Teaching Resources

ChatGPT and AI Composition Tools

Resource Overview

Information about AI tools and their implications for teaching and learning

We will be updating the information and recommendations on this page as we all learn more about this new technology.

What is ChatGPT?

ChatGPT is an AI chatbot that was launched in November 2022 by the company Open AI. A chatbot is an interface that allows someone to have an automated and dynamic “conversation” with a computer that mimics an interaction with an actual person. What makes ChatGPT different from other bots is the extremely high quality of its responses and the wide range of topics it can respond to. ChatGPT is publicly available and free to use, which means almost anyone can use the tool to quickly generate text. Since ChatGPT’s launch other similar advanced AI composition tools have become publicly available.

WashU has created an institutional version of ChatGPT: https://gpt.wustl.edu/chat

What Can ChatGPT and Other AI Tools Do?

AI composition tools can complete a variety of tasks, from writing short essays on almost any topic to generating working software code. We offer some some examples of the ChatGPT’s responses to a number of prompts. Each answer was generated in a matter of seconds. While these examples can give you an idea of the tool’s capabilities, the best way to understand what AI composition tools can do is to try them out yourself.

While ChatGPT is very impressive, it also has some limitations. The tool was trained on data from 2021, so it is not very good at commenting on current events. It also has limited abilities in some languages outside of English, and generally has trouble producing citations for what it writes. Most importantly, ChatGPT and other AI tools can be wrong and sometimes produce plausible sounding, yet nonsensical responses.

Concerns

The ability to quickly and easily generate relatively high-quality text raises a number of concerns. Faculty at WashU and other institutions have wondered how ChatGPT will impact the integrity of take-home exams and essays, the utility of homework, and possibility of AI-generated discussion post responses, among other concerns. This is not resolved by AI-detection software, which tends to have very low accuracy and is unreliable. Many faculty are also worried that focusing too much on policing student work for possible AI use will damage the learning environment and their relationships with students.

Recommendations

As widely available sophisticated AI composition tools are new, the higher education community is still grappling with their potential impact and how best to respond. The Office of the Provost has created a resource Addressing Artificial Intelligence in 2023, and we offer the following complementary suggestions:

Assignment Design

  • Learn how ChatGPT would respond to your class assignments: Try out ChatGPT yourself to understand its capabilities and limitations.
  • Make assignments more authentic and robust: Authentic assessments asks students to complete complex tasks similar to those they would in a non-academic setting.
    • Add non-traditional ways of assessing learning (podcasts, oral exams, projects): Consider other ways of letting students demonstrate the knowledge they’ve learned in their course beyond the traditional college essay.
    • Apply open education practices that involve collaborating with students to create and curate content: Engage students in open education and development of Open Educational Resources (OER) that are student-driven and reflective of the sharing of ideas and material common in various spaces and communities, especially those committed to teaching and helping other instructors!
  • Use local or class-based prompts: One limitation of AI tools is that they do not know what happens in your class. Consider designing prompts related specifically to your course. For example, one could ask students to expand upon a particular class discussion or answer questions requiring knowledge from a lecture slide.
    • Leverage other educational technologies: Instead of using open-ended questions in Canvas, try social annotation tools that require students to engage with a text or video along with their classmates. WashU supports Hypothesis for social annotation of texts and Annoto for collaborative commenting on videos.
  • Focus on the good parts of student voice: Consider how your assignments can allow students to showcase their voice, style, interests, and connection to your class. There are many ways students can show more of their humanity and whole self to learn more about themselves or their peers. Students can even help design prompts or questions in the assignment by discussing what they want to articulate, share, or learn about others in the assignment.
  • Use ChatGPT in your assignments: Help students leverage ChatGPT and other AI tools for their learning. Such tools are going to be available for their present and future work, so including them in assignment design can help them learn how to use them even more effectively.
  • Tackle ChatGPT with Hypothesis: Hypothesis, a social annotation tool supported by WashU, can engage students in collaborative textual analysis of AI generated content, deepening their understanding of the capabilities and limits of the AI tools.
  • Be explicit about when/how students can or cannot use AI tools. For your assignments, identify skills that are crucial for your students to practice and those that are less crucial and would benefit from AI support. Create/adapt a spectrum articulating ways students and AI can contribute to the assignment. and be clear with students about the boundary that is acceptable. See these example tables for potential spectrums of AI-Student Collaboration that you can copy/use in your class.
    • For example, boundaries can vary across assignments and across language classes. At the intro level, students may need to practice grammar, so students must proofread on their own for each assignment. For other classes, proofreading and getting feedback from AI may be acceptable, but writing the core arguments and content may not be. For other course, AI could be used to generate possible ideas that students then curate, build on, revise, or synthesize.
  • Have students cite and attribute use of AI-generated text in text or in a table. Make explicit when and how students leverage AI for assignments so that you and they are also aware of the impact of their input and output from AI. You can engage with the WUSTL Library and your subject librarians to get support for your students on citation.

Course Design

  • Design your course with attention to AI as a collaborator: AI will likely play a role in the future of our workplaces, labs, and society as a whole. Consider the role of AI in your course and outline for yourself how you expect AI to be a collaborator on class activities so that it becomes clear to students as well.
  • Update your syllabus language: Our syllabus template now has additional language added specifying that the unauthorized use of AI-content generators is a form of academic dishonesty.
  • Incentivize the learning process, not just the product: Asking students to submit one final perfect paper for an A may increase the odds of cheating. Instead, provide points for behaviors and habits associated with learning such as articulating ideas, failing, improving, and reflecting on one’s learning. For essays, this may be outlining, drafting, reviewing, and revising, which are essential skills for students to develop in the thinking and writing process. You can have students turn in intermediate assignments, work on pieces in class, or leverage peer discussion to help develop final products.
  • Request a 1-1 consultation: Get more specific recommendations for your class by working with one of our educational developers.

Further Resources

Basic Information

News Outlets

Blog Posts

Additional

Have suggestions?

If you have suggestions of resources we might add to these pages, please contact us:

ctl@wustl.edu(314) 935-6810Mon - Fri, 8:30 a.m. - 5:00 p.m.