Talking with Your Students about AI

Content

Clear communication with students about AI policies is essential. An individual student may be in 5 different courses that have 5 different policies around AI use. Some may explicitly encourage (or even require) regular interaction with AI while others may ban it altogether. As an instructor, it is your role to ensure that students know what the expectations are every time they start working on assignments for your course.  

Throughout this resource, we will encourage you to think about AI through the student lens. How you communicate your policies and educational values is in some ways even more important than what policy you set to begin with. In this resource, you’ll find information on: 

  • Considering AI policies from the student perspective 
  • How and when to talk to students about AI 
  • Activities and assignments that develop critical AI literacy

Student-AI-Perspective

Considering AI policies from the student perspective 

As an instructor, you have likely had many experiences where you thought you had created crystal-clear guidance for students, but it turns out they had a radically different interpretation than intended. This happens a lot with AI policies as well – people think they’ve crafted something that conveys to students exactly what they can and cannot do, only for students to accidentally end up violating the policy.  

Before moving forward, we encourage you to try this activity. It lists a typical AI policy at the top, although we encourage you to instead use your own course policy for the activity to provide you with more useful information. Then, it provides a range of different circumstances where students might use AI throughout the course. Your job is to determine whether the instructor is intending to prohibit or allow AI use as well as whether a student would likely interpret the policy as prohibiting or allowing AI use.  

As you complete the activity, think about these 3 guiding questions: 

  • What items seemed most unclear? 
  • Where might there be different interpretations between students and instructors? 
  • What are the implications of those differences? 
  • What does this tell you about communicating with your students about AI? 

When doing this activity, the big thing people quickly notice is that the term “AI usage” is much broader than they ever imagined. There are so many tools available that can be used in very nuanced ways. Also, “completing an assignment” has a lot more steps than most instructors may initially consider, which significantly expands the possible ways in which students could use AI.  

So, what does this mean for an instructor? Although we recommend a clear and detailed policy, the answer here is definitely not “create an AI policy that’s 10 pages long to encompass every possible way in which students could use it”. Instead, you should identify the places where there is most likely to be student misunderstanding of your policy and supplement the written policy with examples and discussions in class. You might also consider adding a few sentences of explanation at the top of your assignment instructions clarifying which AI tools might be relevant and how students can vs can’t use them.  

When-to-Talk-to-Students-about-AI

How and when to talk to students about AI 

Students frequently say that they’re overwhelmed by AI expectations and wish that instructors would talk about it more. Instructors are used to talking about academic integrity briefly on the first day of class then never bringing it up again. That worked because across classes, there was a general consensus on what constituted academic misconduct. AI has changed the situation because now different courses have radically different policies around which behaviors are considered an academic conduct violation. This makes it essential that instructors talk with students about AI not just at the beginning of the semester, but at important points throughout. 

This section provides guidance on how to approach AI at the start of term as well as suggestions for maintaining the discussion throughout. 

Beginning of the semester 

You’ll want to think carefully about how you integrate AI into the broader conversation about academic integrity. Tone maters a lot in terms of whether students see you as a trusted partner that wants to help them learn vs an adversary creating stringent rules to be ignored or tested.  

  • State your values. Use your educational philosophy as a starting point for the conversation. The goal of your AI policy is to ensure that students are meeting learning objectives and being prepared for what comes after your course. By stating this up front, you position yourself as a partner in their education rather than someone enforcing rules for their own sake. 
  • Avoid being accusatory. Students are starting to feel a sense of “guilty until proven innocent.” Saying things like “everyone is constantly cheating” only makes this worse because it tells students that you assume from the beginning that all of them are going to engage in academic misconduct. If that’s the case, then there’s no reason for them not to cheat.  
  • Make it a conversation. Ask students about their experiences and engage with genuine curiosity. Talk about what policies they’re encountering in different classes and how that impacts their approach to their education. Ask what AI tools they use and how they find them useful vs problematic.  
  • Use an activity. You could use the one described above in the Considering AI Policies from the Student Perspective section or something that asks them to compare and contrast policies across their courses. You may even want to develop something tailored to an early assignment in your course – give students the instructions, your AI policy, and a range of possible situations involving AI use, then discuss what would be appropriate vs inappropriate.  

Throughout the semester 

Don’t assume that students will remember the conversation you have on the first day of class. They’re having similar conversations in all their other courses and it can be really challenging to remember which policy discussion fits with which course. Instead, frame your AI policy as an ongoing discussion that will last the full semester. 

  • Use AI literacy assignments. These can be helpful in underscoring to students why it is important for them to learn things instead of using AI for everything. See the following section for examples. 
  • Be careful if any of these assignments ask students to use AI in a way that would otherwise violate course policies. For example, some instructors that ban AI will have in-class activities using it to intentionally highlight its problems. This can be really confusing for students, so you’ll want to pair those activities with a clear statement that they are only allowed to use AI for specific activities, not other coursework. 
  • Re-state your policy on each assignment. This may seem like unnecessary hand-holding, but it’s very useful for students and reduces the probability of confusion. This can also be a good space to provide more information about different tools they may encounter that are embedded within other programs and not clearly labeled as AI help.  
  • Have a discussion after the first assignment. Many instructors find that, despite their best efforts, students still engage in inappropriate AI use on the first assignment. This can be an opportunity to re-open the AI conversation. You might even consider saying something like “this one time, I will allow anyone that inappropriately used AI to re-do the assignment, no questions asked. After this, all violations will be submitted to the conduct board.” 
  • Lead by example. When you use AI for a task relevant to your course, talk with your students about why it was an appropriate problem for using AI, how you prompted it to get a useful response, and what steps you took to verify or edit the output before using it.  
  • Follow your policy. A main reason students use AI is that they believe nobody gets in trouble for it. When you find violations, it is important to follow the university’s academic misconduct process. 

Developing-Critical-AI-Literacy

Developing Critical AI Literacy 

An important part of talking with students about AI is helping them understand how it actually works and what potential problems there are with using it. Many students see AI as no different from any other technology and inherently trust the results. While student skepticism is increasing, not all students have had explicit instruction in what to look out for when it comes to AI  

There are a wide range of course activities you can use to help students develop AI literacy. We recommend you use something that is most relevant to the ways that students are likely to use (or misuse) AI within your specific course or discipline. You’ll also want to think carefully about the goals of the assignment since there are a number of nuanced aspects to AI literacy.  

Whether you end up using one of the activities below or something else, the most crucial part of the experience is engaging in robust discussion and reflection. In response to class time being limited, instructors will sometimes rush the reflection element or move it to an out-of-class format. We highly recommend taking the time to do this part in class to ensure that students fully engage in the conversation. This is to ensure that students grasp the essential points you intended for the activity to demonstrate. (It is also partially because many instructors are finding that students offload reflection activities to AI tools, thus foregoing the entire point of the exercise…)

How AI works  

Many students treat AI as being the same as any other technology, but it functions very differently, which has an impact on its utility. To help illustrate this to students, consider using an activity highlighting the similarities and differences between generative AI tools and other common technologies such as search engines and calculators. 

Student Activity: Examination of AI Types

A possible activity for this is to have students fill out a table like the one below, changing the labels on the axes to fit the types of technologies they’re likely to use in your course.  

Feature AI Chatbot Search Engine Calculator Traditional Software 
Follows fixed rules     
Generates new content     
Always produces the same answer to the same prompt     
Learns from data     
Requires explicit programming     

 

After filling out the table, have a discussion about the types of tasks you’ll be completing in your course and whether each feature is positive or negative. For example, in a math class, it’s likely unhelpful for a technology to produce different answers from the same prompt. However, this could become an asset in a business class where you’re having students develop new business ideas.  

You may also want to illustrate some of these concepts by having students use different tools to answer the same questions and compare results. These could include both fact-based questions and higher-level tasks like “compare and contrast” or “analyze the strengths and weaknesses of …” For this activity, you’ll want to test the tools ahead of time to ensure they produce data useful for illustrating your point. 

Academic Integrity 

AI has had a significant impact on academic integrity. In many national surveys, there is a significant proportion of students reporting using AI in violation of academic integrity policies. When developing course activities around this topic, don’t think of it as purely understanding course AI policies. Instead, this is an opportunity to teach students about ethical decision-making more generally and to consider what factors might be involved in bringing about academic integrity violations.

Student Activity: Academic Integrity

As you think about activities for this, we recommend first reading our Course Policies and Practices that Encourage Integrity resource. This frames academic misconduct as the result of otherwise ethical people making a momentary decision to act unethically. In that resource, we discuss course policies to help combat academic misconduct. In a classroom activity, the goal is to help students understand those same forces and consider how they might alter their workflows to avoid making unethical choices.  

To start this activity, you might want to give students an anonymous survey with questions like those below. Note that in very small courses where anonymity cannot be maintained, you might want to skip the first question and just talk in terms of what they’ve seen from other students.  

  • Have you ever used AI in a way that violated your course AI policy? 
  • Have any of your friends ever used AI in a way that violated their course AI policy? 
  • If either of those answers was yes, why did the misuse of AI happen?  

When looking at the survey results, you’re likely to see something resembling the list below, which comes from several research studies on academic misconduct. If you’re uncomfortable using the survey described above, you can start by presenting students with this list instead. 

  • Learning is hard, but AI is easy. 
  • Not understanding AI policies. 
  • Not understanding that the tools they use are AI 
  • Not understanding assignment instructions. 
  • Running out of time. 
  • Not seeing the value of an assessment. 
  • Fear of failure. 
  • Fear of reaching out to the instructor. 
  • Believing other students are cheating. 
  • Believing they won’t get caught.  

The first thing to notice about the list is that none of the factors are really inherent to the individual. Instead, they are based on situational factors that impact decision-making in a specific moment. The other thing to notice is that other than the first few, they’re not even about AI itself – they’re actually the same reasons that have always shown up in cases of academic misconduct. This means that AI isn’t as much a completely new problem as it is a technology that has made pre-existing problems worse. 

The activity for students will then involve asking them to go through the list above and rank them from “I am very likely to use AI for this reason” to “I am very unlikely to use AI for this reason”. Then for their Top 5 Risk Factors, have them identify specific strategies they might implement to avoid engaging in inappropriate AI use. (You might want to encourage them to work through the whole list, although that may be prohibitively time-consuming to do in class).  

For example, if their biggest risk factor is running out of time, they may want to devise and enforce a strict study schedule. If it is not understanding assignment instructions, they may want to set aside time to begin working on assignments during your office hours to ensure they have the opportunity to ask questions.  

We recommend ending the activity with a class-wide discussion since some of the items, such as believing others are cheating or that they won’t get caught, may require instructor explanation of how you set up course practices to reduce AI misconduct and how you will be treating violations of your course policies.  

Demonstrating bias 

AI models were trained using information freely available on the internet with little editing. This means that AI output suffers from all the biases present in the training materials. We provide some general example activities demonstrating AI bias below, but when tailoring them to your course, it is important to start by thinking about what biases exist within your field that are likely to be present within AI-generated output.  

Student Activity: Demonstrating Bias

An important aspect of bias often shown in AI output is the exclusion of particular people and perspectives. One way to help illustrate this phenomenon is to have students generate a list of key figures in your field (both historical and present-day). You can then have them categorize the list based on factors such as: 

  • Gender  
  • Race/ethnicity 
  • Nation of origin 
  • Historical period 
  • Political leanings 
  • Perspective within the field 

An alternative way to approach this activity is to have students use AI to do an investigation of a specific topic within your field that is presently unsettled with multiple potentially valid perspectives. When using this option, you’ll want to ensure students ask for specific sources supporting each perspective so they can also examine potential biases in sources rather than just the output.  

With either approach, it is recommended you have students use different AI tools (ChatGPT, Copilot, Claude, Gemini, etc.) since each tool is trained differently, and biases can show up in a variety of ways. You’ll also want to run a test of the activity on your own so you have an idea of the possible outputs ahead of time and can choose instructions and topics that best illustrate bias to your students. 

After collecting the AI output, you’ll want to have students discuss questions like: 

  • Which groups/perspectives are over- vs underrepresented relative to their contributions to the discipline? 
  • Are there specific figures/perspectives missing from some or all outputs? Is there a pattern connecting them? 
  • Is the output particularly focused on any period of time? 
  • How do these patterns fit with known bias within the field? 
  • What does this tell us about using AI within our field of study? 
  • What steps is it important to take to mitigate the impacts of this bias when using AI tools?

Misinformation 

The generative nature of AI tools results in output that often contains incorrect information. The degree to which this happens is dependent on the topic and format of the query. AI output is created probabilistically, meaning it is constructed based on “what is the most likely next word in a sentence” rather than “what is the correct answer to this question?” This can result in incorrect responses to even simple questions like “how many R’s are in the word strawberry?” Examples like this may seem entertaining and harmless, but they actually illustrate important lessons about the need to always verify AI output rather than taking it at face value. 

Student Activity 1: Misinformation from Neutral Prompting

The goal of this type of activity is to get students to understand the implications of AI hallucinations and misinformation for their own AI usage. To set up this activity, you will want to choose a topic that your students already know a decent amount about. You’ll also want to create a prompt that produces somewhat lengthy, detailed output to work with. If your course involves finding and citing external sources, you’ll also want to include a directive to provide citations. You may want to have some students ask for direct links to citations while others leave out that directive to illustrate that hallucinations are more common if you don’t ask for links.  

A challenge with this type of activity is that AI output is highly variable: sometimes it might produce completely correct information, and other times have significant errors. This is why it is essential that you test your prompt shortly before the class period you conduct the activity. You should also have students use a range of different AI tools (ChatGPT, Copilot, Claude, Gemini, etc.) and even compare the free vs paid versions of tools if you have access to paid accounts.  

After students obtain their individual output, they should go through and note: 

  • Is each factual sentence correct vs incorrect 
  • Is each citation real vs hallucinated? If real, does it actually provide evidence for the claim it is linked to? 

In small groups or as a full class, have a broader discussion about the implications: 

  • Which specific claims/types of information were most likely to be accurate vs incorrect? 
  • Were any specific AI tools more or less prone to factual errors? What about falsified references? 
  • How would you rate the quality of references used? Are they of a high enough standard that they would be acceptable for coursework?  
  • This is a topic where we already know a lot. Have any of you used AI to research a topic you were unfamiliar with? What are the implications of that?   
  • What steps can you take when using AI to ensure accuracy? 
Student Activity 2: Misinformation from Biased Prompting

For the previous activity, we set up a prompt that is factual and based on settled evidence. An additional demonstration you might try is to push the AI to take a specific perspective. The goal here is to demonstrate that AI tends to be sycophantic and produce a response in alignment with the beliefs of the person using it. It is therefore important to be very careful with your prompts if you want the response to produce academically validated results. 

To demonstrate this, take a topic that is either unsettled within your field or settled by academics but with significant existing misinformation on the internet. Break students into groups and have each group write 3 different sets of prompts: one from each of the conflicting perspectives and one with a neutral ask for evidence for both claims. Make sure they ask for sources to back up all claims. You might also want to ask different groups of students to use different AI tools, but you’ll want to ensure that within a single group, they use the same tool for all 3 prompts to ensure consistency.  

After running the 3 different prompts, students answer questions like these: 

  • When asking for evidence from a specific perspective, did the output include any discussion of the opposing perspective? 
  • How does the quality of evidence compare across the 3 prompts in terms of how often they used real sources rather than hallucinating them and how much they cited peer-reviewed academic sources rather than other websites? 
  • Examine the evidence in the “neutral perspective” output: is it truly neutral? If not, does it lean more toward the consensus view from academics in the field or the alternative view?  
  • Imagine you are someone completely new to the field and you’re trying to figure out what’s most likely to be true. Did the neutral prompt actually accomplish this? Is it possible to write a prompt that guarantees this kind of output?  
  • What does this tell us about using AI for different kinds of tasks?  
  • What steps can you take when using AI to ensure accuracy? 

AI vs human competition 

Often, students turn to AI tools because they trust them more than they trust themselves. It can therefore be helpful to set up course activities that demonstrate the knowledge and value that their perspective brings. Depending on the situation, these activities can either be used to demonstrate that human work is superior on its own, or that collaborating with an AI tool results in higher-quality output than either could do individually.  

Student Activity: AI vs Human Competition

The starting point for this activity is generally a prompt developed by the instructor that is very similar in structure and content to what students will likely encounter in their coursework. It should be something where there is a wide range of possible responses rather than purely factual – for example, asking students to apply a specific concept from class to a real-world scenario. You’ll want the responses to be relatively short (just a couple of paragraphs) since students need enough time to complete the activity on their own. You might also consider using this as a way of collecting writing samples from students that you can later use to investigate instances of potential academic misconduct!

 In class, you’ll have students write out their answer to the prompt, then put the prompt into an AI tool. We recommend doing the two tasks in this order to ensure that the human-generated response isn’t influenced by reading the AI output. You might also want to ensure that different students use a range of AI tools since the output may be slightly different, and this can help determine which tools may be of best use for your course (if you choose to take the activity in that direction).  

Once both sets of writing have been created, you can either put students in small groups or have a full class discussion about what they notice. Some possible topics of discussion include: 

  • Look at the AI-generated responses. What patterns do you notice? How similar are they in terms of writing style, quality, topic, depth of analysis, novelty etc.?  
  • Look at the human-generated responses and answer the same set of questions.  
  • What are the similarities and differences between the human vs AI responses? 
  • What is better about the human-generated response? The AI-generated response?  
  • If you were an instructor, which set of responses would you rather read? Why?  
  • AI writing tends to be relatively generic – if you read a bunch of responses to the same prompt, they all sound pretty much the same even if the details differ. What are the implications for this if we think big picture about how much humans are relying on AI when creating published writing? 
  • What implications does this have for your own writing? How will you ensure that, if you do use AI, the final product maintains your own voice and perspective?  

Depending on the goals for the activity, you might leave the conversation there or take an additional step of having the student see if they can work with the AI to create a collaborative product that’s better than either original document.  

The first step in that process is making an important decision: do you start by having the AI produce feedback/edits on the human writing, or by having the students edit the AI writing? You might turn that decision into a discussion with students about which approach is most likely to result in the student keeping their own voice and perspective. If time allows, you might want to use both of these tactics to demonstrate the differences between where they end up.  

Then, you’ll give students time to engage in an iterative process that uses AI to provide feedback on the human-generated writing, and the student makes edits to the AI-generated writing. You’ll want to structure this in a way that makes sense given the type of assignment and writing conventions for your discipline.  

When making edits to the AI output, the student will want to make edits like: 

  • Ensuring the voice of the writing sounds like them 
  • Ensuring that the ideas presented reflect their views and interest in the topic 
  • Ensuring the writing avoids misinformation and bias 
  • Ensuring all references are real and appropriate 

When asking the AI for feedback, the student may want to use prompts like: 

  • Fix all spelling and grammar mistakes 
  • Ensure all conventions related to the specific writing style are correct 
  • What are the 3 things a reader might be most confused about? 
  • What feedback do you have regarding my thesis statement and the quality of supporting evidence for it? 
  • Which arguments are the weakest? Provide suggestions for how they might be improved 
  • Provide suggestions for how the introduction could better draw in readers and set up my argument. 

After running multiple editing iterations, have students compare the finished draft with the original and reflect on questions like: 

  • In what ways is the final draft similar to the original? In what ways does it differ? 
  • Which seems better? Why? 
  • If a rubric is available, have the student score both versions to see which would receive a better grade. 
  • Does the final draft feel “yours”? In other words, does it read like something you would have written and reflect your ideas/values? 
  • What did you learn from this exercise about working with AI to improve your writing? Where is it useful and what are some potential problems to look out for? 

Other-Ethical-Concerns

Other ethical concerns 

There is a wide range of other ethical concerns regarding AI (environmental impacts/privacy concerns/copyright / etc.) that you may want to integrate into your course. While we don’t provide specific examples due to the wide variety of topics, you can use many of the above examples and tailor them to these use cases. For example, applying the Misinformation or Bias activities could help students understand the ways in which AI tools produce inaccurate information regarding some topics. You could also use the How AI Works activity to have students investigate these topics using AI, a traditional search engine, and an academic database to compare the results.  

AI-Student-Attribution

Assistance in Developing Course Material  

If you would like help developing a plan for communicating with your students about AI or assistance with anything else related to AI in your course, please reach out to instructional designers designated for your college.  

If you have any feedback about the content of this page or ideas for additional content to include, please reach out to the page developer Amy Ort (aort@unl.edu).