When someone beyond the instructor will see the work, students prepare differently. The task itself also becomes more specific. An assignment that says “address this particular situation with these particular constraints” makes generic AI output less useful unless the student supplies and verifies local information and exercises professional judgment.

Try This: Redesign One Assessment

Select a summative assessment from one of your courses. Work through these steps.

Audit for authenticity. Does the task resemble something a professional would do? Does it have a real or realistic audience? Does it require judgment beyond recall?
Identify the disciplinary practice. What form of reasoning, evidence use, problem framing, or communication must students practice? Where could AI support that practice, and where could it bypass it?
Audit for UDL alignment. Do students have choices in format or approach? Are there scaffolds supporting different learners? Are expectations clear?
Check for structural vs. discursive. If a student wanted to use AI inappropriately, would they need to violate an instruction (discursive defense), or would the task design make substitution less useful or provide direct evidence of student capability (structural defense)?
Identify one change. Based on your audit, what single revision would most increase either authenticity or UDL alignment?
Draft revised instructions. Write a paragraph describing what the revised assignment would look like. Consider using an AI chatbot to brainstorm options, then apply your judgment to select and refine the best approach.

If you use an AI chatbot during this activity, return to R-T-C-I. Give the AI a role, the assignment redesign task, the course context, and clear instructions about the kind of options you want. Then choose which suggestions fit your students and your outcomes.

You can save the implementation decision for later. This activity practices the analysis that makes redesign possible.

Assessment Idea Bank

If you want more examples before choosing a redesign, see UCF’s Show Your Work: Assessment in the Age of AI in Further Reading. Its short entries cover AI-supported co-creation, learning without AI, and deliberate AI friction across multiple disciplines and assessment formats. Use the strategies as starting points, then evaluate each one against your outcomes, students, workload, and institutional requirements. Some entries depend on current tool limitations. Treat them as design experiments, not AI-proof solutions.

Looking Ahead

This chapter introduced authentic assessment and UDL as approaches that can support intrinsic motivation. The next chapter, Why Process-Oriented Assessment Works Well in the AI Era, turns to process-oriented assessment. It examines strategies like iterative drafting, AI transparency statements, and structured peer and AI feedback, all aimed at making the learning process visible and accountable.

References

CAST. (2024). Universal Design for Learning Guidelines version 3.0. Source link

CAST. (n.d.). UDL and assessment. UDL on Campus. Source link

Corbin, T., Dawson, P., & Liu, D. (2025). Talk is cheap: Why structural assessment changes are needed for a time of GenAI. Assessment & Evaluation in Higher Education, 50(7), 1087-1097. DOI record

Cordingley, M. (2024). Authentic assessment examples of practice. Centre for Academic Innovation and Development, University of Chester. PDF source

Davis, S., Kilmister, M., Mereles, A., & Khamis, A. (2023, January 26). Exhibiting history: OBL assessment online. Public History Weekly, 11(1). Source link

Eaton, L. (2025, April 10). Blending AI, OER, and UDL. AI + Education = Simplified. Source link

Furze, L. (2025, August 18). Five principles for rethinking assessment with Gen AI. Source link

Furze, L. (2025, November 24). Good enough and better than me: Two problematic student perspectives on Gen AI. Source link

Gulya, J. (2025, August 10). If we’re going to adapt to the age of AI, we’ll need to chip away at transactional education. The AI Edventure. Source link

Gulya, J. (2026, February 25). What happened when I asked an AI agent to interact with a chatbot? The AI Edventure. Source link

Liu, G., Christian, B., Dumbalska, T., Bakker, M. A., & Dubey, R. (2026). AI assistance reduces persistence and hurts independent performance. arXiv. Source link

Marsh, A., & Hagan, A. (2024, December). Beyond memorization: The power of authentic assessments in STEM learning. Land-Grant Press. Source link

Mills, A. (2025, August 23). Strategies for reducing the likelihood of AI misuse. Anna Mills’ Substack. Source link

Perkins, M., Furze, L., Roe, J., & MacVaugh, J. (n.d.). The AI Assessment Scale. Source link

Perkins, M., Roe, J., & Furze, L. (2025). Reimagining the Artificial Intelligence Assessment Scale (AIAS): A refined framework for educational assessment. Journal of University Teaching and Learning Practice, 22(7). DOI record

Perkins, M., Roe, J., & Furze, L. (2025). How (not) to use the AI Assessment Scale. Journal of Applied Learning and Teaching, 8(2), 14-23. DOI record

Schrum, K., Abbot, S., Loughry, A., & Catalano, D. C. J. (2024). “I wanted to know”: Engaging learners in the history of higher education through authentic digital assessment. The History Teacher, 57(2), 154-177. PDF source

Shaw, S. D., & Nave, G. (2026). Thinking—Fast, Slow, and Artificial: How AI is reshaping human reasoning and the rise of cognitive surrender. SSRN working paper. DOI record

Sokhanvar, Z., Salehi, K., & Sokhanvar, F. (2021). Advantages of authentic assessment for improving the learning experience and employability skills of higher education students: A systematic literature review. Studies in Educational Evaluation, 70, 101030. DOI record

Stolpe, K., Larsson, A., & Johansson Falck, M. (2026). Discipline-Specific AI literacy (DiSAIL): A theoretical framework for situated engagement with generative AI in education. International Journal of Technology and Design Education, 36, 1917–1932. DOI record

Underwood, T. (2025, April 10). Looking for the next world: Possible risks of cognitive offloading in an AI education landscape [Guest post with introduction by N. Potkalitsky]. Educating AI. Source link

University of Central Florida Libraries. (n.d.). Teaching Repository of AI-Infused Learning (TRAIIL). Source link

Villarroel, V., Bloxham, S., Bruna, D., Bruna, C., & Herrera-Seda, C. (2018). Authentic assessment: Creating a blueprint for course design. Assessment & Evaluation in Higher Education, 43(5), 840-854. DOI record

Watkins, M. (2025, July 29). Some things need to be grown, not graded, and definitely not automated. Rhetorica. Source link

License

Icon for the Creative Commons Attribution-NonCommercial 4.0 International License

A Guide to Teaching and Learning with Artificial Intelligence Copyright © by Jason Blomquist; Liza Long; and Joel Gladd is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, except where otherwise noted.