Social Convoys: Those who accompany us in our journeys through life
The “Social Convoy Model” of relationships was developed by [21]. This model is vital to the general understanding of changes in social engagement over the life course. It highlights that the life circumstances and goals pursued at different life stages affect what people want from their social network and contacts. In other words, earlier life experience matters for one’s later social engagement.

Figure (6) illustrates an example of the Convoy Model. The center represents the individual, “Me.” The three larger circles represent that person’s convoy:

1. The Inner Circle (1st tier): People who are very close to the focal person. They are perceived as critical support givers. Membership in this inner circle is relatively stable throughout life. For example, a close friend who now lives far away may remain in the inner circle.
2. The Middle Circle (2nd tier): Often family, friends, or co-workers. Compared to those in the 1st tier, the focal person perceives support from 2nd tier members as more relevant on the roles that they fill in his or her life. A person’s relationship with his or her 2nd tier contacts may be less stable over the life course than those within the 1st tier, as many environmental changes may affect the nature of their relationships. An example could be mothers whose children went to the same school for 6 years who may become close, but the relationship changes after their children go to different schools later.
3. The Outer Circle (3rd tier): People who are least close to the focal person, but have nevertheless been identified as sources of support. Compared to the 1st and 2nd tiers, the interactions between the focal person and the 3rd tier members are even more dependent on each member’s role. Examples include co-workers, supervisors, and neighbors.

Experience sampling: A barometer of our daily feelings
Observing and measuring individuals’ emotional responses to daily experiences in a rigorous way has long been a challenge for social scientists. Bringing people into a laboratory setting, while useful in many contexts, disrupts daily life and places people in an unfamiliar setting. This can reduce the ecological validity of a study, which refers to how similar the study setting is to that of the real-world setting that the researchers are interested in. Study sessions conducted in laboratory settings also may not capture a complete picture of individuals’ emotional experiences, which often fluctuate throughout a given day or week. To overcome these obstacles, researchers are taking advantage of new technologies to more reliably and completely capture people’s emotions and experiences in their daily settings.

One of the first such efforts is the Experience Sampling Method (ESM), which was initially created in the late 1970s [3] and continues to be used today. ESM involves a researcher alerting participants at random times throughout the day (e.g., with a beeper or other paging device). At each of these time points, the participants provide immediate responses to questions about their current context and how they are feeling in that moment (see Figure 1 for an example of an experience sampling questionnaire). In order to explore whether emotional experience changes with age, Carstensen and colleagues [4] combined ESM with a longitudinal design. They alerted participants five times per day over one week to gather rich emotional experience data, then repeated this experience sampling five and ten years later, also for one week each. The researchers found that older age is associated with greater emotional well-being and emotional stability.

Figure 1: Emotional experience questionnaire5. Participants rate their feeling of 19 emotions on a 7-point scale, partly seen here. The list of emotions included 8 positive emotions (e.g., pride, happiness) and 11 negative emotions (e.g., anger, guilt).

English and Carstensen [6] later used the data captured in the Carstensen et al. [4] study (described previously) to examine how people’s emotional experiences might be affected by their social connections. The researchers used the daily experience sampling data to calculate, on average, how positive and negative a person felt in a week. They then analyzed this emotion data with data from a social network assessment that had been collected once during each wave for information about the composition of each person’s social network and whether the participant associated positive or negative emotions with each social connection. The researchers found that the emotions associated with people’s social connections influenced their daily emotional well-being, particularly if the connections were associated with negative emotions.

A modified version of ESM – the Day Reconstruction Method (DRM) – was later developed by Kahneman and colleagues [7] in attempt to minimize the potentially disruptive effect of frequent alerts. The DRM approach asks participants are to reconstruct the previous day by breaking it into structured “episodes” (e.g., “commuting to work,” and “at lunch”). Similar to experience sampling, participants then answer questions about the context of each episode and how they felt during it. Building the context around each episode helps increase the accuracy of participants’ recall of recent memories. This method is less expensive than ESM and does not disturb daily activities. One study used DRM to explore why extraverts are more likely to feel positive affect than introverts [8]. While other methods tend to conflate extraversion and global positive affect, the DRM approach allowed the researchers to separate these two constructs by using a global measure of extraversion and episodic measures of positive affect, thus avoiding overlap and providing richer data. They found that extraverts and introverts enjoy social interactions to a similar extent, but extraverts are more likely to interact with others more frequently, which helps explain why they experience greater positivity.

The administration of experience sampling and similar methods has become simpler and more practical thanks to the ubiquity of smartphones. While previous experience sampling researchers alerted participants using pagers and phone calls, these alerts and questionnaires can now be pushed directly to participants’ smartphones, which makes real-time data collection considerably easier and more efficient. Smartphones can also be used to collect and store the large volumes of data that are typically collected in ESM and DSM research. For example, in another study, researchers wanted to find out how often older adults reminisce in a conversation [9]. However, traditional self-report methods presented challenges, as many people were unaware of when they were reminiscing and others could not accurately recall when they reminisced. The methodological solution was to have participants carry a smartphone with recording software that would randomly record 30 seconds of their speech throughout the day. With this sporadic sampling, the researchers were able to obtain a much more accurate measurement of how often people were exhibiting this reminiscing behavior and with which social partners, and how reminiscing was related to well-being.

Screenomics: Capturing a digital footprint
Much of daily life now revolves around using social technology. It is so ubiquitous that instead of simply asking if people use technology, researchers now employ methods that allow them to measure how people are using it. Byron Reeves, Nilam Ram, and colleagues have developed a new method for examining the conscious and unconscious choices people make and the patterns they create as they switch between texting, email, social media, and other apps on their devices. Their approach involves building what they call a “Screenome” [10].

To build a Screenome, the researchers install an app on participants’ smartphones or computers that takes screenshots of their activities on the device every five seconds that the device is in use throughout the duration of the study. This is similar to the Experience Sampling Method in that it captures snapshots of an individual’s activity, but the advantage of this approach is that it does not rely on participants to provide responses. The Screenome is a powerful tool for investigating many aspects of the way people use technology across multiple platforms. This level of detail being measured in real time had, until now, been difficult to capture in social technology research.

In one example of the Screenome method, Reeves et al. examined the way a cohabiting couple interacted with each other and with their friends and families online [10]. Figure 2 shows each partner’s social media use on two different devices across 29 hours. Person A can be seen switching quickly between social media apps in the evening on their laptop, while Person B uses social media on their phone more sporadically throughout the night. This type of data can be broken down further to examine other patterns, such as which platforms they used with which social partners (e.g., one of the partners switched between several different platforms but only used one platform to communicate with family).

Figure 2: A cohabiting couple’s social media use across 29 hours10. The color bars represent use of a social media platform, the black bars represent time that the device was on but a social media platform was not being used, and the gray space indicates when the device was not in use. The upper panel magnifies a 3-hour time period for Person A’s social media use and shows whether communications or exchanges with another person were synchronous (i.e., communicating at the same time) or asynchronous (i.e., time elapsed between messages or they were switching between different tasks).

Digital lexical: Predicting individual outcomes from social media posts
The words people say and write can reveal rich information about their thoughts, feelings, beliefs, and relationships. Without asking direct questions, researchers can listen to someone’s conversation or read a sample of their writing to gain a better understanding of that person and how they interact with others. This “lexical” approach involves identifying key words and counting the frequencies of those words in a person’s text or speech, while considering the circumstantial information surrounding those words.

In a classic study, Walter Weintraub [11] noted that first-person pronouns such as I, me, and my were used much more frequently by people suffering from depression than by a control group. He also applied this method to news conferences with U.S. presidents to analyze their personalities and speaking styles [12]. These early studies were an accomplishment not only because of the ideas and findings they contributed to the field, but because they were done before technology was sufficiently powerful to automate the process (e.g., the researchers often counted words by hand).

The lexical approach has since become widely popular thanks to the increasingly fast speed of computer processors and improved storage capacity. During the time the technology was being developed, researchers also took many years to reach a consensus about which words and other aspects of language are most indicative of the psychological states they were trying to measure [13]. In the 1990s, all of these factors came together in a text analysis computer application called Linguistic Inquiry and Word Count (LIWC) [14], which was developed by researchers at the University of Texas at Austin.

This natural language processing software categorizes words and phrases according to an extensive dictionary that the researchers developed; for example, the words “buddy” and “pal” are categorized under “friend.” These predefined categories can be used to examine variables of interest. For example, researchers were interested in whether the language a couple used with each other was associated with the quality of their relationship [15]. They used LIWC to scan compilations of heterosexual couples’ instant messages and calculate the percentage of words that fit into categories such as “positive emotion” and “negative emotion.” They then analyzed this linguistic data alongside a separately collected survey about relationship quality. Their findings included: men’s use of “positive emotion” words (e.g., “nice,” “happy”) was positively associated with relationship satisfaction for both themselves and their partner, while women’s use of “positive negation” words (e.g., “not nice,” “not happy”) was negatively associated with relationship satisfaction for both themselves and their partner.

In this age of social technology, the amount of written content available for lexical analysis has skyrocketed as people are producing massive amounts of content online. In addition to the previous example involving instant messages, lexical software has been applied to social networking sites to make health predictions about individuals. Such prediction models incorporate the text of the posts, post length, posting frequency and timing, and demographic information. By applying these models on Facebook, Eichstaedt and colleagues were able to accurately predict depression three months prior to an official medical diagnosis [16]. This new application offers promising pathways to detect health-related conditions such as depression or loneliness, which are often underdiagnosed and undertreated.