Beyond group differences: specificity of nonverbal behavior and interpersonal communication to depression severity

ABSTRACT

Depression is one of the most prevalent mental health disorders and a leading cause of disability worldwide. AVEC 2013 heralds the first systematic effort to detect presence of depression from nonverbal behavior. This keynote addresses three related issues. Specificity. Are differences between depressed and non-depressed persons specific to depression or are they common to the types of people most likely to become depressed? Depression is strongly related to stable individual differences in neuroticism, introversion, and conscientiousness. Differences in nonverbal behavior between those with and without depression could indicate personality differences rather than depression. Do they?

Functions. What can non-verbal behavior tell us about possible functions or mechanisms of depression? Two alternative hypotheses are Affective Dysregulation and Social Risk Avoidance. To contrast these hypotheses, fine-grained analyses of facial expression are needed that can distinguish between displays of negative emotion. In particular, between negative displays that elicit approach or affiliation (e.g., sadness) and those that elicit avoidance (contempt and disgust).

Interpersonal effects. Early work proposed that depression has strong interpersonal effects. Recent work in psychopathology has tended to neglect the possible effects of depression on interaction partners and the influence of context. Does context matter for depression detection? How might depression negatively impact interaction partners?

In this keynote, I explore these issues from the vantage of longitudinal research in depression. The findings suggest that nonverbal behavior in depression can be automatically measured, is highly specific to severity of depression, and is a strong indicator of change over the course of treatment. Avoidance of social risk appears to be a critical function of depression. Depression strongly impacts the actions of others. Automated detection of depression may be optimized by exploiting social context, paying careful attention to displays of affiliation and risk aversion or rejection and the communicative displays of interaction partners.

Index Terms

• Applied computing

  • Law, social and behavioral sciences

    • Psychology