As an instructor, you almost certainly use artificial intelligence (AI) daily in your academic work without knowing or caring how it works. Email spam filtering and Microsoft Word\u2019s grammar checker are just two examples.\u00a0<\/p>\n
However, a new AI chatbot technology just released, ChatGPT, is a tool you cannot ignore.\u00a0<\/p>\n
To help you better understand this new tool and how it can be used, here are five steps you can take to become more familiar with ChatGPT.<\/p>\n
1.\u00a0 \u00a0Start exploring ChatGPT\u00a0<\/strong><\/p>\n The first step is to start learning to use ChatGPT and to understand its potential. If you don\u2019t have an account already, create one at https:\/\/chat.openai.com\/<\/a>.\u00a0<\/p>\n Once logged in, you see a prompt to enter text. ChatGPT is designed to understand natural language so just use ordinary language for your inputs. As ChatGPT itself suggests, begin to \u201cexperiment with different types of inputs, such as short phrases, questions and longer paragraphs, to see how the model responds and what kind of output it generates.\u201d ChatGPT can perform a variety of functions. It can generate content on topics you suggest, create articles, blogs, poetry and product ads, summarize text, translate text to other languages and even produce computer code.<\/p>\n Try entering questions you consider in your lectures or paste in your assignment questions. Compare the answers ChatGPT generates for accuracy and quality of writing.\u00a0<\/p>\n You can also give ChatGPT specific instructions, such as \u201cWrite an article of 1,000 words on topic\u2026\u201d or \u201cSummarize in 100 words the following text\u2026\u201d or \u201cImprove this text\u2026\u201d or \u201cWrite an essay in the style of [e.g., author, magazine or newspaper]\u201d or \u201cTranslate into French\u2026\u201d.\u00a0<\/p>\n Feel free to try out as many prompts as you like. If you find ChatGPT\u2019s response inadequate, tell it to \u201cgive more details about this\u201d or \u201cexpand upon your answer\u201d because it can remember your previous questions. You can also just click on the Regenerate Response button that appears after it replies to your prompt, which causes the tool to respond with more detail.<\/p>\n As you become more familiar with ChatGPT, you begin to see its strengths as well as its limitations. Occasionally, sentences may appear nonsensical. This is because ChatGPT uses statistical patterns to generate text without understanding the meaning. If you ask ChatGPT about current events, you\u2019ll notice it cannot respond. The current version does not contain information more recent than 2021 in its dataset, and it is not connected to the Internet (at this point). Since ChatGPT is trained with data taken from the Internet, its responses may also reflect biases, stereotypes and offensive language found on the Internet. Moreover, ChatGPT is trained not to respond to malicious questions or otherwise inappropriate questions.<\/p>\n 2.\u00a0\u00a0 \u00a0Participate in professional learning<\/strong><\/p>\n After becoming more familiar with ChatGPT, you can take advantage of professional opportunities to learn more about the tool.\u00a0<\/p>\n Higher education institutions are working to understand its implications<\/a> for academic integrity and possibly even their future existence. Discussions are being held by academic administrators, admissions officers and faculty groups about how to respond. Many institutions are offering workshops and webinars for faculty to learn about ChatGPT, including York University<\/a>, Concordia University<\/a> and University of California Irvine<\/a>. Check your own institution or search for others that may be open to outsiders.<\/p>\n Beyond these more formal offerings, you can look for YouTube videos that demonstrate how to use ChatGPT. Many are now available despite ChatGPT\u2019s newness. For example, there is a popular one by Adrian Twarog<\/a>. You may also want to watch academic discussions<\/a> about the tool\u2019s potential impact on higher education.<\/p>\n 3.\u00a0\u00a0 \u00a0Try ChatGPT in your courses<\/strong><\/p>\n Although many institutions want to ban ChatGPT outright and rewrite academic integrity policies to include it and similar AI technologies, students will still use it.\u00a0<\/p>\n Therefore, consider introducing ChatGPT in your courses and discuss with students how it could be legitimately used. Brainstorming, researching specific topics, summarizing complex concepts, and improving written expression are several applications to try.<\/p>\n Another idea is to have students write on an assigned topic and compare their paper to the response ChatGPT generates. Students could note the differences and share them in class.<\/p>\n You could also have students find appropriate citations for ChatGPT-generated responses. This is a valuable exercise since ChatGPT doesn\u2019t provide citations for its output unless you specifically ask for them. Students could be asked to do independent research outside of ChatGPT to find these references or to validate the ones the tool produces. In the process, they may discover faulty statements generated by ChatGPT that demonstrate firsthand the tool\u2019s limitations.<\/p>\n 4.\u00a0\u00a0 \u00a0Share teaching ideas<\/strong><\/p>\n Hearing how colleagues use ChatGPT in their classes and sharing your own uses can be a significant way to become more familiar with the tool. Since the technology is so new, one of the best ways to find out more about how others are using it in their teaching is to turn to social media. For example, search Twitter for #chatgpt and find dialogues such as this one<\/a> where there\u2019s a discussion thread on how faculty are using the tool. You can try searching hashtag combinations, such as #chatgpt<\/a> #highered<\/a> or #chatgpt #edtech<\/a>.\u00a0<\/p>\n LinkedIn is another source to get ideas and share your own about teaching with ChatGPT. Although LinkedIn\u2019s orientation is more toward business, you\u2019ll find discussions, articles and workshop announcements related to higher education. You might also try searching TikTok and Facebook.<\/p>\n 5.\u00a0\u00a0 \u00a0Learn to identify ChatGPT-generated text.\u00a0<\/strong><\/p>\n If you suspect a student has plagiarized, there\u2019s no easy way to recognize whether text is generated by a human or ChatGPT. With some practice, however, you may get a better sense of the text\u2019s authenticity.\u00a0<\/p>\n ChatGPT generated the following characteristics of text generated by language models (e.g., ChatGPT) that you might watch for:<\/p>\n \u2022\u00a0\u00a0 \u00a0Coherence and fluency: Text generated by language models can often be coherent but may lack the fluency and naturalness of human-written text and may have grammatical errors or unnatural word choices.<\/p>\n \u2022\u00a0\u00a0 \u00a0Repetition and lack of originality: Text generated by language models may repeat certain phrases or information and may lack originality in comparison to human-written text.<\/p>\n \u2022\u00a0\u00a0 \u00a0Lack of context: Text generated by language models may not be able to capture the nuances and subtleties of context the way a human writer can.<\/p>\n \u2022\u00a0\u00a0 \u00a0Lack of creativity: Text generated by language models may not be able to come up with new ideas or creative approaches the way a human can.<\/p>\n \u2022\u00a0\u00a0 \u00a0Lack of understanding of the intended audience: Text generated by language models may not be able to adapt its style and language to fit the intended audience.<\/p>\n Not surprisingly, AI-based systems are being developed to detect machine-generated text. Recent tools garnering interest from education institutions around the world are ChatZero<\/a> and AICheatCheck<\/a>. All you need to do is paste text into its input window and the tool assesses the likelihood of it being AI-generated.\u00a0<\/p>\n Unfortunately, you cannot trust these tools to accurately detect AI-based (or human-generated) text. For example, ChatZero\u2019s score identified the above bullet points copied directly from ChatGPT were \u201clikely\u201d human written. AICheatCheck gave a 93.77% assurance. Additionally, the current version of the popular tool Turnitin Originality detects AI-generated text to some extent, but the company is working on a product<\/a> specifically to detect ChatGPT-generated work.<\/p>\n \u00a0<\/p>\n Therefore, until technological solutions work significantly better, faculty and instructors should use their judgement to assess the origin of their students\u2019 work.\u00a0<\/p>\n What\u2019s next?<\/p>\n