Skip to main content Skip to navigation

From {Solution Synthesis} to {Student Attempt Synthesis} for Block-Based Visual Programming Tasks

Project Overview

The document examines the integration of generative AI in education, emphasizing its application in block-based visual programming for novice learners. It identifies the common challenges these students encounter in programming tasks and proposes AI-driven programming tutors as a viable solution for delivering personalized support. To assess the effectiveness of various AI methodologies in understanding student behavior and generating programming attempts, the paper introduces a benchmark known as StudentSyn. Findings reveal that while certain AI techniques, such as NeurSS and SymSS, demonstrate potential in improving educational outcomes, there is still a considerable performance disparity when compared to human experts in teaching contexts. Overall, the document underscores the promise of generative AI in enhancing educational experiences while also highlighting the need for further advancements to bridge the gap between AI systems and human instructional effectiveness.

Key Applications

Student modeling and attempt synthesis using AI-driven tutors

Context: Block-based visual programming tasks for beginners, particularly in educational settings such as Hour of Code initiatives.

Implementation: The implementation involves creating a benchmark (StudentSyn) to synthesize student attempts based on reference tasks and observed student behavior.

Outcomes: The AI-driven models aim to predict student misconceptions and improve personalized feedback through synthesized programming attempts.

Challenges: Challenges include data scarcity, variability in student behavior, and the complexity of synthesizing student attempts accurately.

Implementation Barriers

Data Scarcity

Limited availability of real-world datasets for student attempts in block-based visual programming.

Proposed Solutions: Creating synthetic datasets and using techniques that can operate effectively with limited data.

Variability in Student Behavior

High variability in the way students approach programming tasks, leading to diverse and unpredictable attempts.

Proposed Solutions: Developing sophisticated models to capture and predict diverse student behaviors and misconceptions.

Project Team

Adish Singla

Researcher

Nikitas Theodoropoulos

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Adish Singla, Nikitas Theodoropoulos

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

Let us know you agree to cookies