WISE Causal Models: Wisdom Infused Semantics Enhanced Causal Models - A Study in Suicidality Diagnosis

Author(s)

Kaushik Roy, University of South Carolina - ColumbiaFollow
Yuxin Zi, University of South Carolina - ColumbiaFollow
Vignesh Narayanan, University of South Carolina - ColumbiaFollow
Manas Gaur, University of Maryland Baltimore CountyFollow
Sanjay Chandrasekar
Amit Sheth, University of South Carolina - ColumbiaFollow

Document Type

Article

Abstract

The COVID-19 Pandemic has highlighted the gap between the number of mental health care seekers and care providers. Netizens have taken to internet-based platforms such as Reddit to express their experiences. Mental illness diagnosis processes have clinically accepted causal interpretations and semantics. Curiously, mental illness diagnosis accuracy is low relative to similar well-studied illnesses. Motivated by this discrepancy, we propose Wisdom Infused Semantics Enhanced (WISE) causal models, inspired by the wisdom of the crowd idea that learns from a collective agreement among causal models and their semantics for mental illness diagnoses. We use suicidality diagnosis task descriptions, datasets, and baseline methods to evaluate the effectiveness of WISE causal models. Our experiments show that learning WISE causal models improve performance on these tasks.

Publication Info

Preprint version 2022.

© The Authors, 2022

APA Citation

Roy, K., Zi, Y., Narayanan, V., Gaur, M., Chandrasekar, S., & Sheth, A. (2022). WISE Causal Models: Wisdom Infused Semantics Enhanced Causal Models - A Study in Suicidality Diagnosis. [Preprint]