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
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
Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI
2019 · 8.250 Zit.
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
2019 · 8.109 Zit.
High-performance medicine: the convergence of human and artificial intelligence
2018 · 7.482 Zit.
Proceedings of the 19th International Joint Conference on Artificial Intelligence
2005 · 5.776 Zit.
Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI)
2018 · 5.434 Zit.