Using Artificial Intelligence (AI) As An External Examiner

Dr.Tayyaba Azhar; Kinza Aslam; Zakia Saleem; Ahsan Sethi; Tahseen Fatima

doi:10.51273/esc23.251319323

Authors

Tayyaba Azhar FMH college of Medicine and Dentistry, Lahore
Kinza Aslam
Zakia Saleem The University of Lahore
Ahsan Sethi
Tahseen Fatima

DOI:

https://doi.org/10.51273/esc23.251319323

Keywords:

AI, ChatGPT, Automatedscoring, human scoring

Abstract

Objective: To access the validity of ChatGPT on AI assisted tool for evaluating essay questions.
Material and Methods: This was a cross-sectional quantitative study conducted at University College of
Medicine and Dentistry from June till August 2023. Eighteen questions were selected from fifteen exit tests
of Certificate in HPE course. Each of the answers were independently graded by two assessors with doctorate
in HPE. The same answers were then reevaluated using ChatGPT. The inter-rater reliability was determined
using Kappa test.

Results: The agreement between ChatGPT and examiner scores varied on various items. Weak agreement was observed for questions 8 and 9, moderate agreement for questions 2, 3, and 5, and strong kappa agreement
for questions 1, 4, 6, and 7.

Conclusion: Artificial intelligence assisted tools such as ChatGPT is a reality but its use in assessing essay questions would require massive training data from expert assessors. Once appropriately trained, it may replicate assessment decisions across the full range of subject.