🎧 AI Сan Detect Loneliness from a Person's Voice

A lonely person can be identified by the way they speak, according to an experiment by researchers in Germany.

First, 96 volunteers were assessed for their subjective sense of loneliness using the UCLA Loneliness Scale questionnaire. The researchers then recorded their speech in two types of tasks: describing a picture and talking about their personal life.

Using this data, the team trained an AI model that analyzes 78 acoustic features in speech, including rhythm, pauses, intonation, and voice quality, and predicts whether a person is lonely. In the end, the AI was able to detect loneliness from a combination of many weak signals—what the person is actually saying did not matter to the algorithm.

👨 What do lonely people sound like?

Among lonely women, speech was marked by prolonged pauses and uneven volume. Among men, indicators included, for example, a higher-pitched voice and shorter pauses between syllables.

At the same time, the effect appeared only in the picture-description task. In free personal narratives, the markers of loneliness disappeared. The researchers believe that strong emotions may override these subtle cues.

@hiaimediaen