AI VERSUS HUMAN GRADERS: ASSESSING THE ROLE OF LARGE LANGUAGE MODELS IN HIGHER EDUCATION DOI Open Access
Mahlatse Ragolane, Shahiem Patel,

Pranisha Salikram

и другие.

Innovare Journal of Social Sciences, Год журнала: 2025, Номер unknown, С. 1 - 10

Опубликована: Янв. 1, 2025

While artificial intelligence (AI) grading is seeing an increase in use and adoption, traditional educational practices are also forced to adapt function together with AI, especially assessment grading. In retrospect, human grading, on the other hand, has long been cornerstone of assessment. Conventionally, educators have assessed student work based established criteria, providing feedback intended support learning development. offers nuanced understanding personalized feedback, it subject limitations such as inconsistencies, biases, significant time demands. This paper explores role large language models (LLMs), ChatGPT-3.5 ChatGPT-4, processes higher education compares their effectiveness that methods. The study uses both qualitative quantitative methodologies, research extends across multiple academic programs modules, a comprehensive how AI can complement or replace graders. 1, we focused (n=195) scripts (n=3) modules compared GPT 3.5, 4, Manually marked exhibited average 24% mark difference. Subsequently, (n=20) were using GPT-4, which yielded more precise evaluation. Total 4% difference results. There individual instances where marks higher, but this could not naturally be marker judgment. Study 2, results from first highlighted need for memorandum; thus, identified (n=4341), among (n=3508) used. found remains efficient when memorandum well-structured. It was while excels scalability, graders excel interpreting complex answers, evaluating creativity, picking up plagiarism. 3, evaluated formative assessments 4 (statistics n=602, Business Statistics n=859 Logistics Management n=522). third demonstrated marking tools effectively manage demands assessments, particularly questions objective structured, Management. initial error 102 importance well-designed memorandum. concludes reduce burden should integrated into hybrid model markers systems tandem achieve fairness, accuracy, quality assessments. contributes ongoing debates about future by emphasizing well-structured discretion achieving balanced effective solutions.

Язык: Английский

The Implementation of Artificial Intelligence in South African Higher Education Institutions: Opportunities and Challenges DOI Creative Commons
Shahiem Patel, Mahlatse Ragolane

Technium Education and Humanities, Год журнала: 2024, Номер 9, С. 51 - 65

Опубликована: Июль 25, 2024

This paper examines the strategic implementation of Artificial Intelligence (AI) in South African Higher Education (HE) institutions and its potential opportunities challenges. It posits that AI can significantly enhance educational outcomes administrative efficiency HE institutions, but successful integration necessitates addressing infrastructure limitations, ethical concerns, frameworks. The study employs a qualitative research methodology using secondary sources. Findings reveal substantial benefits, such as improved efficiency, personalized learning, data-driven decision-making, often impeded by challenges like inadequate infrastructure, socio-economic disparities, issues related to data privacy algorithmic bias. importance planning frameworks, AI8-Point Model, is emphasized for effective HE. Recommendations include investing technological developing policies adopting Collaboration among policymakers, educators, technology providers essential navigate complexities operational

Язык: Английский

Процитировано

6
AI VERSUS HUMAN GRADERS: ASSESSING THE ROLE OF LARGE LANGUAGE MODELS IN HIGHER EDUCATION DOI Open Access
Mahlatse Ragolane, Shahiem Patel,

Pranisha Salikram

и другие.

Innovare Journal of Social Sciences, Год журнала: 2025, Номер unknown, С. 1 - 10

Опубликована: Янв. 1, 2025

While artificial intelligence (AI) grading is seeing an increase in use and adoption, traditional educational practices are also forced to adapt function together with AI, especially assessment grading. In retrospect, human grading, on the other hand, has long been cornerstone of assessment. Conventionally, educators have assessed student work based established criteria, providing feedback intended support learning development. offers nuanced understanding personalized feedback, it subject limitations such as inconsistencies, biases, significant time demands. This paper explores role large language models (LLMs), ChatGPT-3.5 ChatGPT-4, processes higher education compares their effectiveness that methods. The study uses both qualitative quantitative methodologies, research extends across multiple academic programs modules, a comprehensive how AI can complement or replace graders. 1, we focused (n=195) scripts (n=3) modules compared GPT 3.5, 4, Manually marked exhibited average 24% mark difference. Subsequently, (n=20) were using GPT-4, which yielded more precise evaluation. Total 4% difference results. There individual instances where marks higher, but this could not naturally be marker judgment. Study 2, results from first highlighted need for memorandum; thus, identified (n=4341), among (n=3508) used. found remains efficient when memorandum well-structured. It was while excels scalability, graders excel interpreting complex answers, evaluating creativity, picking up plagiarism. 3, evaluated formative assessments 4 (statistics n=602, Business Statistics n=859 Logistics Management n=522). third demonstrated marking tools effectively manage demands assessments, particularly questions objective structured, Management. initial error 102 importance well-designed memorandum. concludes reduce burden should integrated into hybrid model markers systems tandem achieve fairness, accuracy, quality assessments. contributes ongoing debates about future by emphasizing well-structured discretion achieving balanced effective solutions.

Язык: Английский

Процитировано

0