AI Writing Tools

The widespread adoption of OpenAI’s ChatGPT in late 2022 brought unprecedented attention to the capabilities of generative AI in all types of communication. Generative AI systems are software applications that can create text, images, audio, or video in response to prompts. These systems learn the patterns that exist within very large sets of data and then generate similar content that has some degree of uniqueness.

AI writing tools are one type of generative AI that can support writing tasks by creating human-like text. These systems work by continually predicting the word most likely to come next in each sentence. ChatGPT is a conversational generative AI system, also known as a “chatbot,” that has the ability to recognize plain-text prompts. This system can produce a variety of different styles of text such as emails, essays, scripts, outlines, poems, or song lyrics. The same technology that powers ChatGPT will soon be integrated into popular content creation tools such as Microsoft Word, PowerPoint, and Outlook. Other AI writing tools include Google’s Bard, Jasper, Sudowrite, and Quillbot.

The potential of AI writing tools to modify and augment education is difficult to dismiss. Nevertheless, because these tools are still quite new, instructors and students should carefully consider how, why, and if they should incorporate them into their personal and professional communication.

• Empowering Students – As they prepare students to be professionals, it is important for instructors to give them opportunities to skillfully engage with the technology that will be a part of their chosen discipline. While research on pedagogy that supports AI literacy – or the “ability to understand, use, monitor, and critically reflect on AI applications without necessarily being able to develop AI models themselves” (Laupichler et al., 2022, p. 1) – is limited, the potential of this new technology suggests it will impact all disciplines in some way. By using the AI-powered tools that are integrated into the systems students interact with each day, instructors can help make their courses more relevant while ultimately preparing students for their future professional work.
• Productivity – AI writing tools can support a writer’s productivity. Instructors can utilize these tools to generate initial drafts of many different types of documents including announcements, outlines, and lesson plans. While these first drafts are not perfect, they can work to decrease writer’s block by eliminating the “blank canvas” problem. Additionally, students can also prompt AI writing tools to provide feedback on their writing.

• Customized Course Content – Students come to class with different educational experiences, skill sets, and preferences that can make it difficult to differentiate learning in ways that meet the needs of everyone. AI writing tools can help instructors create content for their courses that exists in multiple formats. For example, ChatGPT can rephrase complex texts at different reading levels. Similarly, generative AI can create custom questions that focus on specific topics that can be used for exam preparation or skill development.
• Language Translation – Generative AI models can also fulfill some language translation tasks which can support students working in non-primary languages. For example, scholars have found that ChatGPT is comparable to commercial translation products on high-resource European languages (Jiao et al., 2023).

• Accuracy – Text generated by ChatGPT may seem plausible but may actually be incorrect or invented, including nonexistent text citations and incorrect mathematical calculations (Azaria, 2022; Southern, 2023). This phenomenon, known as the “hallucination effect,” is a common issue among many natural language processing models (Xiao & Wang, 2021).
• Inequality – Similar to any other fast-growing technology, generative AI models are expensive to build and maintain, meaning that some versions of these tools require a fee or a subscription. While some learners may have the means to purchase more advanced versions of this technology, others may not, which can result in unequal access for students with different economic resources.
• Bias – The inclusion of large-scale datasets as training material for AI systems can amplify systemic, statistical, and human biases (Schwartz et al., 2022). It is important to understand and acknowledge these biases before encouraging students to use AI writing tools.

• Privacy – Generative AI models can collect large amounts of personal information from even the most critical users. It is not always known exactly how this information is used and can cause some data security and privacy issues (Bhutoria, 2022).

The decision to incorporate AI writing tools into a course is complex and nuanced. These tools can empower learners, increase instructor productivity, and support certain learning tasks. However, when considering the use of generative AI, instructors should also seek to understand the capabilities of these AI writing tools, the issues students might have when accessing the tools, the inherent biases of these tools, and the privacy concerns these tools raise. Ultimately, instructors should consider their course learning objectives and whether a specific AI writing tool would help their students meet those targets. For additional guidance, instructors can consult TeachAI, an organization committed to supporting the effective integration of AI into education.

1. Share with students your personal stance on AI writing tools in the classroom. Defining your own thoughts on these tools brings transparency to the issue and clarifies the ethical implication of their usage in the classroom. You likely already discuss norms, expectations, and general class policies with your students. Consider how AI writing tools can be part of these discussions.
2. Develop and include an AI usage policy in your syllabus. After sharing your viewpoint with students on the use of AI writing tools in your course, develop a policy to reflect AI-usage expectations in your classroom and include this policy in your syllabus. Consult page 11 of UIC’s Student Disciplinary Policy, which provides information on Academic Integrity that can be applied to the usage of AI tools.

The following statements may be used as guides or starting points as you develop your own AI usage policy for your course.

• Example statement prohibiting the use of AI writing tools: The use of AI writing tools (including, but not limited to, ChatGPT, Bard, or Sudowrite) is NOT permitted in this course. Students who use these tools for class assignments undermine the goals and learning objectives for this course, reducing the effectiveness of instruction. The instructor may submit student writing to an AI writing detector (e.g., GPTZero) at any point throughout the term. Any confirmed use of AI writing tools will be treated as cheating. Students should reference UIC’s Student Disciplinary Policy for more information.

• Example statement permitting limited use of AI writing tools: The recent advances in AI technology are already transforming the ways humans communicate. In order to prepare students for an AI-infused world, the use of AI writing tools in this class is permitted in some ways. Students are encouraged to use AI writing tools (such as ChatGPT, Bard, or Sudowrite) to generate ideas for their writing and course work in this class, however it is expected that all AI-generated content be reviewed, edited, and verified for accuracy before submission. Please note that you need to cite the specific AI writing tool as a source if you present any significant amount (i.e., more than one sentence) of minimally edited AI-generated text as your own. Please review the APA or MLA guidelines for citing generative AI writing tools.
• Example statement permitting all use of AI writing tools: The recent advances in AI technology are already transforming the ways humans communicate. In order to prepare students for AI-assisted work, the use of AI writing tools is permitted in this course with no restrictions.

Note: Keep in mind that these technologies change at a rapid pace and may require that your policy be updated or clarified regularly. You may also choose to differentiate your policy for different assignments within the same course. To see how other colleges and universities are approaching AI in the classroom, consult this working document: Classroom Policies for AI Generative Tools.

As AI writing tools continue to advance, it is crucial to consider strategies to minimize the risk of student overreliance or misuse when completing their coursework. Here are some potential strategies to reduce reliance or misuse of AI writing tools

• Test out the tool. If you feel comfortable using AI writing tools, test out the type of responses AI writing tools may generate for an assignment in your course. This may shed light on areas of an assignment that you may consider adjusting to promote more authentic work. Consult CATE’s Authentic Assessment Teaching Guide for more information on authentic products, performances, and teaching methods.
• Make student work collaborative. Consider adding group work to your courses, which is an engaged teaching strategy that facilitates student collaboration and makes auto-generated text largely irrelevant to the group work completed in class. Consult CBE Life Science Education’s Evidence-based teaching guide: group work to learn more about applying effective group work strategies to your courses.
• Scaffold student assignments. Segmenting large assignments into smaller, lower-stakes assignments provides students additional opportunities for feedback while emphasizing the importance of revision and progress, which are educational practices consistent with fostering a growth mindset among your students. This could include crafting incremental assignments that build from paper proposals with curated resources, drafting submission of their work, and culminating with final work This strategy enables you to see the entire scope of a student’s process.
• Require accurate source identification and proper citation practices. While AI writing tools are useful for generating ideas on a particular topic, they do not reliably provide accurate references or citations and may even fabricate sources. For research projects, require students to verify all information in their writing with peer-reviewed sources and cite the evidence. Ask students to include citations and direct links to these works in research paper proposals. Application of these references to their final work submissions will make using AI-generated text infeasible.
• Get familiar with your students’ writing or problem-solving processes. Require students to enable track changes throughout assignment components to see their work in progress. Additionally, you might require students to detail their brainstorming process, explain why they chose to write about a particular topic, or show how they arrived at a particular answer to get credit for solving a problem (Cotton et al., 2023).

• Connect assignments to the classroom. Consider connecting assignment prompts to in-class learning activities. AI writing tools do not have access to what happens in class, so requiring the inclusion of ideas that are specific to your courses make output from the AI writing tool less useful.
• Make assignments personal, timely, and locally specific. Include assignment components that integrate experiential knowledge with academic research. This can include self-reflection, specific personal experiences, topics related to their local community, or current events.
• Craft assignments that require higher-level cognitive skills. Assignments that require students to apply, analyze, or evaluate what they have learned diminishes students’ ability to submit AI-generated work as their own.
• Include multimedia assignment opportunities. Consider including assignments that incorporate multiple ways for students to submit their work, such as presentations, videos, or podcasts. Consult CATE’s Universal Design for Learning Teaching Guide to learn more about evidence-based principles that support multimodal assignments.
• Focus on the benefits of learning. Encourage students to focus on the advantages of learning by explaining the purpose of assignments and highlighting the potential drawbacks of relying too heavily on AI. It is important to communicate your expectations for their writing and ensure that your grading practices align with those goals. Emphasize how coursework is relevant to their own lives, and how it leads to personal growth and a sense of achievement. This can help build intrinsic motivation and avoid the perception of assignments being mere exercises to achieve a grade.
• Check that students aren’t submitting AI content for their assignments. There are online tools that attempt to detect AI-generated content, such as GPTZero, Originality.ai, GLTR.io, and OpenAI Classifier. These tools examine the “perplexity” (complexity of text) and “burstiness” (variation in sentence length and structure) and rate a text submission for its probability of being AI generated. However, their detection efficacy is not conclusive. In addition, other tools have appeared on the market that can revise AI-generated text to bypass detection by AI content detectors (Alimardani and Jane, 2023).

Note: Keep in mind that the capabilities of generative AI tools continue to advance, which impacts data output by these tools. Additionally, students will likely continue to build their skills working with these tools, which will impact the AI-generated output they can create. Both factors will impact how effective the strategies described above are in mitigating generative AI reliance or misuse.

While students might readily adopt generative AI technologies and apply them to their own needs, it is important to recognize that there are many different ways to productively incorporate these tools into a course. Educators are already exploring generative AI tools and developing creative classroom activities and assessments. These pedagogical applications of AI writing tools help students learn how to craft prompts that garner useful output, evaluate AI text output for quality, accuracy, and originality, as well as write well-structured and cohesive essays that combine AI-generated content with one’s own writing.

A large collection of educator-generated AI-oriented assignments can be found at Creating a collection of creative ideas to use AI in education. The Stanford Graduate School of Education has also compiled resources for teaching AI literacy that can be found at Curricular Resources about AI for Teaching (CRAFT).

While most AI writing tools are relatively simple to use, there are several skills that students and instructors should develop as they engage with this new technology. For example, writers should carefully plan how they prompt these tools so that they are more likely to receive high-quality content. Writers that use generative AI to support their work should also know how to cite these tools. Furthermore, writers should always critically evaluate the validity of any AI-generated content. Below are strategies to support the development of these skills.

Crafting Prompts for AI Writing Tools

Prompting refers to the process of providing an input or a series of inputs to the AI model in order to initiate a conversation or generate a desired response. Effective prompting is essential to obtain the most useful output from AI writing tools. A prompt can be a question, statement, or instruction, which the model uses to understand the context and generate a relevant, coherent, and helpful response.

Basic Prompting Technique

• Define the target audience: “For an undergraduate introductory non-major Physics course”
• Set the context: “At a large public research university in the United States…”
• Define the task: “Craft a ninety-minute lesson plan”
• Include specific details: “Based on the following learning objective: Explain the properties of waves, including wavelength, frequency, amplitude, and velocity.”
• Complete prompt example: “Craft a ninety-minute lesson plan based on the following learning objective: Explain the properties of waves, including wavelength, frequency, amplitude, and velocity. This lesson plan is intended for an undergraduate introductory non-major Physics course at a large public research university in the United States.”

Chain-of-thought prompting is a technique that uses a series of interconnected and/or iterative prompts or questions to guide the conversation in a specific direction, maintain context or relevance, or dive deeper into a topic. Chain-of-thought prompts follow a logical order or sequence and are often based on the AI tool’s previous responses;

that is, each subsequent prompt is based on the information or context from the previous response, creating a coherent and context-aware conversation. According to Chen et al. (2023), chain-of-thought prompting can improve the accuracy of output when utilized effectively.

Citing Generative AI Sources

In academic scholarship, writers need to cite the sources that they paraphrase, quote, and reference in their work. While major publishing organizations such as the Committee on Publishing Ethics have clearly stated generative AI tools can not be listed as authors on papers, it is important that scholars at all levels be transparent about their use of these tools. When writers paraphrase or quote AI-generated text, they can follow the preliminary guidelines released by the American Psychological Association (APA), the Modern Language Association (MLA), or the Chicago Manual of Style. Additionally, instructors might also seek out discipline-specific guidelines or develop a citation system that supports the needs of their students.

Evaluating Citations Generated by AI Writing Tools

Users should expect generative AI writing tools to produce a mix of real, partially correct, and completely fabricated citations. When prompted a question about fake citations, ChatGPT provides the following advice:

“As an AI language model, I don’t have direct access to a database of scholarly citations. I generated those examples based on my understanding of the topic. To obtain accurate and reliable scholarly citations, I recommend conducting a search in academic databases such as Google Scholar, JSTOR, or PubMed … These databases will provide you with authentic and properly formatted citations that you can include in your research paper” (OpenAI, 2023b).