Towards an Index of Mental Wellbeing in Language The relationship between time orientation, self-focus and mood during prolonged bed-rest

摘要

Monitoring the mental wellbeing and psychosocial states of human operators in safety-critical domains is key to improving crew performance, safety and security. However, acquiring objective data and insights to mental wellbeing and psychosocial states among human operators is difficult and has relied on the subjective reports of operators, which are prone to biases and distortions. Team communications in a broad range of contexts from business organizations to high criticality workplaces such as emergency response, airplane pilots and spaceflight could be significantly improved with quick access to reliable and objective data about the psychosocial health of a team. Data on the psycho-social dimensions of collaborative teams, such as social distance, power dynamics, affect, and a team's comfort working together are typically highly subjective, not readily computationally tractable, and are collected using self-reports such as think-aloud protocols or surveys that can confound the behaviors being studied. By developing a method to collect data using non- or minimally intrusive methods requiring low participant effort coupled with automated data processing, we unshackle researchers from the burdens of hand-coding raw data and enable them to make empirically based discoveries more rapidly. This paper presents a validated, cost-effective and fast alternative to the shortcomings of current assessment methods. We present the results from an automatic text analysis tool applied to large amounts of written text (i.e., journals kept by participants in a bed rest study) in order to identify topics of interest, the emotional valence (positivity or negativity) of topics, as well as changes in these metrics over time. These topics and aspects of the text were identified computationally and automatically. This research was performed on different groups of subjects participating in NASA analog studies, where the primary goal of our investigation was the identification of changes to psychosocial states. Our results show that it is possible to predict mood based on journal entries alone using Latent Semantic Analysis and that we are able to identify non-conscious variables impacting well-being over time.