Social Connection and Anonymity in Health Games
By James Watt, Ph.D.
Professor Emeritus, Communication Sciences; Affiliate Researcher, Center for Health, Intervention, and Prevention; University of Connecticut; Storrs, CT
Anyone who has tried to maintain a regular exercise routine, read a series of interesting but somewhat difficult books, develop an academic thesis, or do some other valuable but not always pleasant activity knows the value of social connections in helping one to persist. That’s why running buddies, book clubs, and faculty advisors exist. It’s harder to stay in bed when your running buddy is waiting at the park or to avoid reading when your book club meeting is coming up. Social networks can also provide group support for difficult tasks or emotional situations. Alcoholics Anonymous (AA) is a prominent example. Smoking cessation groups, medical condition affinity groups like those made up of cancer patients, and other on- and off-line networks provide social support that let individuals know “you’re not in this alone.”
The potential value of social interaction in health games is clear. Positive social pressure and support from a group may be very useful in encouraging healthy behaviors like exercise and healthy eating (Fraser & Spink, 2002) and convincing health games players to continue the game even if they are not doing well. But effective groups depend on honest self-disclosure and emotional connection among members to establish the cohesion that gives them their power. Unfortunately, self-disclosure comes with a risk when communicating about behavior or conditions that are painful or socially negative. To label oneself as an alcoholic, or as obese, or as a cancer patient opens the possibility of being subjected to teasing or bullying or exploitation. In a game, even playing poorly exposes the player to the dangers of ridicule. This fear of negative consequences works against the free communication that makes social interaction effective.
The usual response to this problem is to provide anonymity to group members. AA, for example, makes anonymity of members a central tenet, and many on-line support groups discourage posting of any information that can be used to identify individuals. But anonymity can have negative consequences, too (see Christopherson (2007) for a review of some major issues). For example, while anonymity has been shown to increase group participation and cohesion, it also may increase strong negative language and personally offensive “flaming.”
A typical method for providing anonymity in games is to represent the players with an avatar that protects their actual identity while still representing them with a unique surrogate identity. This can be a visual graphic or picture or even something as simple as a distinctive text “screen name.” (More information about avatars and their potential impact on game design is presented by Nick Yee in The Proteus Effect, another Research Brief on this web site.
Despite its potential for negative side-effects, the power of group interaction to encourage and sustain positive behaviors has convinced some health game designers to incorporate social interaction into their games. The Health Games Research program has funded several projects that use social interaction (albeit without anonymity) as motivational devices. Examples are the Union College project, Seniors Cyber-Cycling with a Virtual Team: Effects on Exercise Behavior, Neuropsychological Function and Physiological Outcomes, and the Cornell University project, Mindless Eating Challenge: Persuasive Mechanisms in Mobile Health Games.
An important research question for health game designers is whether the positive social effects of group communication can be sustained while still providing health game players with the benefits of anonymity in the form of an avatar. Having a player represent him- or herself with an avatar may reduce the player’s risk in disclosing sensitive information to the group and thus encourage him or her to share personal information (Joinson, 2001). But it may also attenuate a player’s feeling of social connectedness with others in their group. Communicating with an avatar is just not the same as communicating with the person behind the avatar. In particular, graphical avatars generally don’t provide important nonverbal information from facial expression, body movement, and other modes of expression. This leads to “emotional distancing” which decreases the benefit of social interaction.
We have been investigating the impact of anonymity provided by avatars on emotional communication and self-disclosure. In a recent study done at the Institute for Creative Technology, University of Southern California (Kang, Watt,& Gratch, 2009), researchers found that increasing the levels of anonymity from none (full visual identification), to intermediate (a graphical avatar), to full (providing no visual representation at all), had two opposing effects: it decreased the sense of emotional connection while at the same time increasing the amount of self-disclosure of intimate information that an individual was willing to share. The diagram below shows these effects and is explained in the Notes About the Statistics at the end of this Brief.
The diagram shows that increases in anonymity simultaneously produce increases in self-disclosure (which is good if the group is going to communicate about sensitive topics) and decreases in emotional connection (which is bad, as the group will be perceived by the individual as more distant and less supportive). Since increases in emotional connection produce increases in self-disclosure, decreasing this variable by increasing anonymity ultimately decreases the amount of self-disclosure. But the effect of anonymity via emotional connection is smaller than the direct effect of anonymity on self-disclosure, as explained in the Notes About Statistics, so the overall effect of increasing anonymity is to increase the communication of intimate information. This should promote group cohesion and honest communication, even though anonymity also somewhat decreases the emotional connection with the other group members and thus might decrease the motivational effectiveness of social interaction in a game. This is a bit of a quandary, as we would like to have both emotional connection and self-disclosure.
But there may be a way to provide the benefits of anonymity without the negative emotional distancing side-effect by designing the avatar to facilitate communication. In an earlier experiment, we found that the use of graphical avatars, like those typically found in games and used in the experiment described above, does decrease the sense of emotional connection between people (Kang, Watt, & Ala, 2008). But we also found that some types of avatars, like a digitally processed video image that contains facial expressions and body dynamics but masks details necessary for recognition, can preserve this emotional connection while still providing anonymity.
Providing both essential nonverbal information and anonymity in game avatars is likely to become easier in the immediate future with incorporation of video as a standard feature in home computers and game consoles and especially in mobile communication devices that are increasingly being used for games. Nonverbally expressive avatars can be fairly easily programmed in this new environment. Adding a social dimension without fear of harassment or ridicule to health games will improve their effectiveness and their potential to reach wide audiences.
Notes about the Statistics: The figure above is a path diagram, where the numbers on the graph are the standardized path coefficients, which range from zero (no effect of one variable on the other) to 1.0 (perfect relationship between variables). Positive numbers indicate that increases in the causal variable (at the open end of the arrow) produce increases in the effect variable (at the arrowhead end). Negative numbers indicate a decrease in the effect variable occurs with an increase in the cause variable. (The diagram and numbers are for a simplified model. The full model is described in Kang, Watt,& Gratch (2009).
Since the direct and indirect effects of anonymity on self-disclosure tend to cancel, it is important to see which is dominant. The path coefficients let us calculate this. Increasing anonymity increases self-disclosure by .27 standard units while decreasing emotional connection by .26 standard units. Since decreasing emotional connection will also decrease self-disclosure by .16 standard unit, the indirect effect on self-disclosure of changing anonymity by one standard unit is -.26 times .16, or -.04, a relatively small decrease. At the same time, increasing anonymity one standard unit directly increases self-disclosure by .27 standard units, so the net effect of anonymity on self-disclosure is .27-.04=.23 standard units.
Fraser, S. N., & Spink, K. S. (2002). Examining the role of social support and group cohesion in exercise compliance. Journal of Behavioral Medicine, 25(3), 233-249.
Joinson, A. N. (2001). Self-disclosure in computer-mediated communication: The role of self-awareness and visual anonymity. European Journal of Social Psychology, 31, 177-191.
Christopherson, K. M. (2007). The positive and negative implications of anonymity in Internet social interactions: “On the Internet, Nobody Knows You’re a Dog”. Computers in Human Behavior, 23(6), 3038-3056.
Kang, S., Watt, J. H., & Ala, S. K. (2008). Social copresence in anonymous social interactions using a mobile video telephone. In Proceeding of the Twenty-Sixth Annual SIGCHI Conference on Human Factors in Computing Systems (Florence, Italy, April 05 - 10, 2008). CHI '08. New York, NY: ACM.
Kang, S., Watt, J. H., & Gratch, J. (2009). Mediation of self-disclosure in human-avatar communication by copresence. (Working Paper, Under Review).