Risk-Taking Among Emerging Adults: A Tree-Based Clustering Approach

• In Event: Poster Session 06
  In Poster Session: PS 06 Topic - Health, Growth, Injury

Abstract

Risk-taking is defined as behaviors associated with probability of undesirable results. Numerous studies have provided evidence that parenting plays an important role in one’s risk-taking; however, there are voids in the literature. First, there are mixed results regarding the relationship between parental behavioral/psychological control and risk-taking. Second, most of the literature on parental control has been conducted with children and adolescents, but as parents are found to exert control over their children beyond adolescence, the study among emerging adults is necessary. The purpose of the current study was to identify the high risk-taking emerging adults and compare them with the normal group regarding their perceived parental control, decision-making, and risk-taking. To identify these emerging adults from the sample, the machine learning approach was applied. The study used tree-based clustering as a technique in machine learning that groups observations by their association so that observations in the same cluster are more similar than those in other clusters.
Data were drawn from 538 college students by using an online survey. The measurements included demographic questions, parental behavioral and psychological control, risk tolerance, risk self-schema, and self-reported risk-taking behaviors. To identify the potential high risk-takers, two unsupervised learning methods, including data cloud geometry tree (DCG-tree; Fushing et al., 2013) and agglomerative hierarchical clustering tree (HC-tree), were used to get clusters of participants based on the pattern of their responses. Next, post hoc tests were conducted to examine the differences between the potential high risk-taking group and normal group.
Among the participants, 22 students showed a special pattern in both DCG-tree and HC-tree and clustered into a group as potential high risk-takers. Compared to the normal group, the potential high risk-taking group was more likely to engage in risk-taking behaviors (e.g., risky driving, substance use), reported lower parental behavioral control (see Figure 1), and perceived higher parental psychological control across the 7 components of psychological control (see Figure 2). In addition, the Wilcoxon rank-sum tests indicated that the high risk-takers could tolerate more risks and were more likely to have a self-schema of being a risk-taker.
This exploratory study suggests that using the tree-based clustering approach can identify the high risk-takers, who show distinctive characteristics that are different from the normal emerging adults. Consistent with the literature, parental control continues to play a role in risk-taking among emerging adults. Despite the fact that children have reached legal adult age, parents often do not stop attempting to control their college-aged children behaviorally and psychologically, and further influence their children’s behavioral outcomes throughout emerging adulthood. Future study may use the machine learning approach and examine more variables as well as demographics to better understand the characteristics of risk-takers in emerging adulthood. Although there is no standard way to quantify the performance of the clustering results, the present study indicates that using the tree-based clustering methods amy help identify the potential high risk-takers among emerging adults and may be helpful in future intervention.

Author

• Catherine P Chou, National Taiwan University
  Presenting Author