Analysis of the effectiveness of the modified Kolb inventory in identifying learning styles in adults

Abstract

Introduction/Justification: Learning in adults is a complex process, influenced by prior experiences, practical needs, and how the individual interprets and applies knowledge in real-world contexts.¹ In this scenario, understanding learning styles becomes fundamental for the development of more effective educational strategies, especially in areas that demand rapid decision-making and practical application of knowledge. The experiential learning model proposed by Kolb² (1984) describes four main styles: divergent, assimilator, convergent, and accommodator, widely used to understand how individuals learn and process information. The identification of these styles traditionally occurs through the Kolb Inventory (KI), an instrument composed of 12 questions with four alternatives each, which takes approximately 20 minutes to complete.³ Despite its relevance and widespread use, the application time can represent a limiting factor, reducing its applicability in educational contexts that require greater agility. Given this limitation, the need arises for more objective instruments that maintain the reliability of the results. In this context, the use of artificial intelligence presents itself as an innovative strategy for simulating and analyzing response patterns, allowing for the controlled evaluation of educational instruments before their application to real individuals. Objectives:
To analyze the efficiency of the Modified Kolb Inventory (MKI) in identifying learning styles compared to the traditional digitized Kolb Inventory, using artificial intelligence agents configured to simulate different learning profiles. Methodology: This is a comparative study in which artificial intelligence agents were developed and configured according to the four learning styles proposed by Kolb² (1984): convergent, assimilator, divergent, and accommodator. The analysis was conducted on two distinct artificial intelligence platforms, GPT and Gemini, with the aim of verifying the consistency of the results in different technological environments. Initially, the agents were subjected to the traditional digitized Kolb Inventory, ensuring standardization in the application of the instrument. Then, they answered the Modified Kolb Inventory, composed of four questions, developed with the aid of artificial intelligence to reduce the application time without compromising the identification of profiles. The learning styles identified in each instrument were analyzed, as well as the time required to complete each application. This approach allowed for direct comparison between the instruments in a controlled environment, reducing external variables and increasing the reliability of the analysis. The use of artificial intelligence agents enabled the standardization of responses and the reduction of individual biases, favoring comparative analysis between the instruments. Results: The results demonstrated high agreement between the profiles identified by the two instruments, with a compatibility index of 95% between the responses obtained on the different artificial intelligence platforms. This result reinforces the ability of the Modified Kolb Inventory to maintain reliability in identifying learning styles, even with the significant reduction in the number of questions. Furthermore, an average reduction of 63% in the application time of the modified instrument was observed compared to the traditional inventory. The profile analysis showed that all styles presented coherent responses between the two instruments, highlighting a greater affinity of the accommodating profile with the reduced format, possibly due to its characteristic of valuing action and objectivity.⁴ The results also demonstrated consistency between the GPT and Gemini platforms, reinforcing the robustness of the adopted methodology. Conclusion: The Modified Kolb Inventory showed high agreement with the traditional inventory, maintaining effectiveness in identifying learning styles, with a significant reduction in application time. These findings indicate that the instrument has potential for application in educational contexts that demand greater speed and efficiency, without compromising the quality of the assessment. The use of artificial intelligence as a simulation tool has proven to be a relevant strategy for the preliminary validation of educational instruments, contributing to the development of more dynamic methodologies aligned with the contemporary demands of education.¹ Future studies with application in real populations are recommended to consolidate the external validity of the instrument.