Skip to main content Skip to navigation

Human Empathy as Encoder: AI-Assisted Depression Assessment in Special Education

Project Overview

The document presents the Human Empathy as Encoder (HEAE) framework, an innovative approach in the field of education that integrates generative AI and teacher empathy to assess depression in special education environments. By incorporating student narratives with a structured 'Empathy Vector' based on teachers' insights, this framework seeks to enhance the accuracy of depression assessments through a collaborative relationship between human judgment and AI analysis. The HEAE framework addresses key ethical considerations, such as data privacy and the complexities involved in interpreting student narratives, while aiming to improve the assessment process. Overall, the use of generative AI in this context demonstrates a commitment to leveraging advanced technology in a way that respects and incorporates human perspectives, ultimately striving to provide better support for students with special educational needs and to promote mental health awareness within educational settings.

Key Applications

Human Empathy as Encoder (HEAE)

Context: Assessing student depression in special education environments

Implementation: The HEAE framework integrates teacher-derived empathy insights into the AI analysis process, using a custom annotation pipeline to generate training data.

Outcomes: Achieved 82.74% accuracy in classifying depression severity into 7 levels. Enhanced ethical responsibility and transparency in assessments.

Challenges: Difficulty in capturing the full richness of teacher empathy and variability across different educators; potential privacy concerns with student data.

Implementation Barriers

Technical Barrier

Challenges in interpreting vague narrative texts for signs of depression and the limitations of existing automated methods.

Proposed Solutions: Development of a multimodal approach that integrates teacher insights into the AI process to enhance accuracy.

Privacy Barrier

Concerns regarding the privacy of sensitive student data when using large language models.

Proposed Solutions: Implementing a privacy-by-design approach that minimizes data transmission and maintains local processing.

Project Team

Boning Zhao

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Boning Zhao

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