Automated Handling of Suicidal Tendencies: A Comparative Analysis of the Performance of Mental Health Support Platforms (Chatbots/AI) Abby and Wellzy in Assessing Graduated Risk Levels

Abstract

This study aims to define the boundaries of clinical perception and conduct a critical evaluation of crisis response sensitivity of two AI mental health platforms (Chatbots), Abby and Wellzzy, by analyzing their initial, automated responses to messages conveying suicidal ideation across a severity spectrum from direct declarations to oblique hints. To systematically navigate this inquiry, the methodology involved crafting three concise, opening-message scenarios which were submitted to each platform, followed by a rigorous analysis of every response against a comprehensive framework, as this framework evaluated critical dimensions including the systems' ability to identify both explicit and implicit warning signs, their response latency, the nature of support offered—categorized as either emotional consolation or concrete referral—alongside the practical utility of the guidance and its suitability for a global user base.
And for greater precision in the analysis, the responses from the two platforms were analyzed according to the RUD scale (Risk, Urgency, Dangerousness). The results unveiled a complex landscape characterized by both unsettling differences and superficial similarities between the platforms, yet underscored profound and systemic shortcomings common to both, as while basic recognition of explicit keywords was evident, the analysis revealed a frequent failure to adequately interpret implicit distress cues, with responses often defaulting to generic, scripted empathy that lacked appropriate urgency. And a complete lack of alignment with the criteria of the RUD scale.