In higher education, students frequently encounter barriers to accessing straightforward and timely academic advice. These challenges are primarily due to the complexity of institutional systems and the overwhelming volume of information dispersed across various platforms.
Advisors are often tasked with a high number of students and insucient time [5]. A second challenge is to provide enough mentorship with the limited number of available college counselors [11].
Due to these challenges students may nd it dicult to obtain quick answers to fundamental questions like course prerequisites and registration processes. This paper describes the implementation of AI-driven chatbots as a solution to streamline academic advising and improve the overall student experience.
This study proposes that AI chatbots can bridge this gap by providing accurate information on course requirements, prerequisites, and general academic policies which are available in student handbooks and bulletin websites maintained by universities but can be hard to access from multiple sources. Additionally, the research examines how these chatbots can be programmed to recognize complex queries that require human intervention, thereby directing students to appropriate advisors for in-depth guidance on matters such as changing majors or accelerated graduation paths. By analyzing existing literature and preliminary results from the Nittany Advisor tool, this paper aims to demonstrate how AI chatbots can enhance the eciency of academic advising, improve student satisfaction, and allow human advisors to focus on more nuanced aspects of student support.
This study draws on empirical research and case studies [5,8,9] to evaluate the deployment and ecacy of AI chatbots in academic settings. A review of the literature reveals that AI applications in education are not only feasible but also benecial in terms of user experience and operational eciency [2]. Their analysis underscores that AI chatbots can be optimized to respond to a range of student inquiries eectively, providing accurate and instant information while integrating with existing institutional databases [6,10,11].
Utilized eectively, chatbots oer the potential to provide immediate responses to frequent questions such as which classes to take and what prerequisites are required [5], thus reducing student frustration and enhancing their self-suciency in managing academic planning.
Further validation comes from the examination of the Advising Virtual Assistant tool [8], which demonstrates the practical benets of chatbots in navigating nancial aid, course registration, and other administrative tasks. These systems are designed to be available ubiquitously, ensuring that students have continuous access to essential information, thus supporting their academic journey outside traditional advising hours. This ubiquitous availability makes advising for basic information-based questions more accessible, as students no longer need to book appointments weeks in advance or work around class schedules and other commitments.
Moreover, these systems make academic advising information more equitable [4]. First-generation students, or students with lower access to resources, can freely use this tool to answer basic questions without feeling intimidated or overwhelmed. By providing a non-judgmental and easily accessible source of information, chatbots can support equity in academics for all students. [4,11] This paper also explores the technical underpinnings of chatbot implementations. An article by Al-Jedaie et al. [1] provides a comprehensive overview of the system architecture, machine learning algorithms, and natural language processing techniques that enable chatbots to understand and respond to student inquiries accurately. Such technical insights are crucial for educational institutions considering the adoption of AI solutions, as they highlight the importance of proper training data, robust integration capabilities, and ongoing system improvements to maintain high performance and relevance [16].
Similar to other assistant tools discussed thus far [1,5,8,14], our tool: Nittany Advisor has all the benets of academic advising chatbots such as ubiquitous availability, equitable access to information, reducing advisor loads, etc. The Nittany Advisor is dierent for two main reasons: it leverages a custom OpenAI Assistant, and it is trained on a database of all university courses and their inter relations.
The Nittany Advisor chatbot is unique because it is built using the OpenAI assistant API [13], which is ne-tuned with a vector database comprising information from major university sources such as several key academic websites and a comprehensive dataset containing all university-oered courses and their relations. The use of the OpenAI assistant API, rather than the regular GPT-4 API, oers additional benets [13]. The assistant API is specically designed to handle conversational ows more eectively, providing more accurate and context-aware responses [13]. This makes the chatbot more suitable for engaging in complex, multi-turn conversations and allows the chatbot to provide accurate and contextspecic answers tailored to student needs.
Additionally, what sets our approach apart is the integration of a comprehensive database containing all the courses oered by the university, along with the relationships between courses. This includes prerequisites, major or minor requirements, and other academic pathways. With this rich dataset, the chatbot can accurately assist students with degree planning, oering clear guidance about course selection, requirements, and scheduling, providing a personalized advising experience. This ability to cross-reference related courses and requirements allows the chatbot to tackle complex queries and oer more in-depth academic support.
The combination of advanced natural language understanding, a ne-tuned database, and a holistic view of course relationships creates a powerful tool that enhances the eciency and quality of academic advising. By automating responses to routine inquiries and providing highly accurate course and program information, our chatbot ensures that human advisors can devote more time to addressing personalized, complex student needs.
Based on our preliminary investigation involving academic advisors, we uncovered a set of 52 common questions that students frequently ask and that the university recommends new students to ask. Of these, 40% were information-based questions that the chatbot could respond to with high student satisfaction and support with data from the course-relation database and academic resources.
These ndings are supported by other studies [5,7,12] on the subject by Katheeri et al. [5]. The study indicated that 94% of students expressed interest in using a chatbot-based academic advising system, and 88% were condent in its usefulness after interacting with it [1]. The chatbot demonstrated an impressive success rate of 85% across twelve unique tasks [5]. This conservative estimate of ⇠40% reects the fact that while chatbots could potentially manage up to 60% of student queries [3], we aim to account for variability in student needs and institutional contexts.
Along with the list of 52 frequently asked questions (FAQs), we also tested the Nittany Advisor's capability to answer FAQs posted on the university's advising websites. We compared the chatbot's responses to 15 of these FAQs with the university's response from the College of Engineering and College of IST websites. The chatbot's responses contained nearly ve times more words than the university's FAQ responses (480% "), mainly for questions regarding procedural information, such as course prerequisites and registration processes, indicating that chatbots excel in oering expanded guidance on processes that students often nd confusing.
The similarity scores between FAQ responses and chatbot answers were low (17.6%), indicating that the chatbot provided more personalized responses rather than reiterating standard FAQ content. The Nittany Advisor's responses averaged 2, 323 characters versus the FAQ's 487 ( E6. = 1836.47 characters or ⇠ 74%), this illustrates that the chatbot's responses provided detailed information likely contributing to increased student satisfaction.
Given these promising preliminary results, it is essential to conduct further studies to assess the impact on actual student advising outcomes and address the challenges of these systems. Moving forward, we plan to conduct blind user studies to validate whether these ndings translate to increased student satisfaction.
In conclusion, AI chatbots hold promise for transforming academic advising. By addressing routine questions and streamlining access to crucial information, these systems enhance institutional efciency and elevate the student experience. These systems provide ubiquitous solutions for advising, making academic information equitable and easily accessible to all students. This paper highlights the importance of a hybrid or mixed-method advising model [15] that leverages the strengths of both AI and human advisors, fostering a more responsive and supportive educational environment. However, more work is required in this area, and with generative AI models improving rapidly, the advising landscape holds exciting potential for further advancement.