Commentary: Social Network Analysis and the Social Work Profession
2015; University of Chicago Press; Volume: 6; Issue: 4 Linguagem: Inglês
10.1086/684138
ISSN2334-2315
AutoresElizabeth M. Tracy, James K. Whittaker,
Tópico(s)Urban, Neighborhood, and Segregation Studies
ResumoPrevious article FreeCommentary: Social Network Analysis and the Social Work ProfessionElizabeth M. Tracy and James K. WhittakerElizabeth M. TracyCase Western Reserve University Search for more articles by this author and James K. WhittakerUniversity of Washington Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreTwo recent issues of the Journal of the Society for Research and Social Work (i.e., Vol. 5, No. 4; Vol. 6, No. 3) included four research articles dealing with current applications of social network analysis. These articles reflected diverse target populations of interest to the social work profession, including homeless youth (Barman-Adhikari, Rice, Winetrobe, & Petering, 2015), dually diagnosed adults (Henwood et al., 2015), incarcerated women (Kriegel, Hsu, & Wenzel, 2015), and service delivery networks for children with behavioral health problems (Bunger, Doogan, & Cao, 2014).Taken together, these articles illustrate the diversity of applications of social network analysis in social work research at the present time. This collection of articles also draws upon a long-standing focus in the social work profession on the reciprocal role of social networks as a reflection of the social environment in the development, maintenance, and amelioration of social problems—that is, social work’s person-in-environment perspective.Social Networks and the Social Work ProfessionSocial work as a profession has been interested in social networks from its earliest days; an example is the work of Mary Richmond and her view of the friendly visitor as a mobilizer between the client and extended family and neighborhood ties that could offer help and support (Richmond, 1922). At the same time, social work has benefited from the variety of disciplines and traditions that have contributed to the development of social network research and analysis. Scott (2000) described the roots of today’s social network analysis as stemming from three main traditions: sociometric analyses depicting group dynamics, sociologists’ study of informal relations within large organizations, and anthropologists’ study of small communities. For an overview of the measurement, terminology, and analytic methods currently used in social network analysis, Rice and Yoshioko-Maxwell (2015) have provided an excellent introduction to both egocentric (personal) and whole-network approaches.The ecological approach in the social work profession offers a strong rationale for social network analysis and intervention, considering the social ecology as both a cause and a solution to many problems (Barth, 1986). Social work takes both a micro and macro approach to practice, helping clients make individual changes as well as helping communities (and in general, the social environment of formal and informal services) to be more supportive of those changes (Whittaker, Schinke, & Gilchrist, 1986). Furthermore, similar to the visual methods used in social network analysis, the social work profession has a tradition of visual displays of information (e.g., Antonnucci, 1986; Hartmann, 1994), and therefore, the use of social network assessment as a form of intervention in and of itself is recognized in social work (e.g., Tracy & Whittaker, 1990, 1991). Pioneering texts in the 1970s and 1980s introduced social network analysis to the profession largely on the basis of the benefits derived by integrating informal sources of help with formal helping systems, with the aim of using professional services more efficiently and to provide a means of empowering clients and communities (e.g., Biegel & Naparstek, 1982; Biegel, Shore & Gordon, 1984; Maguire, 1983; Whittaker & Garbarino, 1983; see Tracy & Brown, 2011, for more on social networks and the social work profession). For example, practitioners were engaged in knowledge generation and the dissemination of knowledge through seminal monographs by Collins and Pancoast (1976) on the function of natural helping networks in day care provision, and through the work of Silverman (1980) on the Widow-to-Widow helping programs generated in agency practice contexts. These seminal works also yielded practical information for social work and identified key questions for future empirical research.Commentary on Current and Future ResearchThe four recently published articles addressing social network analysis draw on these traditions as well as foreshadow the next wave of research on social networks. This commentary discusses these articles in terms of their contributions to (a) a diverse theory base for framing social network analysis in the social work profession, (b) more detailed and nuanced approaches to the measurement of networks, both egocentric (personal) and whole networks, and (c) the variety of research designs that can be used in social network research studies.Theory BaseA variety of theoretical and conceptual frameworks have informed research on social networks. Stress and coping theory is relevant to social network research because of its focus on the availability of personal and social resources to cope with stress (Lazarus & Folkman, 1984). In this tradition, Thoits (1995, p. 64) considered the social network as a social “fund” that people could draw upon when needed. In stress and coping theory, social networks are thought to have both direct and indirect (buffering) effects (Berkman & Syme, 1979; Cassel, 1974; Cobb, 1976). For example, in an examination of network effects on abstinence, Longabaugh, Wirtz, Zweben, and Stout (1998) found both direct positive effects and buffering effects of having abstinent supporters in networks. Alternative models to stress and coping theory include the stress deterioration model, in which stressful life events are thought to disrupt or reduce social support resources (Dean & Ensel, 1982), and the stress prevention model (Gore, 1981; Gottlieb, 1983), in which social networks prevent the occurrence or the perception of stressful events.More recently, exchange theory (Wellman, 1981), rational choice theory (Lin, 1982), social capital theory (Bottrell, 2009) as well as dynamic network theory have been used in social network research (Westaby, Pfaff, & Redding, 2014). Barman-Adhikari et al. (this issue) drew from structural network theory that links network structural properties, such as density, and the positions held by people within the network who exhibit certain behaviors. Bunger et al. (2014) stated that network effectiveness theory was used in their whole-network analysis of the service delivery partnerships among behavioral health agencies. Kriegel et al. (this issue) drew from social exchange theory to examine whether personal networks mediated the association between incarceration and HIV risk; although limited by a cross-sectional design, Kriegel and colleagues found personal networks with alters who engaged in risky sex partially mediated the relationship between incarceration and multiple sex partners and the sex trade.The Kriegel et al. (2015) study is interesting in that it also pointed out the way in which incarceration removes individuals from their typical networks, resulting in overall smaller networks. The same could be said for many similar social work interventions (e.g., residential treatment, foster care) that separate the client from their usual set of interactions. Although this separation can often be a positive influence—such as when separation reduces contact with a negative, abusive, or antisocial network—separation often has unintended consequences and implications for reintroducing the client to their former network posttreatment. Frequently in such instances, reunification with and restoring network relationships is problematic, particularly when placement has occurred.The Henwood et al. (this issue) study touched directly on this theme; these authors examined whether dually diagnosed homeless adults in a Housing First treatment would be more isolated and disconnected as compared with those in traditional treatment, which placed adults in a shared living arrangement before independent housing. Henwood and colleagues found the Housing First participants were not more socially isolated, although they did have a higher proportion of staff members in their networks compared with the traditional treatment group. However, the Housing First participants were also found to be guarded about having close relationships for fear of being exploited. Qualitative interviews with the study participants revealed that those in the traditional housing program found the time-limited nature of the intervention discouraged the formation of new relationships. In addition, those in shared housing arrangements (i.e., traditional treatment) might know a greater number of people but these participants were not integrated into their community. The authors noted that for this group of dually diagnosed clients. larger social factors such as poverty and unemployment might be ongoing contributors to social isolation.This study underscores the way in which an intervention can interact with a client’s current network and either create limitations or facilitate the development of the future network. For example, Min et al. (2013) compared the makeup of the personal networks of women in treatment for substance use disorders. These researchers found that as compared with women who entered intensive outpatient treatment, women who entered residential substance abuse treatment had personal networks that included more people who used substances, more people with whom they had used alcohol and/or drugs, and fewer people from treatment programs or self- help groups. In Min et al.’s sample, network cohesiveness for both groups increased over a 12-month follow-up posttreatment intake; however, although the network composition of women in intensive outpatient treatment remained largely the same 12 months posttreatment, women in residential treatment continued to experience less support from their networks. Moreover, over the follow-up period, women in residential treatment maintained the number of substance-using alters in their network, leading these women to be more vulnerable to relapse in ways not experienced by the women in outpatient treatment.Attachment theory has many implications for network research, such as how support appraisal is influenced by internal working models of attachment (Collins & Feeney, 2004), and how insecure attachments might contribute to stress and the perception of stress (Koopman et al., 2000). Childhood maltreatment disrupts emotional bonds and produces an insecure attachment style. Compromised attachment has a profound impact on developing capacities for regulating negative affect and mobilizing others as support in times of need by limiting an individual’s ability to relate to others and to participate in satisfying social interactions, which in turn, contributes to the development of a compromised personal network later in life (Charuvastra & Cloitre, 2008). Green, Furrer, and McAllister (2007) reported that, as compared with mothers with secure attachment styles, those with insecure attachment styles perceived less social support from their networks. Similarly, Suchman, McMahon, Slade, and Luthar (2005) found the early attachment experiences of drug-dependent mothers mediated their ability to perceive network support, even when support was available.Trauma theory and the role of previous trauma and childhood maltreatment in shaping future family and network relationships are related to attachment theory and are relevant to many of the populations served by social workers. This trauma theme runs through several of the articles under consideration in this commentary. In the literature review of their article, Kriegel et al. (2015) cited Green et al.’s (2012) finding that women with histories of childhood physical abuse were more likely to have networks characterized by high risk and lesser-quality connections; consequently, a history of child abuse was included as one of the covariates in Kriegel and colleagues’ path analysis (although this covariate was not statistically significant in the final model). Barman-Adhikari et al. (2015) noted that youth often engage in substance use as a way to cope with the effects of exposure to family violence; however, the study analysis did not include trauma-related variables or account for the effect of trauma on network relationships. Henwood et al. (2015) did not discuss trauma in their article, although their dually diagnosed client sample would likely have experienced traumatic events. However, as previously noted, Henwood and colleagues pointed out larger structural factors present among their dually diagnosed client sample that could serve as barriers to network involvement. Min, Tracy, and Park’s (2014) study of personal networks in a sample of women with substance-use disorders who were receiving treatment demonstrated the consistent effects of trauma symptoms on the quality of network relationships, such as closeness. Overall, the role that previous attachments and trauma play in shaping the amount and quality of relationships within a network needs to be the focus of continued research.Measurement of NetworksUnderstanding the manner in which social network data are collected is of central importance to the interpretation and application of knowledge gained from the data. The four articles under consideration illustrated a range of approaches to data collection. Kriegel et al. (2015) used a personal network interview approach, which is a commonly used method. This approach starts with a name generator that asks respondents for the names of 20 people with whom they have had contact in the past 12 months (i.e., face-to-face, phone, mail, or e-mail); the name generator is followed with alter-specific questions, including alter behavior, type of relationship, and frequency of contact. Kriegel and colleagues stated that they chose to reduce the research burden in their study by asking two alter-specific questions for only 12 contacts named in the first step; the 12 names were selected via a stratified probability sample from the 20 names generated, and given the nature of their research questions, Kriegel et al.’s study design over sampled respondents’ sex partners. Henwood et al. (2015) used a mixed-methods approach, relying on extracting information on network size, composition, and quality from transcripts of semistructured interviews. In addition, at three time points (i.e., baseline, 6 months, and 12 months), two research team members worked independently to develop network case summaries that included overall quality as well as changes in relationships; these summaries were then coded for analysis, and Henwood et al. pointed out a high level of agreement between coders. The quality of relationship changes over time was coded as either growing distant, growing closer, or remaining the same. This method of quantifying qualitative data is innovative in that network data were extracted from routinely collected information, and this approach differs substantially from the name generator approach. Henwood and his colleagues noted that the initial (baseline) interview might have overlooked some network connections, and more importantly, that their data collection method could not provide a count of network members at each time point.In the third article, Barman-Adhikari et al. (2015) used an event-based approach to data collection that captured a person’s position in a network to examine how network structure and position interact with peer influence among homeless youth. Their findings showed an individual’s connections to substance-using peers and that individual’s position in the network influenced rates of substance use among homeless youth. Notably, Barman-Adhikari and colleagues primarily focused on sociometric data to show how behavior can be influenced by location within the network, not just network composition or quality of re lationships. The sociometric map was drawn through an interview process with the youth participants, in which each youth nominated people with whom they had interacted. In a process similar to that used in egocentric (personal) network data collection, the first step included nominating or generating a list of contact names, and then the interviewers asked questions about each alter listed. In the next step, study participants were linked using a sociomatrix based on matches of nominated names; this adjacency matrix enabled the researchers to compute structural network variables (e.g.. centrality, components, isolates) and to use specialized computer software to generate visual representations of social networks.Barman-Adhikari et al.’s (2015) findings reinforced those from prior research on composition of personal networks in that having peers who used substances was associated with a greater likelihood of the study youth also using substances. However, the results also showed the position a youth held in the network was associated with substance use. Specifically, these authors found that popular, well-connected youth were more likely to use methamphetamine, whereas youth in peripheral network positions were more likely to use heroin. Interestingly, cocaine use did not follow a pattern, and the authors suggested that this finding might have been impacted by questions that asked about lifetime cocaine use rather than current use; thus, the data on cocaine use might reflect past, not present, networks. Future studies should continue this focus on network structure because it opens new areas for possible intervention. For example, in a longitudinal study of low-income women in substance abuse treatment, Park, Tracy, and Min (2015) found a greater number of nonsubstance using isolates in the network at 6 months posttreatment intake was associated with less likelihood of substance use by 12 months posttreatment intake.A similar whole-network approach was used in the Bunger et al. (2014) study of service delivery networks based on agencies reporting sharing of referrals and staff expertise with other agencies. Follow-up interviews over a 2-year period allowed the researchers to illustrate the evolution and growth of service delivery systems, and enabled the authors to move beyond description to examine mechanisms that help explain the ways in which networks evolve. Moreover, Bunger and colleagues’ study demonstrated how a network analysis approach can be used to study macro-level factors of interest to social work practitioners and administrators.Research Designs and Implications for Future ResearchTaken together, these four articles are fine examples of the range of applications possible for social network analysis in terms of the client populations included and the data collection methods used. In addition, the four studies used a variety of research designs: mixed methods, egocentric-network analysis, whole-network analysis, path analysis, as well as longitudinal and cross-sectional designs. Notably. each of these studies goes beyond merely describing the properties of social networks to examine the role of social networks in relation to specific outcomes relevant to the sample included in the study.As with any body of research, these studies contain some design limitations that future research will need to address. With respect to social network interventions, Ertel, Glymour, and Berkman (2009) identified the following limitations of social network research: (a) the focus on method of delivery without considering the timing of delivery, (b) the use of patient-only samples rather than community-based samples, and (c) extensive reliance on cross-sectional designs. At this time, longitudinal analysis of network data appears to be the most promising method to better examine causal relationships and to identify network stability and change over time. Examples of data-rich longitudinal studies include studies that have examined network changes over the life span (Wrzus, Hanel, Wagner, & Neyer, 2013) and network changes after significant life events (Stone, Jason, Stevens, & Light, 2014). As compared with cross-sectional studies, longitudinal studies using repeated observations over time, can distinguish within-individual changes from inter-individual differences. Longitudinal analyses examining changes (or lack of changes) in the composition and structure of personal networks and correlates related to those changes (or lack of changes) will provide better understanding of personal networks, which can improve treatment efforts by specifying modifiable targets.Future research should also focus on the use of latent class analysis to identify underlying patterns in network relationships (Buckman, Bates, & Morgenstern, 2008). Qualitative studies, as exemplified by Henwood et al. (2015), that explore the ways in which people experience and manage their network relationships would be helpful in creating a better understanding of the mechanism for change in networks. A qualitative study conducted by Brown, Tracy, Jun, Park, and Min (2015) described how women in recovery added individuals to their networks, and how the women managed existing relationships (i.e., distancing. isolating some members) to diminish the negative effects on their recovery. As more research is conducted using network structural variables, we might, as Barman-Adhikari et al. suggested, be better able to target interventions to subgroups within the network. As indicated earlier, more research is needed to explicate both the role of trauma in shaping network relationships and the role of other societal barriers to network involvement.Finally, more research using true experimental designs with comparison or control groups is needed to explore network interventions. Much remains unknown regarding the timing and sequencing of social network interventions and how to target interventions more precisely and individually. For example, the type of network that is helpful might differ depending on the stage of treatment of the client (Tracy & Johnson, 2007). As Tracy, Munson, Peterson, and Floersch (2010) pointed out, social workers must remain mindful of the complex positive and negative effects of social networks—or what Freisthler, Holmes, and Wolf (2014) referred to as a “dark” side of social support.In sum, the four studies discussed here illuminate both substantive findings and key implications in domains of high interest for social work practice, as well as shed light on potentially fruitful methodological pathways for future investigations in social network analysis. Although all of this work is surely to the good, it somehow seems inadequate to answer Brekke’s call for a science of social work (2012), which several of the investigators of the studies discussed here invoked as at least a partial stimulus to their present studies. Even when envisioning a greatly expanded corpus of empirical network analysis as exemplified in these excellent efforts, the direct effect on social work practice must of necessity await the kind of careful and systematic process of translational research that will render specific findings into useful guidelines and stratagems for practice. Given the sheer size and complexity of social work’s practice sphere and the profession’s still rudimentary scientific enterprise that Brekke (2012) succinctly identified, perhaps, in addition to more thoughtfully crafted individual studies such as those reviewed here, something larger and more systemic is needed to truly expand social work science and its impact on social work practice.One clue to such undertakings may be found in some of the innovative projects, studies, and organizational structures that served to inspire an earlier generation of social network investigators. For Collins and Pancoast, these thought-provoking efforts would surely include the ambitious community intervention in England known as the Dinnington Neighborhood Services Project (Bayley, Seyd, & Tennant, 1989), which sought to bridge the gap between formal and informal helping sources in a local authority, as well as the elegant and thick description by Young and Willmott (1957) of “family and kinship” in East London. For Phyllis Silverman, the list of inspiring works must include Gerald Caplan’s (1964) innovative Laboratory for Community Psychiatry at Harvard, which created a fluid space where professional mental health service providers and lay helpers were allowed to not only exist simultaneously but also to learn from one another. Finally, scholars will always be drawn to the many contributions of Bill Reid for creatively using the agency practice context as a crucible for the generation of new knowledge and for whom the social agency was indeed a “research machine” (1978). Social network analysis offers both some fruitful pathways for future empirical study and some tantalizing schema for what Bronfenbrenner (1979) long ago referred to as the “transforming experiment.” The timing is right for such enterprises—reducing social isolation and strengthening ties among people has been proposed as one of the grand challenges for social work (Lubben, Gironda, Sabbath, Kong, & Johnson, 2015). Let the work begin on both fronts.NotesElizabeth M. Tracy is the Grace Longwell Coyle Professor in Social Work and Associate Dean for Research and Training at the Jack, Joseph, and Morton Mandel School of Applied Social Sciences at Case Western Reserve University.James K. Whittaker is the Charles O. Cressey Endowed Professor Emeritus in the School of Social Work at the University of Washington.Correspondence regarding this article should be sent to Dr. Elizabeth Tracy, 11235 Bellflower Rd., Cleveland, OH 44107-7164 or via e-mail to [email protected]ReferencesAntonucci, T. C. (1986). Hierarchical mapping technique. Generations: Journal of the American Society on Aging, 10(40), 10–12.First citation in articleGoogle ScholarBarth, R. P. (1986). Social and cognitive treatment of children and adolescents. San Francisco, CA: Jossey-Bass.First citation in articleGoogle ScholarBarman-Adhikari, A., Rice, E., Winetrobe, H., & Petering, R. (2015). Social network correlates of methamphetamine, heroin, and cocaine use in a sociometric network of homeless youth. Journal of the Society for Social Work and Research, 6, 433–457. http://dx.doi.org/10.1086/682709First citation in articleLinkGoogle ScholarBayley, M., Seyd, R., & Tennant, A. (1989). Local health and welfare: Is partnership possible? A study of the Dinnington Project. Aldershot, United Kingdom: Gower.First citation in articleGoogle ScholarBerkman, L. F., & Syme, S. (1979). Social networks, host resistance, and mortality: A nine-year follow-up study of Alameda County residents. American Journal of Epidemiology, 109, 186–204.First citation in articleCrossrefGoogle ScholarBiegel, D. E., & Naparstek, A. J. (Eds.). (1982). Community support systems and mental health: Practice, policy, and research. New York, NY: Springer.First citation in articleGoogle ScholarBiegel, D. E., Shore, B. K., & Gordon, E. (1984). Building support networks of the elderly: Theory and applications. Newbury Park, CA: Sage.First citation in articleGoogle ScholarBottrell, D. (2009). Dealing with disadvantage: Resilience and the social capital of young people's networks. Youth & Society, 40, 476–501. http://dx.doi.org/10.1177/0044118X08327518First citation in articleCrossrefGoogle ScholarBrekke, J. S. (2012). Shaping a science of social work. Research on Social Work Practice, 22, 455–464. http://dx.doi.org/10.1177/1049731512441263First citation in articleCrossrefGoogle ScholarBronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press.First citation in articleGoogle ScholarBrown, S., Tracy, E. M., Jun, M., Park, H., & Min, M. O. (2015). Personal network recovery enablers and relapse risks for women with substance dependence. Qualitative Health Research, 25, 371–385. http://dx.doi.org/10.1177/1049732314551055First citation in articleCrossrefGoogle ScholarBuckman, J. F., Bates, M. E., & Morgenstern, J. (2008). Social support and cognitive impairment in clients receiving treatment for alcohol-and drug-use disorders: A replication study. Journal of Studies of Alcohol and Drugs, 69, 738–746. http://dx.doi.org/10.15288/jsad.2008.69.738First citation in articleCrossrefGoogle ScholarBunger, A. C., Doogan, N. J., & Cao, Y. (2014). Building service delivery networks: Partnership evolution among children’s behavioral health agencies in response to new funding. Journal of the Society for Social Work and Research, 5, 513–538. http://dx.doi.org/10.1086/679224First citation in articleLinkGoogle ScholarCaplan, G. (1964). Principles of preventive psychiatry. New York, NY: Basic Books.First citation in articleGoogle ScholarCassel, J. (1974). Psychosocial processes and “stress”: Theoretical formulation. International Journal of Health Services, 4, 471–482. http://dx.doi.org/10.2190/WF7X-Y1L0-BFKH-9QU2First citation in articleCrossrefGoogle ScholarCharuvastra, A., & Cloitre, M. (2008). Social bonds and posttraumatic stress disorder. Annual Review of Psychology, 59, 301–328. http://dx.doi/org/10.1146/annurev.psych.58.110405.085650First citation in articleCrossrefGoogle ScholarCobb, S. (1976). Social support as a moderator of life stress. Psychosomatic Medicine, 38, 300–314. http://dx.doi.org/10.1097/00006842-197609000-00003First citation in articleCrossrefGoogle ScholarCollins, A. H., & Pancoast, D. L. (1976). Natural helping networks: A strategy for prevention. Washington, DC: National Association of Social Workers.First citation in articleGoogle ScholarCollins,
Referência(s)