AI Symptom Checkers Found Inaccurate, Risking Public Health

The online symptom checkers are AI systems that analyze user input to provide health diagnoses and recommendations. The study highlights that these AI systems often provide inaccurate or misleading information, which can indirectly lead to harm by causing users to misjudge the severity of their health conditions or delay seeking proper medical care. This constitutes indirect harm to health (a), as users may make poor health decisions based on faulty AI outputs. Therefore, this event qualifies as an AI Incident due to the realized harm stemming from the use of AI systems in health diagnosis.

Online symptom checkers are AI systems that analyze symptoms to provide diagnostic suggestions and triage advice. The study reveals that these systems are often inaccurate, with only 36% correct diagnoses and inappropriate emergency advice in many cases. This misdiagnosis and poor triage can cause harm to individuals by delaying proper medical care or causing unnecessary emergency visits, thus impacting health outcomes. The harm is realized and directly linked to the AI systems' outputs. Hence, this qualifies as an AI Incident due to indirect harm to health caused by the AI systems' malfunction or limitations.

The symptom checker apps are AI systems as they use AI algorithms to infer diagnoses and triage recommendations from user input. The study shows these AI systems' use has led to inaccurate diagnoses and triage advice, which can indirectly cause harm to users' health by misleading them about the seriousness of their conditions or delaying necessary care. This constitutes indirect harm to health (a) due to the AI systems' use. Therefore, this event qualifies as an AI Incident because the AI systems' use has directly or indirectly led to harm risks that are realized or highly plausible, as supported by the study's findings and expert caution.

Online symptom checkers are AI systems that generate medical predictions based on user symptoms. The research shows these systems frequently provide incorrect diagnoses and inaccurate advice on seeking medical help, which can lead to harm if users rely on them instead of professional medical consultation. This constitutes indirect harm to health due to the AI system's use, fitting the definition of an AI Incident.

The article explicitly describes symptom checkers as algorithm-based programs that provide diagnostic and triage advice, which fits the definition of AI systems. It reports that these systems often provide inaccurate or misleading advice, which can indirectly lead to harm to individuals' health if they rely on incorrect diagnoses or triage recommendations. Therefore, the event involves the use of AI systems leading to indirect harm to health, qualifying it as an AI Incident under the framework.

The article explicitly discusses AI-based symptom checkers (AI systems) and their performance in diagnosis and triage advice. While no direct harm is reported, the inaccuracies and lack of quality control imply a credible risk of harm to users' health (harm category a). The AI systems' use could plausibly lead to incidents such as misdiagnosis or inappropriate healthcare seeking behavior. Hence, this qualifies as an AI Hazard rather than an AI Incident. It is not Complementary Information because it is not an update or response to a prior incident, nor is it unrelated as it clearly involves AI systems and their impact on health.

The symptom checkers are AI systems that infer diagnoses and triage advice from user symptoms. The study shows these systems produce correct diagnoses only about a third of the time and provide triage advice that is only accurate about half the time, with a tendency toward risk-averse recommendations that may cause unnecessary urgent care visits. This inaccurate output can directly or indirectly harm users' health by misleading them about the seriousness of their conditions or prompting unnecessary medical actions. The harm is realized as users rely on these AI systems for health decisions, fulfilling the criteria for an AI Incident under the framework.

The article involves AI systems (symptom checker apps using AI algorithms) and discusses their use and performance. While it identifies significant risks and potential for harm (misdiagnosis, incorrect triage advice), it does not report a concrete event where these apps have directly or indirectly caused injury, rights violations, or other harms. Therefore, it does not meet the threshold for an AI Incident. It also does not focus primarily on a future plausible harm scenario or a near miss, so it is not an AI Hazard. The article provides important contextual information about AI system performance and risks, which enhances understanding of the AI ecosystem and informs future risk assessment. Hence, it fits best as Complementary Information.

Online symptom checkers are AI systems that analyze user input to provide medical diagnoses. The article states that these systems are inaccurate most of the time, which can indirectly lead to harm by causing misdiagnosis or delayed proper treatment. Although no specific incident of harm is described, the unreliability of these AI systems in medical diagnosis represents a plausible risk of harm to health. However, since the article focuses on the general unreliability and potential danger rather than a specific realized harm event, this qualifies as an AI Hazard rather than an AI Incident.