Abstract
Ensuring ethical compliance in research proposals submitted by students to academic institutions is essential. However, manually assessing these proposals for informed consent requirements can be time-consuming for research advisers and ethics reviewers. Most existing AI systems that assess ethical compliance focus on legal and financial applications rather than on evaluating informed consent requirements for academic research proposals. To address this gap in the literature, this paper presents the development of an AI system that assesses informed consent requirements in research proposals submitted by students to academic institutions, using a hybrid natural language processing approach. This system accepts research proposals in DOCX and PDF document formats, extracts the sections of the proposals that deal with ethical considerations of the proposed research, and utilizes both rule-based and machine learning techniques to determine whether the requirements of MC-Res Form 16 are satisfied in the submitted research proposals. The machine learning approach uses techniques such as TF-IDF feature extraction and a Support Vector Machine (SVM) classifier to determine whether a proposal includes informed consent requirements, either explicitly or implicitly stated. When combined, these two methods more accurately determine whether a research proposal meets all informed consent requirements. The system was evaluated on 20 research proposals from students, containing 180 individual items to evaluate for informed consent requirements. The system achieved an accuracy of 0.90, precision of 0.96, recall of 0.93, and an F1-score of 0.94 in identifying informed consent requirements in these research proposals. Furthermore, the system provided explanations for its decisions, thereby improving its interpretability for researchers, their advisers, and academic institutions overall.
