One word can change it all: GenAI places an emphasis on entered pseudoscientific terms for patients seeking advice on urinary tract infections

Abstract

There are an increasing number of identified benefits for employing generative artificial intelligence (GenAI) in the presentation of health information. For patients, GenAI is particularly appealing as it presents information in a way that is digestible and understandable to a broad audience. This includes tools such as ChatGPT, Gemini, Claude, Deepseek, and Copilot. However, the practice runs the risk of perpetuating misinformation and presenting potentially hazardous advice. With the rising development of antimicrobial resistance in urinary tract infections (UTIs), this disorder presents an area where it is important to ‘get the message right’. In some cases, a GenAI response to a health query can be considerably influenced by the addition of a single word to the user’s entered prompt. In particular, this study identified particular words of concern related to UTI treatments, such as ‘homeopathy’, ‘crystals’ and ‘star sign’. When a single pseudoscientific term (usually only a single word) was entered, the response tended to emphasise this, returning misinformation. This article provides insights into where concern should be taken if patients mention using GenAI for their health advice. The outcome is that health professionals should be specifically aware of the tendency for GenAI to emphasise pseudoscientific concepts if prompted, and its tendency to present them as effective interventions for UTIs. When counselling patients, a discussion of this concept would be necessary to clarify the limitations of this technology in health education.

Ähnliche Arbeiten