Project Baymax: Designing an AI Companion for Emotional Support and Therapy-Inspired Care
Loading...
Authors
Zelele, Andrias
Date of Issue
2026-05-04
Type
Thesis
Language
en_US
Subject Keywords
Research Subject Categories::TECHNOLOGY::Information technology::Computer science::Computer science , Research Subject Categories::INTERDISCIPLINARY RESEARCH AREAS::Caring sciences::Occupational therapy
Other Titles
Abstract
College students frequently experience stress, anxiety, and loneliness, yet many hesitate to seek immediate support due to stigma or limited access to mental health resources (American College Health Association, 2023; Eisenberg et al., 2007). This study explores the development of an AI-based companion designed to provide low-pressure, emotionally supportive interactions in everyday situations. Inspired by the character Baymax from the Disney animated film Big Hero 6 (Hall & Williams, 2014), the system emphasizes calm, ethical, and non-clinical communication rather than attempting to replace professional care.
The prototype is implemented as a console-based application that integrates emotion detection, intent recognition, and rule-based response generation. A pretrained natural language processing model is used to identify emotional tone in user input, while a lightweight intent classifier distinguishes between emotional expressions and general conversation. A safety module detects high-risk language and redirects users toward appropriate real-world support resources.
Evaluation across 262 simulated user interactions shows that the system can reliably detect basic emotional states, achieving an average confidence score of approximately 0.81 in emotion classification. A hybrid refinement layer reduces low-confidence predictions and improves handling of ambiguous inputs. However, limitations remain in recognizing subtle emotional expressions and preventing misclassification of neutral statements.
These findings demonstrate that interpretable, modular AI systems can provide meaningful emotional support interactions while maintaining ethical safeguards. The project contributes to ongoing efforts to develop accessible and responsible AI tools that complement, rather than replace, traditional mental health support systems.