Network Diversity and Affect Dynamics: The Role of Personality Traits

Abstract

People divide their time unequally among their social contacts due to time constraints and varying strength of relationships. It was found that high diversity of social communication, dividing time more evenly among social contacts, is correlated with economic well-being both at macro and micro levels. Besides economic well-being, it is not clear how the diversity of social communication is also associated with the two components of individuals' subjective well-being, positive and negative affect. Specifically, positive affect and negative affect are two independent dimensions representing the experience (feeling) of emotions. In this paper, we investigate the relationship between the daily diversity of social communication and dynamic affect states that people experience in their daily lives. We collected two high-resolution datasets that capture affect scores via daily experience sampling surveys and social interaction through wearable sensing technologies: sociometric badges for face-to-face interaction and smart phones for mobile phone calls. We found that communication diversity correlates with desirable affect states--e.g. an increase in the positive affect state or a decrease in the negative affect state--for some personality types, but correlates with undesirable affect states for others. For example, diversity in phone calls is experienced as good by introverts, but bad by extroverts; diversity in face-to-face interaction is experienced as good by people who tend to be positive by nature (trait) but bad for people who tend to be not positive by nature. More broadly, the moderating effect of personality type on the relationship between diversity and affect was detected without any knowledge of the type of social tie or the content of communication. This provides further support for the power of unobtrusive sensing in understanding social dynamics, and in measuring the effect of potential interventions designed to improve well-being.

Fig 1. At the beginning and the end of the experiments, participants filled a survey capturing their (stable) traits.

(i) Framework of the Sociometric Badges Dataset: Participants filled 3 daily surveys to measure (dynamic) states. We take the daily average of states and calculate the daily diversity in communication that took place before the last filled survey. (ii) Sociometric Badges are used to track face-to-face interactions by means of infrared (IR) sensors. (iii) Framework of the Mobile Territorial Lab project: Participants filled one daily survey to measure daily (dynamic) states. We calculate the diversity in communication that took place before the daily filled survey. (iv) Smart phones tracks the daily call social networks of participants.

Fig 2. Sample daily networks of the two datasets.

(i) The network of the community of participants is constructed based on the sociometric badges dataset in one typical day. The infrared sensors can only detect the infrared sensors of other participants, so the network includes face-to-face interactions within the community of participants. (ii) The call network is constructed including people within or outside the participant community in one typical day. MTL participants are coloured with dark blue while non-participants are coloured with green. In both networks, the thickness of edges represents the intensities of communication between two nodes. We can see that individuals are inclined to divide their time unequally among their social contacts.

Fig 3. (a) Distribution of total time spent on phone calls by participants (b) Distribution of total time spent communicating through face-to-face interaction (c) Distribution of personal network sizes via mobile phones calls (d) Distribution of personal network sizes via face-to-face interaction.

Generally, few individuals have very long phone calls or face-to-face interactions. Also, few individuals have a high number of social contacts.

Fig 4. Distribution of dynamic HPA state by participant and trait in the sociometric badges dataset.

Each boxplot describes the distribution of the high positive affect (HPA) state score for each participant. The colour intensity indicates the individual’s score of the corresponding trait (HPA trait). The dark blue colour means that a participant has a low score in the HPA trait, while the light blue colour means that a participant has a high score in the HPA trait. The boxplots are sorted in an increasing order according to the score of the HPA trait. Generally, people who have high scores in the trait (darker colours) tend to have high scores in the scores of the corresponding state as well. The same applies to people who have low scores in the trait (lighter colours) who tend to have low scores in the corresponding state as well. However, there are many exceptions whereby the dispositional trait of HPA do not explain the daily scores of the state.

Fig 5. The bar plot demonstrates the mediating effect of traits in the relationship between the diversity measure and the score of the dynamic affect state.

Generally, high scores of the trait correspond to high scores of the state. However, if we look for within-trait variation of dynamic state, we can see that diversity plays different roles in different levels of the trait. For example, when the trait level is low, the score of the state is relatively high for high scores of Gini and low for low scores of Gini. In contrast, when the trait level is high, the score of the state is relatively high for low scores of Gini and low for high scores of Gini.