Brian Schwartz

Hiba Arnaout, Anmol Goel, H. Andrew Schwartz et al.

Although artificial intelligence (AI) shows growing promise for mental health care, current approaches to evaluating AI tools in this domain remain fragmented and poorly aligned with clinical practice, social context, and first-hand user experience. This paper argues for a rethinking of responsible evaluation -- what is measured, by whom, and for what purpose -- by introducing an interdisciplinary framework that integrates clinical soundness, social context, and equity, providing a structured basis for evaluation. Through an analysis of 135 recent *CL publications, we identify recurring limitations, including over-reliance on generic metrics that do not capture clinical validity, therapeutic appropriateness, or user experience, limited participation from mental health professionals, and insufficient attention to safety and equity. To address these gaps, we propose a taxonomy of AI mental health support types -- assessment-, intervention-, and information synthesis-oriented -- each with distinct risks and evaluative requirements, and illustrate its use through case studies.

Shaina Raza, Brian Schwartz

There are a few challenges related to the task of biomedical named entity recognition, which are: the existing methods consider a fewer number of biomedical entities (e.g., disease, symptom, proteins, genes); and these methods do not consider the social determinants of health (age, gender, employment, race), which are the non-medical factors related to patients' health. We propose a machine learning pipeline that improves on previous efforts in the following ways: first, it recognizes many biomedical entity types other than the standard ones; second, it considers non-clinical factors related to patient's health. This pipeline also consists of stages, such as preprocessing, tokenization, mapping embedding lookup and named entity recognition task to extract biomedical named entities from the free texts. We present a new dataset that we prepare by curating the COVID-19 case reports. The proposed approach outperforms the baseline methods on five benchmark datasets with macro-and micro-average F1 scores around 90, as well as our dataset with a macro-and micro-average F1 score of 95.25 and 93.18 respectively.