Creating Assignments that Outsmart AI GPT-3 Software

Teaching & Learning Guide

Welcome to the Teaching & Learning Guide for Creating Assignments that Outsmart AI Software! 

This guide includes best practices and tips designed to help you: 

What is AI GPT-3 Software? 

Generative Pre-trained Transformer 3 (GPT-3) software is an autoregressive language model that produces human-like text. With the rise of GPT-3 software, instructors may be concerned with students’ use of programs and services to buy assignment submissions and responses. 


One major concern of GPT-3 software is that since text is generated based on a prompt, it is not copied from somewhere else. This means plagiarism detection tools like Turnitin will not flag AI-generated essays, at this time. Turnitin is currently working on AI detection.


With more students exploring GPT-3 for writing assignments, many education leaders and EdTech specialists have suggested that now is the time for instructors to shift their teaching with AI in mind, or even incorporate AI into their teaching. While it is concerning that traditional essay-writing may be less impactful to student learning, there are ways to write assignments and assessments that won’t allow students to use GPT-3 software. 


Teaching continues to adapt based on technology available. With the rise of grammar and spell-check software, many instructors no longer assess grammar and spelling, for example. Adjusting to this type of software is just part of that continued teaching adaptation. 


In this guide, we’ll discuss some strategies for writing assignment prompts and assessment questions that outsmart GPT-3 software. 

Writing Assignment and Assessment Prompts

Tips for creating prompts and assignments: 

Use Course Context. 

Write questions and prompts based on past coursework. GPT-3 software algorithms can’t decipher indexicality - this means that they can’t write a response if it is based in the context of the course. For example, you can ask students to summarize a discussion or reference a lecture. You might ask something like, “Based on our discussions from module two, what are the two main arguments in X theory?”

Compare and Contrast. 

GPT-3 Software also can’t compare and contrast texts. Asking students to do analytical comparisons will make it difficult for them to rely on this software. These comparisons can be more effective when you require them to choose course texts. You could even shorten this type of prompt for a quiz question by asking students to compare the argument or point made in a certain part of one text with a specific section of another. 

Find what's missing. 

Asking students to find what’s missing from a list is a great way to help them develop problem solving skills. AI software also can’t analyze a list. This type of prompt could be in the form of giving students a list of formulas, data, or terms and asking them to identify the missing formula, data, or term. 

Analyze their own essays.

If you suspect that students may be using GPT-3 essays, you can expand the workflow. Ask students to analyze their essays or work, based on course concepts. This will ensure that even if students do use GPT-3, they will still have to complete the analytical step themselves.

Use GPT-3 essays.

You can use GPT-3 as a learning tool. For example, generate one or more essays about a topic and ask students to evaluate them. They can serve as helpful templates or examples for students, while discussions about them can help students write even better essays. Plus, it’s a subtle way to let students know you are familiar with GPT-3 software.

Scaffolded assignments.

Scaffolding means students are asked to complete a large project in smaller parts. This could mean asking students to complete an annotated bibliography, then milestones toward their paper. Or it could mean asking students to complete worksheets to help them complete the end assignment. The key is to check the parts as students complete them, so that you can continually track student progress.

Case studies.

Case studies are another great way for students to build problem-solving skills and also get practice in their field. Providing a case study or option of a few case studies helps ensure students have fewer options for seeking software-written papers. With a case study assignment, you could ask students to summarize the problem and propose a solution. Or, you could require students to use course theories or frameworks to explain the case.

Apply concepts to real world scenarios.

There are numerous ways to ask students to apply course concepts outside of class. You could ask students to reflect on where they’ve noticed the concepts in their everyday lives. Or you might ask students to reflect on a provided scenario using tools they’ve learned in the course.

References

https://www.chronicle.com/article/will-artificial-intelligence-kill-college-writing

Resources for Instructors at TWU

Instructional Design Partners

Instructional Design Partners in the Center for Development, Design, & Delivery design and present learning solutions to continually enhance institutional and instructor performance. We collaborate closely with instructors to translate course objectives into meaningful, customized courses tailored to each instructor’s specific needs, leveraging an aptitude for design and development, along with excellent problem-solving and analytical skills. 

Our technical expertise encompasses a range of programs and best practices, including Canvas, Quality Assurance, Universal Design, and more. Instructional Designers partner with academic components to answer questions about teaching and learning in one-on-one consultations, small group work, symposia, and workshops.

Technical Support

To request technical support, submit a Technology Service Desk email to start a ticket.