Skip to main content Skip to navigation

Transferring Domain Knowledge with (X)AI-Based Learning Systems

Project Overview

The document explores the application of Explainable Artificial Intelligence (XAI) in educational settings, focusing on its role in improving the training of novices in high-stakes fields, such as medical image classification. Through a study assessing XAI's effectiveness, it is demonstrated that XAI enhances knowledge transfer from experts to novices by utilizing example-based learning methods. The findings indicate that learners, regardless of their cognitive styles, experience improved learning performance when XAI is employed. This highlights the potential of generative AI to support personalized learning experiences and bridge gaps in understanding, ultimately leading to better educational outcomes in complex domains.

Key Applications

XAI-based learning system for teaching novices in mammography image classification

Context: Training novices to classify mammography images using a dataset of expert-labeled images

Implementation: Participants were divided into two groups: one receiving only examples and the other receiving examples with explanations generated by XAI.

Outcomes: The study found that participants who received explanations showed improved learning performance, particularly those with a visual cognitive style.

Challenges: The study faced challenges such as potential cognitive overload from too much information and the need for further research on varying explanation modalities.

Implementation Barriers

Technical barrier

The need for sophisticated AI systems that can provide explanations, which can be resource-intensive to develop.

Proposed Solutions: Utilizing existing AI frameworks and collaborating with AI developers to create effective XAI systems.

Cognitive barrier

Differing cognitive styles among novices may affect how they process information from XAI systems.

Proposed Solutions: Tailoring explanations to align with individual cognitive styles to maximize learning effectiveness.

Project Team

Philipp Spitzer

Researcher

Niklas Kühl

Researcher

Marc Goutier

Researcher

Manuel Kaschura

Researcher

Gerhard Satzger

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Philipp Spitzer, Niklas Kühl, Marc Goutier, Manuel Kaschura, Gerhard Satzger

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