Feasibility of AHU Nebras™: A No-Code Customized GPT for Graduate Student Support in Academic and Administrative Regulations

Abstract

Accessing institutional policies in higher education can be cumbersome; that is particularly true whenever digital systems are underdeveloped. Customized GPTs, which represent generative AI tools, might offer a promising solution for improving information accessibility. The current study aims to explore the design and feasibility evaluation of Nebras, a customized GPT created to support graduate students in retrieving accurate university regulations and procedures from the Deanship of Scientific Research and Graduate Studies in a university in Jordan called Al-Hussein Bin Talal University (AHU). The GPT was designed using OpenAI’s platform without using any coding. The procedure of design and development of the Nebras was presented. In addition, to evaluate its performance, a set of 100 real-world questions was submitted to Nebras. The questions were generated from frequently asked real-life inquiries collected from Deanship staff, reflecting the most common issues they face daily. The accuracy of the responses was checked on a scale from 1 (very poor) to 5 (excellent) from a panel of reviewers. The results showed a high mean score for the accuracy of the responses (M = 4.10, SD = 1.32), indicating that the system’s responses were rated as Good or Excellent. A further analysis of the accuracy of the responses showed that the reasons for the scores between 1 to 3 were categorized into three types: ones related to the quality or types of the questions (prompts), ones related to the quality and comprehensiveness of the content of the provided institutional documents loaded to the customized GPT, and ones related to the GPT itself. AHU Nebras shows promising potential as an accessible, AI-powered tool for enhancing regulatory support in higher education institutions, particularly within developing countries. Building on these feasibility results, the study recommends enhancing the quality and the comprehensiveness of the set of the instruction feed up to the system based on reviewed academic and administrative regulations, accompanying the pilot launch of the custom GPT with simple manual of use and short training for the potential users, and conducting a pilot study to evaluate AHU Nebras’s effectiveness and user experience within real academic contexts.