ChatGPT: Changing Education, One Prompt at a Time, A quantitative study of factors influencing students' usage of ChatGPT

Sammanfattning: An AI tool that has gained massive media attention in the past months is the language model ChatGPT. The rapid adoption of the tool by students made the authors of this study curious about exploring the chatbot further, using technology adoption theory. This study aims to understand how the original constructs of UTAUT2 (Unified Theory of Acceptance and Use of Technology 2): Performance Experienced, Effort Experienced, Hedonic Motivation, Social Influence & Facilitating Conditions, influence students' usage of ChatGPT. To better fit the novel technology of ChatGPT, the authors have extended UTAUT2 with Trust, Accessibility, Humanization, Data Privacy Risks & Creativity. Ten hypotheses have been formulated based on the constructs presented above. To test the hypotheses, data on students at Stockholm School of Economics have been collected through an online questionnaire (N = 205). The data was analyzed through multiple regressions, with three different dependent variables of usage frequency (1) "How often do you use ChatGPT?" (2) "How many questions do you ask ChatGPT on average per week?" and (3) "When was the last time you used ChatGPT?". The author's extension of UTAUT2, adding additional constructs, resulted in a greater ability to explain the usage frequency of ChatGPT among students. Due to reliability and validity issues, the following constructs were removed from the multiple regressions: Social Influence, Facilitating Conditions, Accessibility, Humanization and Data Privacy Risks. Out of the remaining five constructs, Performance Experienced, Creativity & Hedonic Motivation influenced students' usage of ChatGPT. Performance Experienced was shown to be the strongest predictor of students' usage of ChatGPT, while Effort Experienced was non-significant and Hedonic Motivation had a significant negative effect. Regarding the contributing constructs, Creativity showed a significant positive effect on student's usage of ChatGPT and Trust showed a non-significance effect. This study contributes with its findings that Creativity is an important significant predictor of students' usage of ChatGPT and that Performance Expectancy can also be generalized to this context. Moreover, the dependent variable of usage has been studied, instead of the mediator of behavioral intention, which leads to more valid results and further contribution to the literature. Lastly, the findings indicate that UTAUT2 has limitations when predicting usage of ChatGPT and therefore new models and theories need to be explored. These can be built upon the findings of this study in order to provide a more comprehensive understanding of ChatGPT usage.