Artificial intelligence programs, once the stuff of science fiction, have become broadly available to students and instructors alike. Although the effect of such programs on learning and skill development are unknown at this time, students are already using AI technology in many assistive capacities, and sometimes in ways that might be considered academic misconduct. It is therefore important that all instructors consider the availability of AI when designing their courses and assignments.

What do we mean by AI? #

AI comes in many forms, from assistive programs that help to polish and perfect our work to generative programs that create new content. 

Applications that allow us to check grammar and spelling are already incorporated into approved Microsoft products, and are used by faculty and students alike. Grammarly, a more sophisticated stand-alone platform, performs similar checks and helps writers modify the tone of written work to increase impact. This software is approved for use in some classes, but is not available for download through the self-service platform.

At the time of this writing, the only UWSP approved options for generative language programs are those that are covered under the university’s Microsoft contract. The Bing browser, when run on Microsoft Edge, has a chat function that combines the large language model generative AI algorithms of ChatGPT 4 with the internet search capacity of Bing. Unlike the stand-alone ChatGPT applications, the chat function in Bing will include web references in the essays it generates. There are some limitations on the functionality at this time, since conversations in the chat cannot exceed 10 questions. Microsoft has announced that the same chat algorithms will be incorporated into Word before the end of the calendar year, and possibly before the beginning of the Fall 2023 term.Image generating AI is also available to faculty and students. Currently, Adobe Creative Suites is offering a beta version of Adobe Express, which has text-to-image functionality. Microsoft Edge also has an image generator. Although it is not currently available through institutional accounts, it will likely be integrated into the platform for all users soon.

Options for addressing AI use in your courses #

There are three basic strategies for dealing with AI in your courses. You should choose the option that is best suited to your context.

Be clear about your expectations for student work #

Students are more likely to meet your expectations when you have taken the time to make your expectations clear to them. In addition to discussing your AI policies during your class, include a statement in your syllabus that outlines your policy on AI use. Two examples can be found below..

• “Since writing, analytical, and critical thinking skills are part of the learning outcomes of this course, all writing assignments should be prepared by the student. Developing strong competencies in this area will prepare you for a competitive workplace. Therefore, AI generated submissions are not permitted and will be treated as plagiarism.” Offered freely by Jill Hogan, Higher Ed Discussion of AI Writing Facebook Group.
• “One goal of this course is for you to work on developing the discipline-specific writing skills that you will need to be successful as a professional in this field. I want to acknowledge that recent buzz about ChatGPT and other generative AI tools poses some interesting questions about the need for developing these skills, and how such tools can be used in higher education. Given that this technology is still in its infancy and that my goal is for you to develop your skills as writers, the unauthorized use of ChatGPT or other AI writing tools, is not permitted in this course. Students found to be using such tools will be considered as engaging in conduct aimed at making false representations of a student’s academic performance, and will be subject to disciplinary action as defined in the UWSP Academic Misconduct Policies.”

AI-proof assignments #

The available evidence indicates that students are relying on AI tools when completing assignments, including both assistive AI and generative AI. The sad news is that there is no way to prevent them from doing this. The key to preventing AI use on assignments is careful assignment design, in line with evidence-based best-practices for teaching and learning. The following recommendations are provided to guide you in your assignment design.

Provide context for students by making the purpose of the assignment transparent #

When students don’t value what you are asking them to do—when they don’t see or understand how the process is connected to the acquisition of knowledge, or the formation of skills, they are less likely to engage in the assignment with the level of commitment that you’d like to see. Making the learning goals of the assignment clear to the students can help to improve their valuation of the work, and will lead to less overt cheating with AI. 

Carefully consider the alignment of the assignment with course learning outcomes #

In many cases, we provide writing assignments as a means of evaluating student achievement of learning outcomes. If the learning outcome is not specifically tied to writing skills, it may be that some other form of assessment could be substituted. This may be a presentation, a podcast, or a proctored exam. For learning outcomes that reflect the knowledge and understanding levels of Bloom’s cognitive taxonomy, standard exams may provide a more accurate reflection of student’s ability to recall facts, definitions, and provide simple explanations and descriptions than take-home papers that allow the students to (at worst) submit AI generated content, or, verify their answers prior to submission.

If the learning outcome in question is tied to writing skills, consider extemporaneous demonstration of writing skills. In days of yore, blue-book exams were often used to assess students’ ability to write. Such assessments may be a more accurate reflection of students’ unaided skills, as they will not have access to AI assistance as they compose their responses to prompts. Although this type of exam has been replaced through the years with take-home writing assignments to limit student stress, it is possible to reduce the stress of these exams through exposure. Consider including smaller-stakes extemporaneous writing assignments prior to summative assessments using this modality.  

Scaffold writing assignments to emphasize the writing and research processes, not the final outcome #

Like Athena emerging from the forehead of Zeus, generative AI spits forth essays completely formed. Reworking your assignments to emphasize steps in the preparation of the final essay or paper may make it more difficult for students to substitute the work of AI for their own.

1. Research projects often begins with brainstorming. Consider making this an assignment. Ask students to reflect on several different ideas or approaches to the final work, weighing the merits of each.
2. Brainstorming is usually followed by research. Having students specify articles and books that might be useful sources for their paper/project in advance of drafting their papers makes it very difficult to get AI to generate a complete paper based on the reference material using a prompt. Because AI is known to “make up” sources, it is important to ask students to provide links to the papers on the journals’ online site. There may be modest benefit in requiring that some proportion of identified sources be recent, since the large-language training models like ChatGPT were not trained on recent materials. (Caveat: Bing’s chat feature has access to just about everything on the web, though it is not clear how thoroughly it integrates content from recent sources into essays it produces.)
3. Drafts are a great opportunity to provide feedback designed to help students improve their performance before a final revision. Allowing students to provide peer feedback will help them strengthen their own writing. Prioritizing the students’ reflections on the feedback they are given and how they plan to use it to improve their final work puts the emphasis not on the final product but on the process of producing that product. Large Language AI models can’t do this.

Flip your classroom to incorporate more extemporaneous classroom activities. #

If writing assignments are intended to help students digest or apply specific content knowledge, consider replacing them with other, ungraded, classroom activities that allow students to digest and discuss readings. Flipped classroom activities are often more engaging for students than traditional lectures. Some options include:

1. Jigsaw discussions in which students prepare for class by reviewing different content. Working in groups during class time, each student will bring different information to the discussion, and through sharing what they know with peers, will help one another to broaden their understanding. Generative AI is of little use for this exercise.
2. Case studies. Students work together to apply content knowledge to specific cases. Because nothing is written, grading is reduced. Students still get practice and receive feedback on their performance prior to assessment. The extemporaneous nature precludes the use of AI to find a “correct” answer.
3. Answering questions about the reading in class as a group. Working with peers, students build a shared understanding of complex topics by discussing and finally composing communal answers to questions about the reading. Though students may sneak AI into the mix, your supervision of the discussions, and questioning of students during the process may limit the utility of the AI.

Include mandatory oral defense of papers as part of your grading rubric #

In days of yore, before word processing, computers, and plagiarism detectors, one of the largest threats to the academic integrity of student papers was file caches held by student organizations (e.g. the “fraternity paper files”) that provided a platform for students to submit re-typed copies of old papers. The only way for professors to detect this type of misconduct was to interview each student about the paper they had submitted, probing the students’ knowledge of the submission, the topic, and the references within the paper. This technique, though time consuming, also holds promise in the age of AI. It cannot prevent the use of AI by students, but if incorporated into grading rubrics, will provide extra accountability for the knowledge you are attempting to assess, and may serve as a deterrent to unfettered use of generative AI to write papers. 

This approach to verifying the authorship of papers may be more equitable than questioning only students whose submissions strike instructors as sketchy, as it prevents instructors from beginning with biased assumptions about the credibility of any paper submitted. By instituting a level playing field, there can be no inadvertent targeting of specific students or demographic groups for enhanced scrutiny. 

If an oral defense meeting leads to suspicion of academic misconduct, follow the procedure for dealing with academic misconduct cases outlined below. Because students are afforded legal rights to have an advocate and to defend themselves against any charges of misconduct, the oral defense may not be used as a misconduct meeting.

Incorporate AI into assignments #

Because it is likely that students will use generative AI assistance in their assignments anyway, it may be useful in some contexts to include the use of AI prior to or as part of an assignment. If you decide that this is the best option, please make sure that the use of AI is aligned with or reflected in the learning outcomes of your course/lesson, and approved for use at UWSP. Some options for incorporation are presented below.

Discuss the benefits and costs of using AI as part of assignment completion #

AI may be a useful shortcut for some work that is more meaningful, but the AI output is only as good as the algorithms and training material allow it to be. Making it clear to students where the weaknesses of AI are as they pertain to your discipline will help students contextualize the use of this technology.

Demonstrate the limitations of AI generated content #

A useful exercise is to require students to critique AI generated content. The specific areas of critique will vary from discipline to discipline and assignment to assignment, per your experience with the accuracy of AI tools in your discipline. Some of the areas to focus on are content accuracy, potential bias, missing information, attribution, and writing style. This is an excellent way to introduce students to the importance of information literacy and actual content knowledge, as it is impossible to perform such a critique without having the knowledge, understanding, and writing skills that most writing assignments are intended to assess.

Include AI drafts of assignments as part of process work. #

Allow students to draft some of their materials using AI. Using criteria similar to those outlined for demonstrating the limitations of AI generated content, students may critique these drafts, using them as a starting point for further work. Because AI can fabricate content and references, validating content and providing references from the literature is a useful exercise.

Include reflection assignments about errors in AI output and changes made. #

If AI content is used as a starting point, it is important to emphasize as part of the assignment critical evaluation of AI content. Have students keep track of errors they have found and changes they have made, explaining their reasoning in all cases

UWSP Approved Technology #

Some AI applications are approved for purchase and use in classes, but others are not. Approved software must not pose information security concerns or potential threats, must be supportable in the campus computing environment, must work with other reviewed technologies, and must be available through UW-System purchasing practices. Although students may use unapproved applications outside of class of their own volition, instructors cannot compel students to do so. This document presents the UWSP policy for use of non-approved information technology applications, but in general, it is best when planning assignments to use only approved software. Although the first section of this document provides a summary of what is available, you can check whether specific applications are approved for use with the UWSP software catalog. You may also apply for approval of new software you’d like to use. 

Dealing with Academic Misconduct associated with AI use #

Many instructors view the use of generative AI in assignments as a form of academic misconduct. Although Turnitin, the antiplagiarism software used at UWSP offers AI detection, this feature is not reliable. It provides a percentage estimate of the likelihood that content was generated by AI, but has been known to flag work that has used predictive text AI, including text reworked by grammar-correcting software. Because this type of AI is generally not forbidden on assignments, this software can’t be used as a definitive AI detector or as sole evidence of misconduct.

If you suspect that a student has used generative AI inappropriately, and that the use may constitute academic misconduct, do not succumb to the temptation to impose a penalty without further inquiry. The student has legal rights in such cases, so it is important to follow the approved process for handling suspected academic misconduct.

1. Request a meeting with the student using the meeting request template provided by DOS which outlines the student’s legal rights. You must also complete the academic misconduct reporting form available on the Dean of Students (DOS) web page.
2. When you meet with the student, approach the situation with an open mind. Think of it as both a fact-finding mission and a teachable moment in which you can engage the student in a conversation about the suspected misconduct. During the meeting, you will assess responsibility and determine whether/what sanctions are appropriate. It may be helpful to ask the student questions about the paper and the material you suspect the student has used generative AI to produce. This meeting is also a venue for students to explain their perspective and present evidence of their innocence.
3. Reach a decision about whether misconduct occurred, and what penalties against the student are warranted. 
4. Inform the student of your decision and their right to appeal that decision using the Academic Misconduct Outcome letter template. The DOS office should also be informed of the outcome of the case.

The academic misconduct process can be time consuming. It is better to avoid the potential of misconduct by careful assignment design. If you have large numbers of students engaging in AI related misconduct, a reevaluation of assignment design may be warranted. 

Summary #

The landscape of AI is changing rapidly. Adjustments to the way we teach and the way we assess student learning will be required. CITL is available to help you reevaluate your assignments to help make them less amenable to AI use, or to help you redesign your assignments to include AI. Visit the CITL website to schedule a consultation.