Non-Suicidal Self-Injury: a study about the evolution of stable maladaptive strategies

Non-Suicidal Self-Injury (NSSI) is a maladaptive behavior and it is defined as "the deliberate, self-inflicted destruction of body tissue without suicidal intent and for purposes not socially sanctioned". Although under-reported, NSSI is often revealed in community samples, and a recurring comorbidity was found with suicide, which highlights the gravity of NSSI itself, as it may lead to serious injury or even mortality. It may occur in all phases of life, but individuals appear to start such behavior very early in their life, especially in adolescence and young adulthood, with possible incidence rates that considerably vary among studies (i.e., approximately from 5% to 38%). This dissertation aims to thoroughly explore the NSSI topic. Firstly, the behavior was investigated from a psychological and experimental perspective, through observational data. After, two different approaches applying Agent-Based Modeling (ABM) were introduced to study the role of diverse factors affecting NSSI. Two main theoretical models were considered to better understand NSSI behavior, and as a background for the implementation of the numerical model. Particularly, the Experiential Avoidance Model claims NSSI is a negatively reinforced strategy for terminating unwanted emotional arousal. On the other hand, the Integrated Theoretical Model of the Development and Maintenance of Self-Injury focused on the description of the functions of NSSI (i.e., interpersonal and intrapersonal functions), and it introduces how distal factors may increase an intrapersonal and interpersonal vulnerability, and how specific factors may lead to engage in NSSI to cope with stress. Hence, the latter was suitable to thoroughly identify the complexity of factors involved in NSSI. Besides literature has showed the presence of several factors that affects NSSI, previous studies have not investigated the specific role of diverse factors in affecting NSSI. Particularly, several studies were more focused on the individual-related risk factors, omitting to adequately analyze interpersonal factors or other similar psychosocial dimensions, such as adolescent dynamics and risk behaviors in virtual social networks. To fill this gap, a few preliminary studies were conducted: we have analyzed the online social dynamics of adolescents, and the role of interpersonal and family relationships in their life, to detect possible risk factors. For what concerns the numerical modeling, we have focused on the NSSI dynamics in adolescence, through the Agent-Based Modeling (ABM). As a first approach, main NSSI risk factors (i.e., Inner Factor, Outer Factor, Media Factor) were selected from literature, assuming them as increasing the probability of self-injury. Moreover, three network topologies (i.e., Uniform, Gaussian, Exponential) and a probability to experience stressful events were settled as fixed parameters. As a dynamical parameter, the Peer Factor was introduced as the density of self-injurers in the network. The mathematical model was described, and the numerical simulations conducted were introduced, first considering the effect of each risk factor singularly, then contemplating the combined effect of all risk factors. As a second approach, a preliminary validation of the mathematical model was conducted. Risk factors were parametrized through real data collected in three secondary schools, and numerical simulations were started comparing results with those obtained through the survey. Moreover, the network was finally divided in two subclusters presenting a different density of self-injurers. Hence, the propagation of self-injurers from a subcluster to another was analyzed, as well as the number of connections between subclusters increased along numerical simulations. Results revealed a relevant effect of both risk factors and the dynamical parameter Peer Factor on NSSI dynamics, considering both results from the non-parametrized model and results from the parametrized model. On the contrary, topology displayed little or none effect on the probability of self-injury. Moreover, the model appears to adequately reproduce NSSI real trends, obtaining similar results than real data, confirming its good effectiveness in describing NSSI dynamics. Such findings highlighted interesting implications about the complex dynamics of the phenomenon, and might represent a starting point to implement an integrated model reproducing similar maladaptive behaviors (e.g., gamble).