Addressing Unhappiness in Elderly Care: Challenges in Causal AI and Span Categorization Solutions
Asmae Elidrissi, My Abdelouahed Sabri · 2024
Causal Artificial Intelligence (AI) represents a transformative advancement in AI research, focusing on the identification and understanding of cause-and-effect relationships between variables. Unlike traditional AI, which primarily identifies patterns and correlations, causal AI aims to explain how specific events influence one another. This study explores the application of causal AI in Natural Language Processing (NLP) to develop a model capable of extracting the root causes of unhappiness among elderly residents in retirement homes from dialogue data.Our research reveals that the current state of causal AI in NLP is insufficient for effectively identifying these root causes. Consequently, we employed an alternative approach using span categorization, which allowed for a more precise analysis of textual spans within dialogues. This paper outlines our methodologies, from the initial exploration of causal AI to the challenges encountered and the eventual shift to span categorization.Our findings highlight current limitations in causal AI for NLP applications and offer insights into practical solutions for understanding the emotional states of elderly individuals through dialogue analysis. This work contributes to the broader field of AI-driven social support systems and underscores the need for further advancements in causal AI to fully realize its potential in complex NLP tasks.