This technical note presents a conceptual and ethical analysis of the use of voice-based artificial intelligence systems in mental health contexts, particularly regarding attempts to infer depression from vocal signals. Based on extensive research and observation, the author highlights that voice is one of the most unstable and error-prone manifestations of human functioning, being highly sensitive to context, environment, physical condition, and transient emotional states. From the perspective of the Theory of Fundamental Belief (TCF/TFB), voice appears as a late expression of functioning and cannot be treated as a structural indicator of mental health conditions. Attempts to use voice as a primary signal for diagnosis risk inverting the order of functioning and may lead to serious methodological and ethical errors. The author, a Brazilian researcher, emphasizes that current research trends, including initiatives observed in Brazil, must clearly distinguish between signal detection and clinical diagnosis. This note does not oppose technological research, but calls for caution, proper framing, and responsibility to prevent misuse and harm. This material is explanatory and non-prescriptive, intended to clarify structural limits rather than propose methods or diagnostic tools.
CHRISTIAN MONTGOMERY (2026) studied this question.