Adapt Assessments to Mitigate Inappropriate or Unauthorized Use of Generative AI

Person using device and speech bubbles with warning icons appearing around icon of human face with the letters AI

The integration of GenAI provides an unprecedented opportunity for faculty to reimagine how to assess student learning. Faculty can create unique, authentic and scaffolded assessments that provide students with multiple, and even personalized options for demonstration of learning.

Offering different assignment options can help to support inclusive learning environments. Offering students a choice in assessment strategies can be helpful in supporting diverse learners and aligns with Universal Design for Learning guidelines by providing multiple means for students to express their learning.

In this section

Authentic assessments

Ask yourself: ‘What am I, or professionals like me required to do in our role? What knowledge and skills would a graduate be expected to bring to the job? What are key requirements for successful work in the field?’ Then develop a similar task, or a series of scaffolded tasks, in alignment with your course and program learning outcomes. For example, if conducting interviews is part of the job, have students role play, record their interview, and watch it back to provide their own assessment of and feedback on their interview techniques. If they are required to write reports, create personalized scenarios and have them complete the same type of reports in response to the scenario.

Ideally, authentic assessments:

Mirror realistic tasks
Present a cognitive challenge
Require students to apply learning in a novel or creative way, systematically evaluate something to render a judgement, or assess a situation to devise an alternative approach or solution.

Tip

For ideas, see Alternative Assessment Worldwide, the Authentic Assessment Toolbox, or the University of Queensland Assessment Ideas Factory.

Focus on the process, not the product

Designing personalized authentic assessments that GenAI cannot complete effectively is important. Jason Lodge (2023) from the University of Queensland reminds us that:

While generative AI can increasingly reproduce or even surpass human performance in the production of certain artefacts, it cannot replicate the human learning journey, with all its accompanying challenges, discoveries, and moments of insight. It can simulate this journey but not replicate it. The ability to trace this journey, through the assessment of learning processes, ensures the ongoing relevance and integrity of assessment in a way that a focus on outputs cannot.

Requiring students to submit something that reveals the process they followed to create that product can help reduce the likelihood that the assignment was created by GenAI. For example, you can view the editing history in Google Docs or Word, or students can explain in videos using Microsoft Flip or Canvas Studio.

Adjust instructional and assessment strategies

To mitigate the use of GenAI, consider adjusting instructional and assessment strategies by:

Bringing assignments, or parts of them, into the classroom

  • While completing all assignments in class may not be feasible, integrating specific elements of assignments can help you monitor student work more closely.
  • You can ensure that students are developing their critical thinking and writing skills while promoting academic integrity.
  • This approach will also allow you to offer feedback as students are developing their work, fostering a more engaging and collaborative learning environment.
  • This rubric can help faculty determine the potential for GenAI misuse in writing assignments to tailor assignments to be more AI resistant.
  • Use peer and/or self-feedback during specified points in the assignment process, and require submission of the assignment versions and feedback.

Designing non-textual and alternative assessments

  • Ask students to respond to non-textual resources, such as images, diagrams, or videos.
  • Allow students to express their learning using methods other than writing (e.g., a mind map, a timeline, an infographic, a video, etc.)
  • Students could also be asked to do presentations in class with Q&A. Though they might use an AI text generator to develop part of what they present, they will still need to understand the material enough to effectively present it and answer questions from other students.
  • Consider issuing two-stage exams.
  • Have your students create open educational resources such as videos or text.
  • Try peer assessment.

Incorporating more disciplinary, situational, and individual-based questions

  • Ask students to discuss their own individual experiences or views on course topics, or to provide a specifically disciplinary or course-informed response to real or fictional case studies.
  • Connect assessments to specific points discussed in class, on discussion boards, and the like, or to other courses students may have taken before.

Consider Bloom's Taxonomy

Consider Bloom's Taxonomy as a reference for evaluating and making changes to aligned course activities and assessments that account for Generative AI capabilities.

Screenshot of Bloom's Taxonomy Revisited
Bloom's Taxonomy Revisited by Oregon State University Ecampus. This image is CC-BY-4.0.

Prevent AI-generated work

  • Require students to share their process or submit “in process” portions of academic work for feedback.
  • Submit your assignment prompts and questions to ChatGPT yourself to become familiar with the responses and the writing style of the output.
  • See this AI misuse rubric for other ideas for tailoring your assignments to be more AI-resistant.

Due to differences in writing style, you will, most likely, be able to tell if a student uses AI-generated text the same way you can with other plagiarized text. There are emerging tools that can detect AI-generated text; however, they are unreliable, often producing inconsistent results for the same piece of work. They can be very easy to beat, may have significant false positive rates and may collect and use student work and data.

Here are some suggestions for what to do when you suspect AI-generated text, without relying on GenAI detection tools:

  1. Check references and sourced information. Are the sources accurate and relevant? GenAI may create false sources of information or scholarly works.
  2. Compare the assignment against other samples of writing and academic work from the student. Is the language and writing convention consistent?
  3. Open a discussion with the student. Ask them to explain or expand on specific pieces of information in the submission. How did they arrive at an idea or conclusion? How did they go about locating their references?

References