A hybrid MCDM approach for unveiling ChatGPT's effect on students' learning

Texto completo

The education sector received significant benefits from ChatGPT as an Artificial Intelligence (AI) chatbot because it strengthens human-machine teamwork while making education more efficient and adapting to individual students and reducing their workload. This research paper focuses on investigating three primary research objectives concerning the integration of ChatGPT into educational settings. Firstly, the study aims to assess the effects of ChatGPT on students' learning outcomes with identifying the main cause and effect impacts. Secondly, it seeks to identify and analyze various strategies that can be employed to effectively utilize ChatGPT for educational purposes. Lastly, the research endeavors to establish the priority levels of these strategies to optimize the effective implementation of ChatGPT in educational contexts. To achieve these goals, a hybrid multi-criteria decision-modeling (MCDM) approach, including Decision Making Trial and Evaluation Laboratory (DEMATEL) and newly developed Comprehensive Distance Based Ranking (COBRA) methods, have been integrated. Using data gathered from student questionnaire responses, DEMATEL method identified the main factors impacting student behavior and academic practices by classifying twelve effects into cause-and-effect categories. This categorization ensures successful ChatGPT integration in education by assisting policymakers and academics in creating targeted approaches to address the root cause rather than surface-level effects. In a focus group discussion and MCDM analysis based on academicians' survey responses, the strategy that minimizes AI dependence that improves critical thinking and promotes self-directed learning was identified as ‘Customize, timely and area specific assignments’. In order to validate the outcomes derived from the proposed model, alternative MCDM methodologies, including COmplex PRsitional ASsessment (COPRAS), Combined Compromise Solution (CoCoSo), Multi-Attributive Border Approximation Area Comparison (MABAC), and Measurement of Alternatives and Ranking According to the COmpromise Solution (MARCOS), have been employed. Subsequently, to assess the robustness of the proposed model, sensitivity analysis has been conducted under various experimental conditions. After rigorous evaluation, the COBRA technique emerged as the most appropriate among the models considered, exhibiting comprehensive calculations and demonstrating the highest Spearman's correlation with other models.