With the evolution of generative artificial intelligence, it’s important for instructors to consider how we evaluate student learning, the relationship of disciplinary skills and knowledge to what we observe as their proxies, and what the key skills are that we need to be sure our students can demonstrate unaided by the time they graduate. Addressing this at the program level can help us to plan assessments that reliably and accurately evaluate those skills and knowledge.
The following table provides an overview of the possible impact of AI on assessments. Though, it is critical that instructors keep in mind that altering assessments requires reflection on and consideration of pre-determined learning outcomes and goals; and some alternative assessment types included in the table may disadvantage students with disabilities.
The Centre for Teaching and Learning also has numerous resources on Assessments and Alternative Assessment Design
Assessment Approach |
Susceptibility to AI Completion |
Why? |
Incorporating AI into this assessment type |
Alternative Assessments |
Accessibility / equity of alternatives |
|---|---|---|---|---|---|
|
Essays |
High |
AI can generate coherent and contextually relevant text. AI is built into many common tools now and may be hard to turn off |
Focus on documenting process, iterative improvement and co-creation. Use AI to provide feedback, editing advice. Use AI to generate essays that students critique/grade/mark up |
Presentations, group discussions |
May disadvantage some students with disabilities |
|
Mathematical problems |
High |
AI tools can solve well-defined mathematical problems, both from digital symbols and images |
Use AI to create varied problems or data; use to provide alternative explanations / tutorials to students who need it |
Project-based learning; multi-step projects; analyse real-world data sets; written/oral explanation of the problem |
Relatively accessible / equitable |
|
Coding assignments |
High |
AI can generate code for basic and increasingly complex tasks. |
Allow AI-assisted coding using industry standard tools with students explaining their code and debugging processes |
In-class demonstration and discussion of code with individual students |
Relatively accessible / equitable |
|
Research papers |
High |
There are many tools designed specifically to summarise research/sources and help write research papers |
Use AI research assistant tools and document the process; compare specialised research support tools to general tools like ChatGPT; use writing tools to edit or polish a paper; create ‘plain language’ summaries of research papers for public audience and have students compare and critique them |
In class drafting or writing sprints, such as Pomodoro method; use collaborative annotation software; keep version history; short paper plus presentation; poster presentation |
Timed in-class writing can disadvantage students with disabilities; other options relatively accessible / equitable |
|
Synthesis paper |
High |
AI tools are very good at summarising and synthesising from provided works |
Students create their own outline and notes for a synthesis paper first then ask an AI tool to generate the same – compare and contrast these; use AI to suggest/ identify potential novel connections to other literature that may have been missed |
Oral presentation (in person or online); poster presentation; digital story; 3-min thesis format presentation |
Oral presentations and short presentations can disadvantage some students |
|
Practicum/ internship reflections |
High |
Reflections are personal and as generative AI is designed to return human-sounding content, it can create a plausible response if the user provides a small amount of context and personal detail |
Use AI to generate example reflections for students to help guide them – incorporate range of errors/problems and ask students to identify those and suggest alternatives; |
Recorded reflections (audio, video) in situ or at the end of each day, posted to LMS frequently; conversation with supervisor; response to supervisor evaluation (oral or written); |
Relatively accessible / equitable |
|
Discussion forum posts |
High |
These are usually low stakes with broad prompts or a reading for stimulus, and AI can write plausible responses to these prompts, and to other student responses |
Use AI to generate scenarios or cases as stimulus for the discussions; use AI to create exemplar posts; |
Students lead/facilitate discussion (in-class, online, hybrid); students concept map or visually summarise the discussion (the concept map is graded); students submit regular video reflections; use voicethread approach – audio/video discussions rather than text |
Visual representations can disadvantage some students with disabilities; other options relatively accessible / equitable provided students have an appropriate device |
|
Podcast |
Moderate-High |
Generating a script and voice for a podcast are easily achieved, but currently require use of more than one tool. |
Use AI to create music and graphics for the podcast; students draft a script and ask AI to act as an editor/producer and offer feedback on the script or ideas on how to cut it down to a required length; students write a script with multiple characters and get AI to voice some of the characters using the script; students use AI to create different versions of the podcast for different audiences |
Digital story (multimedia); Augmented Reality (AR) story (both could be group based) |
May disadvantage some students with disabilities; students require digital devices to complete project |
|
Case Studies |
Moderate-High |
Cases tend to be complex and nuanced with no ‘right’ answer, but AI can provide responses that analyse multiple subtle indicators in cases. Students will likely need to adapt outputs to develop high quality responses |
Use AI to create or modify diverse case scenarios and/or data for students to address; use AI to answer your case questions and see how it responds, then tweak the cases; |
Simulations; internships |
May pose challenges for students with disabilities and other equity needs |
|
Annotated bibliography |
Moderate-High |
There are many tools designed specifically to summarise research/sources |
Use research-focused AI tools to support the development of the bibliography; get students to identify what sources are missing or over-emphasised |
Concept map |
Many concept mapping tools pose challenges for visually impaired students |
|
Multiple Choice Questions |
Moderate |
While AI can potentially answer factual multiple-choice questions, it can still struggle with complex reasoning, multi-part questions, and nuanced understanding |
Use AI to generate diverse questions and automate grading |
Open-ended questions, case studies, questions with no right answers |
Relatively accessible / equitable |
|
Peer Reviews |
Moderate |
While the review text may be easily generated, the complexity of the peer review process can make it harder to complete entirely using AI |
Use AI to simulate the review process and provide guidance on feedback given to peers to prepare them for unaided reviewing; use AI to generate example papers to review; |
Self-reflections; group critiques |
Relatively accessible; possible equity considerations for minority students |
|
Short answer |
Moderate |
AI can generate plausible-sounding answers, but may lack depth and originality. |
Give students AI-generated practice questions to complete, and then have them get feedback from AI on their answers against a rubric and/or pointing to specific areas to re-read/go over from the course content in the LMS; |
Multimodal answers (text, image, video, audio) |
Relatively accessible / equitable provided students have choice in modes of expression |
|
Lab Reports |
Moderate |
AI tools can summarise and visualise data, and write all sections of reports if appropriately prompted. Some adaptation is likely needed to produce a high quality output |
Use AI to check for grammatical accuracy, consistency, and proper formatting; use AI to analyse and visualise data then students discuss and critique its interpretation/results |
Hands-on experiments; video demonstrations |
May pose accessibility / equity challenges |
|
Oral Presentations |
Moderate |
While the presentation itself will be secure, the content of the presentation (script, slides etc.) can be easily generated by AI |
Use AI to analyze speech patterns, clarity, and content accuracy; use AI to suggest or create visuals to support the presentation; |
Debates; role-plays |
Likely to have the same accessibility concerns as the original |
|
Video presentations |
Moderate |
Video is highly susceptible to AI generation and manipulation, but currently requires some skill and specialised apps to do well |
Write script and get AI to present it; incorporate AI characters in presentation; use AI to provide feedback on script for different audiences or suggestions on how to reach a time limit |
Screening party – present pre-recorded videos then discuss with the students; poster presentation; in-class presentation |
Likely to have similar accessibility concerns as the original |
|
In-class exam - online |
Moderate |
If students have access to digital devices, they will be able to use AI to generate answers, though with close invigilation it may be minimised |
Potential uses will depend on question types / tasks, and the learning outcomes being evaluated |
Oral exam |
May pose accessibility / equity challenges |
|
In-class discussion |
Moderate - Low |
While the student will still have to participate and share, they may use AI to generate ideas and contributions to the discussion |
Students may use AI to check their understanding of ideas in real time before contributing to the discussion |
Device-free discussions; small group collaborative summaries; concept maps/summaries on paper + gallery walk |
May pose some accessibility / equity challenges, especially for visual representations |
|
Digital stories |
Moderate - Low |
Individual elements of digital stories can be created by AI, but producing a quality outcome will require some skill and production management from students |
Use AI to support creation of the various elements of the digital story e.g. graphic development, editing a script, AI characters, writing code for digital elements |
Simulations; internships; poster presentation; viewing party with Q&A; students record a ‘making of’ video; |
Likely to have similar accessibility concerns as the original, but student choice of format may improve this |
|
Portfolios |
Moderate-Low |
AI may be able to create individual components of portfolios, but the complex, multimodal nature of them makes that more challenging to develop coherent portfolios |
Use AI to suggest organisation and presentation options for the portfolio; |
Digital stories; present 3-min summary of portfolio in class |
May pose some accessibility / equity challenges |
|
In-class paper and pencil exam |
Moderate-low |
In-class exams with no access to digital devices may be thought of as secure in the short term, but wearable AI connected devices are rapidly changing that |
Use AI as a tutor before the exam to generate practice questions based on course content and get feedback on answers |
Group exams – students complete exam individually then share with small group of students and explain their answers, then individuals can adapt their answers but must keep the original as well |
Likely to have similar accessibility concerns as the original |
|
Collaborative annotation |
Low |
Social annotation within a document and in response to others is not currently easily achieved by AI |
Use AI as one of the respondents |
Group summaries of a reading in class; concept maps; |
May pose some accessibility / equity challenges |
|
Poster presentation / gallery walk |
Low |
Students may be able to use AI in the content creation, but the construction and explanation during the gallery walk are individual and unaided |
Use AI to edit text, create images, summarise data, suggest a catchy title, or suggest an outline of the poster |
Create poster outlines in class on large sheets of paper; |
Likely to have similar accessibility concerns as the original |
|
Oral viva/exam |
Low |
Students may use AI prior to the meeting with the examiner to summarise or test themselves on concepts, but are generally unaided during the exam. Wearable or implantable devices may change this in the future. |
Encourage students to use AI as a coach prior to the exam (works better if the students have an AI tool trained on the course content) |
Written exam in class; presentation summarising core concepts of the class; digital story summarising the class |
Likely to have similar accessibility concerns as the original |