What Can Faculty Do to Fight Academic Misconduct?
Academic misconduct is a growing problem in higher education, with reports suggesting the number of cases rose dramatically during the pandemic and continue to rise.
The arrival of ChatGPT, a new tool that create answers to questions and helps students write assignments, project briefs and research papers, only exacerbates the problem.
For example, University of East Anglia professor Fahri Karakas asked ChatGPT to describe top mental models for learning and creativity in 13 different fields, including physics, chemistry, biology, psychology and sociology. The tool produced 77 mental model summaries for those fields, each clearly written and about 100 words long. As an illustration, ChatGPT described eight models for psychology, one of which is cited below.
“Flow State Mental Model: This mental model suggests that people have the capacity to reach a highly focused state of creative clarity and peak performance when they are fully absorbed and motivated by their activities and tasks. In this state, the individual can access deeper levels of problem-solving, creativity and productivity. Learning, creativity and personal development can be heavily influenced by this state, as it allows an individual to fully concentrate on the task at hand and put forth their best effort.”
This exercise demonstrates that ChatGPT can provide almost instantaneous, helpful and accurate information. Missing from its responses are citations for sources because it relies only on mapping patterns and associations from the dataset on which it was trained.
Another limitation is that ChatGPT is not connected to the Internet, so it cannot respond to current events. For example, when asked why Russia invaded Ukraine in early 2022, it responded with:
“I'm sorry, but I am unable to provide information about current events as my training data only goes up until 2021 and I am not able to browse the Internet.”
ChatGPT will generate answers to almost any question it’s asked. The accuracy of its response depends on the validity and reliability of the large human-generated dataset on which it was trained. If the dataset contains inaccuracies or biases, these too will be reflected in ChatGPT’s answers.
OpenAI, the developer of ChatGPT, does not disclose the nature of its content except to say it contains “a wide range of text from a variety of sources, including books, articles, websites and other types of written content… [including] text written in a variety of languages and in different styles and formats.”
Indeed, some faculty are concerned the quality of responses seems to be weak, and students who used them “as is” would likely get low or failing grades. On the other hand, ChatGPT performed at or near the passing threshold for the United States Medical Licensing Exam (USMLE) without any specialized training or reinforcement. These observations suggest ChatGPT will generate better answers in some fields than others. Indeed, OpenAI concedes that some knowledge fields may be better represented than others in its dataset.
Regardless of the quality of ChatGPT’s responses and its limitations, some students will inevitably use the tool dishonestly for their coursework, which means faculty must be aware of what they can do to mitigate cheating and plagiarism.
We recommend five steps faculty can take to reduce academic dishonesty.
Paradoxically, ChatGPT itself can offer faculty guidance on steps to reduce the likelihood of students using the tool to cheat and plagiarize. When asked how to do this, ChatGPT responded with these four suggestions:
- Use open-ended assessments: Rather than asking students to answer specific questions or complete predetermined tasks, faculty can design assessments that require students to engage in critical thinking and problem-solving. This can make it more difficult for students to cheat, as they will need to come up with original ideas and solutions rather than simply copying answers from a tool like ChatGPT.
- Use collaborative assessments: Assessing students’ ability to work effectively with others can also help reduce the risk of cheating. This can be done through group projects, collaborative problem-solving tasks and other activities that require students to work together.
- Provide adequate support and resources: If students feel they have the support and resources they need to succeed in their course, they may be less likely to cheat. Faculty can help create a positive, supportive learning environment by offering additional resources and support for struggling students.
- Use a variety of assessment methods: Rather than relying solely on traditional exams or quizzes, faculty can use a variety of assessment methods such as presentations, portfolios and projects. This can help ensure students are evaluated on a wide range of skills and abilities, rather than just their ability to memorize and regurgitate information.
These ChatGPT suggestions make sense and are consistent with general guidelines offered for academic use of Internet sources. Here’s a fifth step faculty can take:
5. Check ChatGPT’s answers to assignment questions: Faculty can do this to spot student answers that plagiarize from the tool. They might carry this exercise further and share with students ChatGPT’s answers to various course-related questions. This serves several purposes:
- Illustrates that the instructor is aware of ChatGPT
- Prompts a discussion about the role ChatGPT might play in the course
- Creates an opportunity to compare and contract ChatGPT’s responses to those of authorities on the topic
- Provides a springboard for students to develop their own answers
Coupled with the above should be a discussion with students about the proper citation of sources, including ChatGPT, and how to avoid plagiarism. Ideally, this topic should be addressed at the program level for all courses. This seldom happens — so individual faculty may need to stress, for example, that copying and pasting without using quotation marks and citing the source is inappropriate.
It's vital to help students understand the appropriate academic uses of a tool that will be a routine part of their future.
The full impact of ChatGPT and other similar tools is yet to come, but ultimately, it will compel faculty to re-examine their teaching, just as the widespread adoption of the Internet did. Attempts to ban its use, like some faculty tried with Wikipedia when it launched, will not be successful.
Faculty must adapt to a world in which ChatGPT can answer questions, generate essays and summarize texts on just about any topic. They need to consider their course learning outcomes and incorporate authentic, open-ended tasks that are not easily solvable by ChatGPT. Finally, they must rethink their assessment practices in order to avoid simply testing for knowledge of facts and incorporate a variety of different assessment strategies.