Teaching in the Age of Generative AI

Generative AI (genAI) tools are trained on massive data sets (of words, images, videos, and code). They use those data sets to predict, automate, and generate output that mimics human-generated content.

For instance, ChatGPT is an artificial intelligence software that composes answers to user-submitted questions or prompts. Its free version (ChatGPT 3.5 as of December 2023) relies on a database of billions of words (scraped from the internet, prior to 2022) and was trained by humans to generate a sequence of words that mimic human writing and conventions. It can also revise its output, based on clarifying ideas or questions or prompts, adjusting the content or tone or style of its generated answers. The term “intelligence,” however, is a bit misleading. The responses ChatGPT generates often include false or unreliable information (often called “hallucinations”). Additionally, ChatGPT lacks the ability to analyze (since it is just predicting text), and it also cannot provide information about current events (it’s dataset currently maxes out at 2021). 

Image generators, such as Dall-E or Midjourney act in similar ways, trained on high volumes of images that have been tagged with descriptive terms that can then be mimicked and reconstituted to match the words provided in the prompt. Video generators also exist, along with code generators such as Codex or GitHub Copilot turn plain language into code or make predictive suggestions for improving computer programming code.

A good place to start building your understanding of how genAI works is with these three articles, written for popular media. Feel free to share them with your students, too.

• How AI chatbots like ChatGPT or Bard work – visual explainer (The Guardian)
• Generative AI exists because of the transformation: This is how it works (Financial Times)
• AI can now generate images out of thin air. See how it works. (The Washington Post)

Experts and pundits all agree that genAI is here to stay. We are in a period of rapid growth and development driven by tech companies eager to lead the way while governments consider what kinds of regulation are needed. Given the proliferation of tools, public consumption of this technology has been widespread and all sectors of the economy are adopting genAI. Some of these uses of genAI will fizzle while others gain traction. Without a doubt, the world we are preparing students to thoughtfully and productively contribute to will have genAI integrated, which means that faculty in higher education, even if reluctantly, need to become familiar with this technology and its implications for their disciplines.

Students will be forevermore living, learning, and working in a world where this kind of sophisticated technology exists (in fact, we can likely assume this is a baseline from which improvements will come). What do they need to know about the limits and possibilities of it? How is genAI showing up or likely to be used in the professions into which your students enter? While none of us has completely clear answers, a college classroom is a great place to begin the dialogue and raise questions about ethics, bias in AI, developing skills vs using a tool, reliance on technology, quality of output, etc.

Perhaps this is obvious but, nonetheless, worth stating: Your strength, which genAI will never have, is your humanity – your ability to connect with students, to mentor students, to facilitate interaction between students, to provide accountability and contextualized feedback, and to bring expert insight and analysis to a learning environment. These strengths are all foundational to help students learn responsible, ethical, and productive use of genAI (along with knowing when to not use it).

Get (More) Comfortable

Because genAI is here to stay, it is a technology worth developing familiarity with. The CTLI recommends creating an account with one or more tools to begin to understand how they work.

For a very introductory tutorial, use Harvard’s 3-Part AI Guide to learn basics and test ChatGPT.

Ethan Mollick and Lilach Mollick (of UPenn’s Wharton School) created a 45-minute, 5-part video tutorial series titled Practical AI for Instructors and Students. Dr. Mollick and Dr. Mollick clearly wholeheartedly embrace genAI in higher education; you may have more hesitation or concerns about its value or utility. However, we think that this tutorial series is a good place to begin if you are brand new or relatively novice to genAI, with information and suggestions for practicing with genAI.

1. Part 1: Introduction to AI for Instructors and Students
2. Part 2: Large Language Models
3. Part 3: Prompting AI
4. Part 4: AI for Teachers
5. Part 5: AI for Students

You will have to create an account to experiment with genAI tools, and when you create an account (especially with a free tool), there are privacy considerations for any information you submit. Some people choose to create a “burner” email account that they use just for genAI tools.

And most importantly, never upload student work without their permission or materials that would violate FERPA student privacy regulations.

As you experiment with genAI tools, track your questions, reactions, concerns, insights, and ideas. Consider what it means, in your discipline, to use genAI effectively and ethically.

You may even want to experiment with using genAI for teaching-related tasks, such as brainstorming icebreaker activities, writing case studies, or generating drafts of low-stakes quiz questions.

---

Develop and Communicate Course-Level Expectations

It is important to have a course-level policy (either written by you or co-written by you and your students) about genAI usage.

Students will be experiencing a range of expectations across their courses, so it is also important to talk with students about why you are making particular decisions about genAI usage in your course.

Many universities suggest creating a genAI course-level policy that falls into one of three categories:

• Integrated. This course’s generative AI policy acknowledges the use of AI is an essential skill in today’s world. By using genAI for specific purposes, students become equipped with relevant skills and tools necessary to thrive in a technology-driven society. With this type of course-level policy, you will want to be explicit about if/how students are expected to indicate their use of genAI including citations. It is also important to help students learn how to evaluate genAI output for accuracy and validity, particularly as hallucinations (made up information or citations) are common with genAI. Consider whether you want students to include copies of their AI chats with assignments they submit where they used AI for assistance or whether you would like them to reflect on the output provided by any genAI tools.
• Allowed. This course’s generative AI policy acknowledges technology, including generative AI, plays a supportive role in learning and feedback. You may have particular assignments or steps of assignments where genAI usage is encouraged and others where it is banned. Communicate clearly when and why those delineation are made. Similar to an unrestricted use case, be explicit about whether and how students are expected to indicate genAI assistance and use citations.
• Prohibited. The use of generative AI is not allowed in this course, with the exception of spellcheck, grammar check and similar tools. This type of policy will be most difficult to monitor and enforce. It is therefore especially important to explain to students how genAI use would negatively impact their learning, substantiating the policy.

The University of Nebraska-Lincoln’s Center for Transformative Teaching has shared sample syllabus statements in these three categories with similar names.

---

Additional Resources

The CTLI has customized an Introduction to AI Literacy guide as a module that you can import into your Canvas course. This module is distributed under a Creative Commons NC-SA license; please describe any changes you make and include the CC-NC-SA license.

To preview and/or import the module, follow these steps:

You can teach with genAI and/or you can teach about genAI in the classroom. AI for Education has developed resources for educators, including these two infographics that use simple mnemonics to coach students on responsible use of AI and effective prompt engineering. If you are teaching about genAI, these are simple strategies to help students learn to use genAI.

---

Teaching with genAI will likely require modifying or creating new assessments, assignments, and in-class activities. These two compendiums (The Assessment Menu and 101 Creative Ideas Booklet) are two places to explore ideas. There are more suggestions (including discipline-specific) in the Additional Resources section.

---

Additional Resources

An understandable initial reaction to this technology is the concern that students will use genAI to cheat – they will feed it a prompt or question (with some additional parameters or guidance) and submit work they didn’t compose.

AI Detection

Simply put, AI Detectors (as of December 2023) don’t work reliably. For instance, Turnitin released an AI detector in spring 2023, touting a high level of accuracy, which they have since acknowledged generates more false positive reports than they originally claimed. The CTLI’s professional judgment is that there are no current tools that can confidently provide faculty with an accurate and reliable assessment of whether or not a piece of writing was composed using genAI and that the existing tools carry high risks of false positives, which undermines trust between students and faculty.

---

Deterring Mis-Use of GenAI

James Lang wrote the wonderful book Cheating Lessons: Learning from Academic Dishonesty. His research indicates that humans are more likely to cheat under certain conditions and that faculty can encourage academic honesty by setting classroom conditions conducive to integrity. These principles apply to academic integrity concerns related to genAI as well as other types of cheating. And, most importantly, they also increase learning outcomes.

• Tap into students’ intrinsic motivation for success.
• Emphasize learning for mastery not performance.
• Lower the stakes.
• Set students up for a high sense of self-efficacy.

Beyond encouraging academic integrity, there are some other approaches that may minimize the inappropriate use of genAI:

• Oral assessments. Asking students to prepare ahead and then engage in an oral assessment either 1:1 with you or as part of a class activity will demonstrate a degree of individual effort. While a student could use a genAI tool in preparation, they would still need to internalize that information in order to demonstrate their learning. This handout of 16 oral assessments may spark an idea you could adopt for your teaching.
• OneDrive version history. Ask students to complete papers in Word Online through their VSC accounts and require their submission be a link to that file in their VSC OneDrive account that gives you access to the version history. The version history will show you a date/time-stamped list of when changes were made and if you compare a previous version to a newer version, you can see what changes were made. If a paper goes from blank to full in a matter of minutes, presumably content was copied into it. While that does not mean a student copied content from a website or genAI tool, it provides you with information from which to have a conversation about that students’ writing process.
• Unessays. Unessays can be used in any discipline, are usually assigned at the beginning of the semester (to give time for students to brainstorm and begin work), and allow students freedom of expression to propose a project and then create a meaningful artifact representing their learning. This blog post by Cate Denial outlines her process and provides examples of some unessays students have submitted in her history classes.
• Paper/pencil assessments. You may find it useful to use classtime to have students complete and submit paper-pencil assessments (worked problems, quizzes, exams) where you can monitor their use of outside resources.
• Two-stage exams. During a two-stage exam model, students take an exam individually, first. They then take the exam a second time, but during this step, they work collaboratively with classmates in small groups of 3-5. Students can then submit a second answer sheet, either individually or collectively, after learning from one another’s ideas and logic. The students’ exam grades are usually a combination of the scores earned during the two stages with the weighting of each portion dependent on the instructor’s goals and philosophical approach (e.g., first stage 40%/second stage 60% or first stage 75%/second stage 25%).

If you would like support to implement any of these ideas, don’t hesitate to schedule a consultation with a CTLI staff member.

---

Additional Resources