Joachim Diederich Psychology Network Pty Ltd School of Information Technology and Electrical Engineering, University of Queensland. Centre for Mental Health, University of Melbourne.
Mental health problems are responsible for significant personal, economic and social burden globally yet they remain poorly diagnosed in a primary health care setting, primarily due to the subjective nature of conventional diagnostic methods (e.g. questionnaires and interviews). The availability of objective methods based on the computational analysis of behaviour (e.g. speech and language) in order to screen for a range of mental health problems therefore meets a widely recognised need and results in significant benefits.
Beyond questionnaires, observer-rating scales and clinical interviews, there are hardly any methods that identify mental health conditions that result in an impairment in the recognition of emotional states. Such a condition is Alexithymia; the inability to identify and describe emotions. Alexithymia is not a recognised disorder, however, the condition overlaps significantly with Aspergers and autism spectrum disorder. It is also possible that frequently observed psychological problems such as anxiety and depression result in distraction and withdrawal including a lack of awareness of emotional expressions. The Mental State Tracker aims at identifying emotional states and to make this information available for clinical practice.
The Mental State Tracker is a mobile application available for Android, Apple and Windows devices. This mobile app is based on artificial intelligence (AI) techniques and combines a number of features: (1) The recording and analysis of speech to detect a number of emotional states. (2) The transcription of speech and the analysis of the resulting texts by various methods to determine emotional issues. (3) The analysis of transcribed speech includes the determination of suicide risk. (4) Standard questionnaires that are commonly used in clinical psychology and psychiatry are also offered and the results are compared with speech and text analysis.
It is possible to record unrestricted speech of several minutes in length and then to receive a detailed report on the distribution of positive and negative emotion words as well as expressions that indicate excessive self-referral. Current research focusses on chatbots that can be used inside the app to provide psychoeduction once an impairment has been identified.