Researchers from the universities of Southern California and Utah have devised a unique computer programme which by analysing the speech of couples can predict whether their marriage or relationship will improve, stay the same, or worsen. The analysis will focus on qualities such as pitch and intensity and compare the results with human behavioural assessments by taking into account familiar features of marital dynamics, such as blame.
The programme is an extended part of a research paper; "The Role of Acoustic Features in Predicting Marital Outcome" where researchers have concluded that acoustic features capture more relevant information than the human-encoded behavioural constructs for predicting marital outcome. The aim of having such a computer-assisted technology to indicate results can prove to be beneficial to the advancement of behavioural outcome studies in psychology in assessing quality of care and treatment methods, according to the research.
The therapy will also be able to pick up speech patterns such as loudness, jitter or shimmer based on which one can measure the shakiness in one's voice. The premise for designing the programme was based on useful information from closely examining patients' voices, gestures and vocabulary in order to make talking therapy more effective.
The scientists worked with video recordings of close to 134 distressed couples who had been married for an average of 10 years and sought therapy for marital problems. The examinations was then divided into three periods: one before therapy began, another after 26 weeks of therapy and a third after two years of treatment. In the videotaped interactions words and body language were also rated for various characteristics.
The computer ratings were then compared to manual ratings by therapists with four possible outcomes -- decline, no change, partial recovery and recovery. The computer-generated results in fact fared a bit better across the board in predicting marital changes with 78% accuracy, beating the human ratings by two percentage points.