A Prescriptive Learning Analytics Framework: Beyond Predictive Modelling and onto Explainable AI with Prescriptive Analytics and ChatGPT
Project Overview
The document discusses the transformative role of generative AI in education, particularly through the implementation of a prescriptive learning analytics framework designed to assist at-risk students in higher education. It critiques traditional predictive models for their lack of interpretability and actionable insights, suggesting that these limitations can be addressed by leveraging explainable AI techniques alongside large language models like ChatGPT. By generating personalized, human-readable feedback, the framework aims to enhance student engagement and retention rates, ultimately improving overall academic performance. The integration of these advanced AI technologies not only provides tailored support for individual learners but also empowers educators with deeper insights into student needs, fostering a more effective learning environment.
Key Applications
Prescriptive Learning Analytics Framework using ChatGPT
Context: Higher Education Institutions (HEI) focusing on at-risk students for retention and completion rates
Implementation: Integrating predictive analytics with prescriptive analytics to generate tailored advice based on student data
Outcomes: Enhanced understanding of at-risk learners, actionable insights provided via natural language feedback, improved retention rates
Challenges: Complexity of predictive models leading to distrust; need for transparency and interpretability; reliance on data-driven methods that may not capture all influencing factors
Implementation Barriers
Technical Barrier
Complexity of predictive models leading to 'black-box' characteristics, reducing trust and interpretability. Educators may distrust predictive analytics tools due to lack of clarity.
Proposed Solutions: Employ explainable AI techniques to enhance model transparency and interpretability. Develop clearer communication of how models work and the rationale behind predictions.
Engagement Barrier
Learners may not engage with LA tools if they cannot understand the output.
Proposed Solutions: Ensure outputs are communicated in an understandable manner, potentially using natural language processing.
Project Team
Teo Susnjak
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Teo Susnjak
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai