AI Assessment Scale

Outlined below are five different levels at which AI may be integrated in assessments, from Level 1 = No AI Use to Level 5 = Full AI Use, following the AI Assessment Scale (Perkins et al. 2024). For each level, we outline appropriate AI use cases, sample student instructions, ideas for assessments and additional resources. Note that these are suggested guidelines that may have to be adapted to your specific context. More generally, you can use guiding questions in the AI decision tree to choose a specific level when designing assessments.
The assessment is completed entirely without any form of generative AI assistance. Ideally, this level is recommended for assessments taking place under supervised settings or in-class formative assessments and activities involving primarily practicing skills. Other reasons include:
- Assess Skills Without Technology
Evaluating mastery of key skills, experiences or knowledge without the use of technology.
- Focus on Tasks Inherently Unsuitable for AI
Tasks themselves do not need AI, for example, playing a physical sport, performing arts, handiwork, oral negotiation skills), or careers related to those tasks limit the use of AI, cybersecurity, data management and privacy.
- Protect Opportunities for Practicing Foundational Skills
Allowing AI will undermine the learning objectives of the class, prevent students from practicing or mastering foundational skills.
- Maintain the Validity of Skill-Based Assessments
Allowing AI may challenge the validity of assessments since students are able to complete the task with AI without necessarily acquiring those skills.
“You must not use AI at any point during the assessment. You must demonstrate your core skills and knowledge.” - (Perkins et al. 2024 AI Assessment Scale)
“You may NOT use gen AI tools on any assignments in this course. Gen AI use in this course will not enhance student learning, nor serve the topic at hand. If gen AI use is suspected on an assignment, I will consult with you and ask that you re-do the assignment using no gen AI within X number of days of the original due date.” - (CU-BFA Suggested Syllabus Statements)
Supporting Transparent and Equitable No-AI Assessment Policies
If generative AI is prohibited, ensure your reasons to do so are centered on what works best for evaluating student learning accurately, rather than resistance to restructuring assessments to be relevant and authentic. Additionally:
Explain the "Why" Behind No-Policies
Explain to students why you are prohibiting the use of AI and what types of support (e.g., study strategies, peer support, scaffolded assignments, reflection, office hours) you plan on offering to help students learn and succeed in the course.Track Progress Through Drafts and Version History
Encourage the use of version history, time-stamped images or screenshots, and intermediate drafts to help students keep track of their progress and receive regular feedback.Design Assessments That Emphasize Reflection and Analysis
Design assessments that require students to engage in reflection on their personal experiences, or engage in analytical thinking in ways that are typically harder to simulate authentically and quickly using AI.Use Technology-Free Assessments with Equity in Mind
You may consider using blue books, scantrons, other types of hand-written, or oral assessment methods to administer assessments in a technology-free setting. However, proactively design assessments to allow for re-takes, alternate modalities and adequate academic accommodations to ensure equity and access in the assessment process.Clarify Expectations Around Academic Integrity
Emphasize the importance and value of academic integrity in your respective discipline, while also discussing what would be considered a violation of the honor code and how it will be addressed.Avoid Overreliance on AI Detectors
Given inherent biases, inaccuracies and privacy concerns associated with AI detectors, we discourage the use of the same to fail students without other means of recourse.Clarify Use of AI-Embedded Tools and Accommodations
Since many tools and search engines, including those that may be approved accommodations, have functions that automatically utilize generative AI (e.g., Google, Adobe), it is important to also clarify if the use of AI-assisted tools, e.g., Grammarly, Turnitin etc., can be utilized at this level.
- Engaging in a Socratic dialogue, debate or discussion on topics, focusing on understanding of concepts, argumentation skills, critical thinking and reflection
- Performing or demonstrating live, individually or in a group. For asynchronous online classes, documentation of such work could be requested through the use of recorded transcripts, pictures or video excerpts.
- Essay or reflection on a personal, place-based experience (e.g., Reflecting on connections between species observed during a nature walk)
- Problems contextualized as real-world case studies.
- Timed, proctored exams including authentic assessments such as specimen identification, conducting experiments.
For more ideas explore:
Additional Resources:
- Mondschein, K. (2023). Avoiding cheating by AI: Lessons from medieval history. Medievalists.net.
- Center for Faculty Development & Innovation. (n.d.). Sample rubric: Writing assignment (no use of generative AI). Southern Illinois University.
- O’Connell, J. (2024). Reimagining oral assessment in the age of AI. eCampus News.
- Desai, H. (2025). What’s worth measuring? The future of assessment in the AI age. UNESCO.
- Mills, A. (2024). Why I'm using AI detection after all, alongside many other strategies. Anna Mills’ Substack.
- Gernsbacher, M. A., Soicher, R. N., & Becker-Blease, K. A. (2020). Four empirically based reasons not to administer time-limited tests. Translational Issues in Psychological Science, 6(2), 175–190.
- Williams, R. (2023). AI-text detection tools are really easy to fool. MIT Technology Review.
- New Student & Family Programs, CU Boulder. (2023). 3 things to know about AI and the Honor Code. University of Colorado Boulder.
- Center for Disability and Access, CU Boulder. (2023). Academic accommodations. University of Colorado Boulder.
- Darby, F. (2024, October 30). 5 small steps for AI skeptics. The Chronicle of Higher Education.
At this level, AI for specific purposes such as initial ideation, brainstorm, outline tasks, or conduct initial research is permitted. However, students are expected to completely write or produce content in the final assessment. In essence, AI can be used to aid development of one’s own ideas and writing but not to generate new content or even edit work for the assignment. Reasons for this level of AI use may include:
- Introduce AI as a Skill-Building Tool
Teaching students to use AI and providing an entry point to learn how to incorporate it in their own workflow.
- Encourage Deeper Engagement and Critical Thinking
Getting students to deeply engage with their ideas through exploring alternatives, and thinking through pros and cons of approaches they may have already come up with.
- Use AI as a Personalized Tutor
Use it as a tutor to help practice skills, improve learning or clarify concepts.
- Support Idea Organization and Collaboration
Students need a way to organize their ideas or aid collaboration, although they may have pre-existing ideas or knowledge on the topic.
- Simulate Complex Scenarios for Deeper Analysis
Simulate innovative or disruptive case scenarios for a deeper discussion or to evaluate analytical skills based on these cases.
"No AI content is allowed in the final submission” - (Perkins et al. 2024 AI Assessment Scale)
“You may use gen AI tools on specific assignments in this course, but their use is limited to the following particular tasks: (explain here– e.g., brainstorming but not for rewriting sections of the assignment, editing. However, the final work must be student-generated, with proper critical evaluation and original analysis or as outlined for each individual assignment.” - (CU-BFA Suggested Syllabus Statements)
Supporting Responsible and Transparent Use of AI in Student Work
- Document AI Use Through Transparent Submissions
Encourage students to provide transcripts, link to their chat interface, or outlines as their intermediate submissions so that timely feedback can be provided and usage of AI can be documented.
- Offer Support Channels to Address AI Misunderstandings
Provide avenues to clarify queries through additional office hours or shared discussion boards, given the possibilities of hallucinations or biases introduced through AI.
- Encourage Credible AI Tools Aligned with Course Content
Recommend the use of tools such as Notebook LM which summarize information from readings specific to the course, increasing the credibility of sources students use.
- Using AI to generate initial concepts or visualizations of an idea students are working on implementing, improving or evaluating the quality of the AI-generated content (e.g., product prototype, sample data generated for analysis)
- Using AI to create case studies that help applying concepts in a new context (e.g., solving application based problems in graph theory, creating a financial plan for a novel business idea generated by AI)
- Using AI to transcribe voice notes, organize summaries (create podcasts from notes to study) or outline next steps based on a collaborative session students had.
- Using AI to explore and identify relevant references or conduct initial research
For more ideas, explore:
Additional Resources
- Goblin Tools. Magic ToDo: AI-powered task management for neurodivergent users.
- University of Colorado Boulder Library. (2024). AI for students: Resources and guidance.
- Brown University Library. (2024). Information discovery with AI: Library guide to generative AI. Center for Computation and Visualization.
At Level 3, AI can be used to improve the clarity or quality of student-created work to improve and refine the final output, but new content should not be AI-generated. Students may be asked to include drafts of student-generated text with no AI use provided in the appendix cases, demonstrating their unique voice, judgment, and point of view.
“AI can be used, but your original work with no AI content must be provided in an appendix” - (Perkins et al. 2024 AI Assessment Scale)
“You may conditionally use gen AI tools on specific assignments in this course to refine output that is otherwise fully student-generated. If you use gen AI tools on assignments in this class, document your usage with the Chicago Manual of Style or appropriate citation guidelines for this course. You will also provide the original draft created by you in the appendix with changes tracked to indicate [enlist what type of edits would you like them to demonstrate] changes made using AI-derived input or feedback. AI-generated textual changes cannot exceed [X%] of the document” - adapted from CU-BFA Suggested Syllabus Statements
Designing Assessments That Support Thoughtful AI Collaboration
- Integrate Self-Reflection on AI-Assisted Writing
Pair assessments with opportunities for self-reflection, asking students to articulate ways in which they plan on applying the feedback in writing with AI-assistance.
- Discuss the Impact of AI on Voice and Expression
Educate students on the risk of AI in standardizing the type of language used, simultaneously limiting variability in expression.
- Adapt Rubrics to Evaluate AI Collaboration
Modify rubrics for the assessment to incorporate relevant criterion that evaluates the quality of AI-assisted editing that students engaged in.
- Using AI to format citations, correct grammar or spell-check in written assignments,
- Using AI to gather feedback on structural changes in evaluating clarity, organization or coherence of arguments in written, oral or visual work.
- Using AI to clarify or unpack rubrics provided by instructors, and use it to self-assess.
For more ideas, explore:
Additional Resources:
- Klebanov, S. (2025, June 29). AI is making everyone sound the same. Morning Brew.
- Southern Illinois University Edwardsville. (2025). Sample rubric 1: Narrative writing with generative AI. Center for Faculty Development and Innovation.
- Moxley, J. M. (2024). Writing with AI: Practice critical AI literacies. Writing Commons.
At this level, tasks may involve AI use at any or all stages of the assessment process including idea generation, gathering feedback or refining content. However, students are required to critically evaluate and modify the output generated by AI, and will be held responsible for veracity of the final submission. These types of assessments may be useful when:
- Center Assessment on AI Literacy Skills
Primary focus of the assessment is on assessing and honing AI literacy skills such as critical thinking, analysis, collaboration etc.,
- Leverage AI for Complex, Time-Intensive Projects
Conducting a larger project which may not be possible to complete in a given term without the use of AI, but still requires human critique and judgement.
- Analyze and Compare AI-Generated vs. Human Work
Comparing AI-generated content across platforms or, to human-generated work.
“You will use AI to complete specific tasks in your assessment. Any AI created content must be cited.” - (Perkins et al. 2024 AI Assessment Scale)
“You may conditionally use gen AI tools in this course on any assignment. Gen AI use is permitted for the purpose of [insert your specific conditions here], with clear attribution and explanation of AI's role in the completion of the assignment. If you use gen AI tools on assignments in this class, document your usage with the Chicago Manual of Style or appropriate citation guidelines for this course” - (CU-BFA Suggested Syllabus Statements)
Ensuring Academic Transparency and Source Integrity in AI Use
- Require Clear Citation of AI Use
Students should clearly cite the use of AI in references, caption AI-generated images, or illustrate in the methodology their use of AI.
- Teach Source Verification and Citation Tracing
Educate students on how to trace original sources of information based on AI-generated output given the potential for AI to hallucinate or fabricate citations.
- Using the same prompt to use across different AI platforms to reflect and analyze variability in responses or images generated by AI.
- Critiquing inaccuracies, biases, limitations as well as opportunities in human and AI-generated visual or textual data and its impact on society
- Using AI for preliminary analysis or visualization of data, followed by manual checking for inconsistencies, errors or alternate ways to solve the problem that students come up
For more ideas, explore:
Additional Resources:
- Mackie, K., & Aspenlieder, E. (2024). The curious educator’s guide to AI: Strategies and exercises for meaningful use in higher ed. Paul R. MacPherson Institute for Leadership, Innovation and Excellence in Teaching.
- Means, T. (2024). Embracing AI innovation: Unlocking human potential. LinkedIn.
- MIT Sloan Teaching & Learning Technologies. (n.d.). When AI gets it wrong: Addressing AI hallucinations and bias. MIT Sloan EdTech.
- University of Northern Colorado. (n.d.). Captioning protocol. Disability Resource Center.
At this level, students can use any gen AI tool for completing the entire assessment, with the focus being on unlocking the potential of AI, exploring new avenues and innovating while learning ways to manage and direct generative AI to perform tasks while also evaluating and analyzing the content. These types of assessment may be useful if:
- Assess Understanding of Generative AI Tools
The focus of the assessment is to evaluate student understanding of gen AI tools
- Allow AI for Non-Essential Process Support
Streamline processes such as editing, transcribing, presentation quality, grammar, etc. (typically the next level of AI use), only if they are not the core skills being assessed in the course, and so AI use doesn’t matter
- Use AI to Support Substantive, Time-Intensive Tasks
The core skills being assessed require substantive groundwork that may not be feasible to complete in the span of a semester (e.g., writing code to analyze data, generating sample dataset, creating storyboards)
- Prepare Students for AI-Integrated Careers
Evaluate student preparation to manage AI-agentic models and work in careers that substantively integrate Artificial Intelligence (AI).
“You may use AI throughout your assessment to support your own work and do not have to specify which content is generated” - (Perkins et al. 2024 AI Assessment Scale)
“You may use gen AI tools in this course, however, you see fit. In fact, some assignments will require the use of such tools. However, keep in mind your own and others' privacy implications and security risks when using gen AI tools. If you use gen AI tools on assignments in this class, document your usage with the Chicago Manual of Style or appropriate citation guidelines for this course.” - (CU-BFA Suggested Syllabus Statements)
Empowering Students to Engage Critically and Ethically with AI
- Build Foundational Knowledge for AI Critique
It is extremely important to make sure students have the foundational knowledge to carefully critique AI output.
- Facilitate Ethical and Societal Conversations Around AI
Encourage open discussions on the ethical, philosophical, human, environmental and social consequences of using generative AI.
- Foster Intrinsic Motivation for Responsible AI Use
Students may not be required to disclose AI use, which requires enough preparation and intrinsic motivation that promotes AI use in line with course goals.
- Using AI-generated code to analyze data collected by students, with students explaining AI’s chosen approach and verifying accuracy.
- Creating AI-generated market analysis reports with data analysis and visualization.
- Building a movie franchise using AI complete with visualizations and images based on a script written by students
- Using AI for preliminary analysis or visualization of data, followed by manual checking for inconsistencies, errors or alternate ways to solve the problem that students come up
- Using tools like Github CoPilot to develop or debug software projects or test code
For more ideas, explore:
- Deep dive with AI Assessment Scale: Level 5
- Contextualization Bot for Aerospace Engineering: Teaching & Learning with AI Repository
- Teaching with text-generation technologies
- The AI Pedagogy Project
Additional Resources:
- McKinsey Global Institute. (2023). Generative AI: How will it affect future jobs and workflows?. McKinsey & Company.
- Means, T. (2024). Instructional minimalism: How a human-AI partnership can transform teaching. LinkedIn
- Means, T. (2024). Threshold of higher education in the imminent AGI era. LinkedIn.