ChatGPT: Friend and foe in the geography subject didactics classroom? Implications for practice
Abstract
Since 30 November 2022 an AI chatbot known as ChatGPT has taken the world by storm, as it can be prompted to answer a wide variety of questions. Almost immediately, reports showed that the chatbot’s answers were good enough to pass a variety of postgraduate examinations, with the result that lecturers were worried that the chatbot could have a negative impact on academic integrity. Some universities and schools reacted by banning ChatGPT on their campuses in fear of an increase in academic dishonesty. Others argued that the chatbot should rather be integrated into classrooms to improve learning experiences, as chatbots are also integrated into workplaces.
Due to the novelty of this chatbot, little research has been conducted to understand whether it can be regarded as a friend or foe in higher education classrooms. As I am a trained geography teacher and have also been a lecturer in geography subject didactics for years, this research aimed to understand whether ChatGPT could be regarded as a friend or a foe in geography subject didactics classrooms. The following questions guided the study:
- How can ChatGPT be misused in the geography subject didactics classroom?
- How can ChatGPT be used in the geography subject didactics classroom?
- How can ChatGPT texts be differentiated from texts written by humans?
- What are the implications for practice?
The research was based on the theoretical assumptions that students would use and misuse ChatGPT due to its:
- relative advantage over the status quo (Rogers 1983)
- trialibility (Rogers 1983)
- observability (Rogers 1983)
- compatibility (Rogers 1983)
- ease of use (Rogers 1983 and Van Staden 2021)
- potential to improve work (Van Staden 2021)
- usefulness (Van Staden 2021).
As the first four factors drove the quick uptake of this disruptive technology I assumed that the last three factors – ease of use, the potential to improve work, and usefulness – could motivate the students to use and misuse ChatGPT in geography subject didactics classrooms.
The research was conducted from a pragmatic worldview, therefore I wanted to make a theoretical contribution and better understand how the practice could be improved. The best methodology for this type of research was action research. As ChatGPT answers questions in a human-like manner and understands questions, I regarded this chatbot not only as the subject of study but also as a participant that could shed more light on its usability in geography subject didactics classrooms.
The research was conducted in an action research framework consisting of five steps. First, I identified the problem (how ChatGPT can be used and misused); then I designed a plan of action (questions I could ask ChatGPT); implemented the plan of action (asked ChatGPT the questions); collected data regarding the possibilities to use and misuse ChatGPT; then reflected on the results and presented the findings in this study. During the third step I followed the same action research cycle for each of the questions: I identified a problem (lesson plans), designed a question to be asked, asked the question, collected data and rephrased the question if necessary until I was sure that ChatGPT could answer the question (or not), and shared the results for each question in this article. After I had asked the subject-related questions I asked for ChatGPT’s opinion about being a friend or a foe in the geography subject didactics classrooms. During the reflection phase of the research I used the results to understand whether ChatGPT could be regarded as a friend or foe in a geography subject didactics classroom.
The most important finding was that ChatGPT can be regarded as a foe in the geography subject didactics classroom. It created a year plan, a lesson plan and a test with a memorandum and assessment guidelines within seconds. The lesson plan was not effective, as the chatbot could not get access to the CAPS documents, but after I had read the necessary private data into the chatbot, it provided an impressive lesson plan. ChatGPT could also set questions on the various levels of Bloom's taxonomy and even correctly indicate at which level the questions were set. If an answer was incorrect, I could adjust my prompts to obtain more accurate responses. It also improved a text I had read which contained many language errors. However, it was not able to identify and correct factual mistakes. Due to the quick responses, quality of Afrikaans language usage, ability to learn, published reports on the efficiency of the chatbot and the credibility of most of the answers, it can be deduced that students might misuse this chatbot to complete their assignments.
According to ChatGPT itself it could be regarded as a foe, as it could hinder the development of various skills, including problem-solving, critical thinking, research, communication and interpersonal skills. It also argued that it could impede the development of unique perspectives. Although the chatbot could be regarded as a foe, I do not support recommendations that it should be banned, as students can use various apps built on the GPT-3 model, and many chatbots are in the developmental process. Instead, I support recommendations that new and innovative assessment practices be developed and used to combat academic dishonesty.
The second finding is as important as the first, as ChatGPT can also be regarded as a friend in the geography subject didactics classroom. It made mistakes, falsified or hallucinated sources and links, could not identify factual errors in a text I had provided, and presented texts confidently with mistakes. The falsification of resources can get students into trouble, as they could be accused of resource plagiarism and falsification of resources.
According to the chatbot itself it can be regarded as a friend, as it can provide additional resources, interactive learning opportunities, personal support, and creative ideas for essays, and it can also provide prompt feedback as it is always available. Although these arguments may be valid, this research shows that ChatGPT is imperfect and can make mistakes. However, these mistakes can be used to create rich learning opportunities.
Based on the findings I do not support Susjnak's recommendation that universities return to invigilated and oral examinations. Instead, I recommend that assignments be designed in such a way that ChatGPT cannot solve them. I recommend a learning-oriented approach to assessment which is based on three pillars, namely that (a) learning rather than assessment tasks be designed, (b) peers be used as assessors and (c) feedback be provided promptly. ChatGPT could, for example, be used to create a first draft of a lesson plan. Then the drafts can be improved and shared with peers for feedback. The feedback can further enhance the lesson plan before submitting it for assessment. The feedback from the lecturer can be used to improve the lesson plan. Following this route, students can be sent into practice with a treasure chest filled with improved lesson plans.
The third finding is that detectors of plagiarism and AI are inadequate. None of these tools was able to identify the texts written by ChatGPT. Therefore it is important to design assignments that ChatGPT cannot complete correctly.
Lastly, based on the findings, it can be assumed that students will use this disruptive technology to complete their assignments due to the following:
- It can provide a relative advantage over the status quo (Rogers 1983)
- It is trialible (Rogers 1983)
- The results are observable (Rogers 1983)
- It is compatible with their needs (Rogers 1983)
- It is easy to use (Rogers 1983 and Van Staden 2021)
- It can improve the quality of work (Van Staden 2021)
- Its usefulness (Van Staden 2021).
This research investigated whether ChatGPT could be regarded as a friend or a foe in the geography subject didactics classroom. Based on the results, this chatbot can be both. As a foe it can be used to be dishonest; therefore I recommend that lecturers design questions that ChatGPT cannot easily answer, assignments it cannot easily carry out. As a friend, it makes mistakes that can be used to create rich learning opportunities in the geography subject didactics classroom.
Keywords: ChatGPT; curriculum studies; didactics; geography; geography subject didactics; higher education; teacher training
- The photo on this article’s featured image was taken by Google DeepMind and obtained from Unsplash.
Lees die volledige artikel in Afrikaans:
ChatGPT: Vriend en vyand in die Geografie-vakdidaktiek-klaskamer? Implikasies vir die praktyk