Online Depression Therapy: Outsourcing Self-assessment 


An increasing number of treatments for depression are being made available online, in an attempt to manage waiting lists and the rising costs of healthcare in the Netherlands. Among the advantages of online therapy are: a reduction of face-to-face time with a therapist, and the opportunity for people to undergo treatment in their own time, at their own pace and in the comfort of their own homes1.  There are also disadvantages in this context. One of the biggest concerns here, is that of the Internet privacy, in consideration of the personal information that is shared via email or video calls. Another, no less important, concern is the difficulty of diagnosing suicidal tendencies since depression is widely associated with suicide (Beskow, 1990). 
According to the media richness theory mediums, such as email and video calls, are less rich and therefore less effective in conveying a message, than interpersonal face-to-face communication (Daft and Lengel (1984). This paper will explore the concept of an online therapy for depression in relation to the information conveyance of a person's emotions. Specifically in the context of self-assessments, which play an important role in monitoring a person's emotional state during treatment. In addition, some suggestions for an experimental setup are being made for testing a software tool that might aid the self-assessment process.
 In general there are two types of online depression therapy: treatment via email or aided self-help courses (van Straten, 2011). This paper will focus on the self-help courses, some of which, are based on the principles of cognitive behavioural therapy. An example of such an Internet therapy is 'Grip op je dip'2 (how to manage a mild depression). This course consists of weekly chat sessions with an expert, homework assignments and 'take home' exercises. To keep track of the emotional state of a person, a weekly self-administered questionnaire, the Edinburgh Depression Scale (EDS), needs to be filled out. The EDS consists of questions about a person's mood and feelings, in certain situations, over the past week. It has been adjusted for online self-administration by Spek, Nyklicek, Cuijpers & Pop (2008). 
 Self-assessments are known to be plagued by bias, which distorts perception and decreases accuracy of the self-assessment test. Two common biases are: the negativity bias, the tendency to overestimate negative feelings, experiences or other kinds of information, compared to their positive counterparts; and the self-serving bias, the tendency to attribute ones successes to internal or personal factors but attributing ones failures to external or situational factors (Campbell & Sedikides, 1999). 
Biases are facilitated by implicit (e.g. unconscious) information processing. Revealing implicit information, by making an effort to make the thought process explicit, reduces the distorting effect of biases on perception. Consequently, it creates opportunity for artificial intervention of the self-assessment process. Unlike the human mind, software, by definition, cannot reason with implicit knowledge. This characteristic could be put to use to avoid biases in current self-assessment methods.  
Emotion detection from facial expressions is an artificial way to obtain explicit information about a person's emotional state. It is considered to be a special form of pattern recognition (Picard, 2000). Facial expression is also asserted to be the most informative channel for machine perception of human emotion (Sebe et al., 2007). This characteristic makes emotion detection from facial expressions especially suitable for obtaining reliable affective data from self-assessments. 
eMotion is a widely used example of emotion-recognition-software (ERS). It creates a three-dimensional map of a person's face using twelve facial markers such as the eyes and corners of the mouth From this map a face-tracking algorithm infers the six universal facial expressions in human beings; happiness, sadness, anger, fear, disgust and surprise. In addition, it can distinguish between positive and negative moods (Martinelli, 2007). eMotion runs on any personal computer and webcam, which makes it an apt candidate for application in online therapies such as 'Grip op je dip'. 
Facial expressions, from a psychological point of view, are a form of nonverbal communication, which conveys a person's emotional, often unconscious, state. In contrast, verbal communication (speaking and writing) is being processed at a conscious level, making the message conveyed explicit. In this respect, the discrepancy between what people say (or write) and what they actually feel might be captured by emotion-recognition-software. If this were the case, software such as eMotion might aid the self-assessment process by increasing accuracy of the conveyed message of a person's self-assessment. To test this an experiment is needed. The assumption would be that eMotion is just as able, or even better, at keeping track of a person's emotional state, compared to the EDS. 
Experimental setups for testing questionnaires such as the EDS already exist. Designing a setup to evaluate eMotion, as the assessor of a person's emotional state, will require a novel approach. For example, not only would the questions be represented by words on a screen, they could just as easily be replaced by audio entirely. In fact, the majority of people are know to be better at expressing their feelings when speaking than they are in writing. This leads to favouring audio over its visual brother, for posing and answering questions. To continue on the path of audio, an avatar could be introduced as a visual cue to ensure the required head pose with regard to the webcam. Additionally, a pop-up window, indicating the remaining time to answer a question, would be welcomed to guide a person while answering a question. Obviously, further options for the experimental setup need to be looked into. 



This paper has explored the concept of online therapy for depression, concerning the self-assessment process that is being used to keep track of a person's emotional state. It has also suggested that modern technology, such as emotion-recognition-software, might increase the accuracy of self-assessment tests, such as the EDS. Finally, integrating eMotion into an online depression therapy such as 'Grip op je dip' could not merely assist the self-assessment process but outsource it, to an artificial agent, all together.




References:
Beskow, J. (1990). Depression and suicide. Pharmacopsychiatry, Vol 23(Suppl 1), Jan 1990, 3-8.

Campbell, W. K., Sedikides, C. (1999). "Self-threat magnifies the self-serving bias: A meta-analytic integration.". Review of General Psychology 3 (1): 23-43.

Daft, R.L. and Lengel, R.H. (1984). Information richness: a new approach to managerial behavior and organizational design. In: Cummings, L.L. & Staw, B.M. (Eds.), Research in organizational behavior 6, (191-233). Homewood, IL: JAI Press.

Martinelli, N. (2007). Emotion-Recognition Software Knows What Makes You Smile.  http://www.wired.com/science/discoveries/news/2007/07/expression_research

Picard, R.W. (2000). Affective Computing. MIT Press, Cambridge, Massachusetts, p166.

Sebe, N., Lew, M.S., Sun, Y., Cohen, I., Gevers, T. and Huang, T.S. (2007).  Authentic facial expression analysis. Image and Vision Computing 25 (2007) 1856-1863.

Spek, V.R.M., Nyklicek, I., Cuijpers, P., & Pop, V.J.M. (2008). Internet administration of the Edinburgh Depression Scale. Journal of Affective Disorders, 106(3), 301-305.

Straten, A. (2011). Internet aided treatment of depression, powerpoint presentating during lecture BASS 2011. 


1 www.gripopjedip.nl