A Dynamic Social Network Approach to Understanding Mutual Aid and Self Organization Among Therapeutic Community Residents

Background: Therapeutic communities (TCs) are based on a system of mutual aid in which residents monitor each other, affirming prosocial behavior and correcting behavior that contravenes TC norms.  TC clinical theory emphasizes the importance of these peer interactions in bringing about change (De Leon, 2000; Perfas, 2012).  The community of residents self-organizes through peer interactions, and this community is itself the method of treatment (De Leon, 2000).  This implies two testable hypotheses.  First, residents should increase prosocial behavior more in response to peer versus staff intervention.  Second, residents should interact in ways that are known to maintain cooperation in groups.  There is extensive evidence that direct reciprocity, reputation effects and positive response to sanctions foster cooperation in groups (Rand & Nowak, 2013). In addition to the increase in prosocial behavior following peer affirmations we would therefore expect evidence that residents reciprocate prosocial behavior, reward peers who act in a prosocial manner and respond to sanctions with prosocial behavior. Finally, De Leon (2000) suggests that TC residents will send affirmations to peers to balance corrections.     

Method:  The data consist of electronic records of peer and staff affirmations and corrections at four corrections-based TC units for men.  We treat the data as a directed social network of affirmations and corrections.  Since TCs define peer affirmations as prosocial acts, our measure of prosocial behavior is therefore the likelihood that residents will affirm peers on any given day.  We sample 100 resident days from each unit (n = 400) and use a generalized linear mixed effects network time series model to analyze the predictors of sending and receiving affirmations and corrections.  The model allows us to control for race, age, time in program and criminogenic factors as measured by the Level of Service Inventory-Revised (Andrews & Bonta, 1992) as individual characteristics and sources of homophily.

Results:Residents increase the number of affirmations they send after receiving a peer affirmation, but not after receiving a staff affirmation. Residents do not respond to peer corrections by increasing affirmations, but respond to staff corrections by decreasing affirmations. Residents directly reciprocate peer affirmations. Residents are more likely to affirm a peer whom they have recently corrected, as predicted in De Leon (2000).  Residents showed homophily with respect to race, age and program entry time.

Conclusions and Implications: This analysis demonstrates that TC residents react more prosocially to behavioral intervention when it comes from peers than when it comes from staff.  This finding validates TC clinical theory.  The finding that TC residents interact in ways known to foster cooperation in groups suggests that the community that is at the core of TC clinical treatment largely self-organizes through peer interactions.  The analysis suggests ways of improving TC treatment through encouraging prosocial peer interactions and potentially using homophily by program entry time to counteract other forms of homophily.