OpenAlex · Aktualisierung stündlich · Letzte Aktualisierung: 18.03.2026, 12:05

Dies ist eine Übersichtsseite mit Metadaten zu dieser wissenschaftlichen Arbeit. Der vollständige Artikel ist beim Verlag verfügbar.

The Effectiveness of Specialised Conversational AI in Preliminary Medical Triage: A Comparative Analysis of Non-Maleficence-by-Design and General-Purpose Architecture

2025·0 Zitationen
Volltext beim Verlag öffnen

0

Zitationen

3

Autoren

2025

Jahr

Abstract

This paper evaluates the safety and effectiveness of Large Language Models (LLMs) in preliminary medical triage through a Structured Comparative Analysis of 20 high-risk clinical inquiries, comparing an unconstrained Google Gemini Advanced (GPM) model with a structurally constrained Med-SafeTriage Prototype (SDM). Findings demonstrate that unconstrained LLM architecture is fundamentally incompatible with patient-safety requirements: GPM showed a 55% safety-adherence failure rate and highly unstable response times, reaching 11.05 seconds, indicating that complex reasoning slows critical decision-making. In contrast, the SDM delivered consistent performance with a 2.16-second response time and 95% safety adherence, underscoring that constrained, safety-optimized architectures better satisfy clinical triage demands.

Ähnliche Arbeiten

Autoren

Institutionen

Themen

Artificial Intelligence in Healthcare and EducationClinical Reasoning and Diagnostic SkillsMachine Learning in Healthcare
Volltext beim Verlag öffnen