Abstract
The daily social life of a person can be captured with different methodologies. Two methods that are especially promising are personal-social-network (PSN) data collection and experience-sampling methodology (ESM). Whereas PSN data collections ask participants to provide information on their social relationships and broader social environment, ESM studies collect intensive longitudinal data on social interactions in daily life using multiple short surveys per day. In combination, the two methods enable detailed insights into someone’s social life, including information on interactions with specific interaction partners from the personal network. Despite many potential uses of such data integration, there are few studies to date using the two methods in conjunction. This is likely due to their complexity and lack of software that allows capturing the full social life of someone while keeping the burden for participants and researchers sufficiently low. In this article, we report on the development of methodology and software for an ESM/PSN integration within the established ESM tool m-Path. We describe results of a first study using the developed tool that illustrate the feasibility of the proposed method combination and show that participants consider the assessments insightful. We further outline study-design choices and ethical considerations when combining the two methodologies. We hope to encourage applications of the presented methods in research and practice across different fields.
Keywords
Humans have been termed “social animals” (Tomasello, 2014). The social context, which includes all aspects of an individual’s interpersonal interactions and relationships, has an impact on most aspects of human life. Social context can, for example, influence mental health and well-being through social support (Hendryx et al., 2009), shape education and career opportunities through information exchange (e.g., Granovetter, 1973), or affect physical health through peer influence on health-related behaviors (Mercken et al., 2007). Thus, social context is an important factor to capture when examining the daily life of a person and when aiming to understand a range of psychological outcomes.
When zooming into everyday life, important dynamics unfold via social interactions such as encountering an acquaintance while grocery shopping, meeting for dinner with friends, or collaborating on a project with a colleague (Brown et al., 2011; Hall & Merolla, 2020; Lin & Lachman, 2021; Sun et al., 2020). Over time, interactions aggregate into social relationships (e.g., friendships) that are part of the ever-evolving social network of a person (Hall, 2019; Holme, 2003; Walker et al., 2018). The relationship a person has with someone, in turn, influences further social interactions (e.g., one discusses more personal issues with a closer friend; Sias & Cahill, 1998). The immediate social context 1 of a person thus involves two connected levels that change on different timescales: momentary social interactions (i.e., verbal exchanges between two or more people in daily life) and more stable but still changing social relationships (i.e., “connections that exist between people who have recurring interactions that are perceived by the participants to have personal meaning,” August & Rook, 2013, p. 1838). In this article, we describe a methodological approach that aims to assess both of these levels and allows researchers to gain a fine-grained view of diverse interpersonal processes.
To date, researchers have often examined the social interactions and social relationships of a participant in separate studies using different methodologies (Langener et al., 2022). Two such methods that have gained popularity are personal-social-network (PSN) data collection (e.g., Reins et al., 2021) and experience-sampling methodology (ESM 2 ; Myin-Germeys & Kuppens, 2021). PSNs provide data on a person’s social relationships (McCarty et al., 2019; Perry et al., 2018) by asking participants to map out the people in their social environment and answer questions about them (e.g., whether they are a friend, family member, or colleague). In addition, ties between the different network members can be assessed (e.g., who knows whom). Note that such a network can be small (i.e., including only family and close friends) or also broad (e.g., including anyone the participant is interacting with in daily life). To capture changes in a person’s social relationships, such networks can be assessed repeatedly over a span of several weeks, months, or even years (e.g., Huxhold et al., 2013). ESM is used on a much shorter timescale (i.e., days or weeks) to measure participants’ experience in daily life by asking participants to report on, for example, their social interactions and mood several times a day using a smartphone. Thereby, a variety of sampling schemes with different assessment frequencies and spacing are possible (Myin-Germeys & Kuppens, 2021).
Why Integrate ESM and PSN?
An integration of ESM and PSN offers various possibilities. First, one can investigate the within-persons interplay of psychological and social variables in daily life while considering specific interaction partners. Researchers can, for example, investigate associations between mood and quality of social interactions with a particular friend or colleague if they know about characteristics of the participant’s relationship with this person. Currently, most ESM studies do not distinguish between different interaction partners but assess only the social role of an interaction partner (e.g., friend, family member, colleague, or superior) and thereby summarize multiple different people in one category (Elmer et al., 2023). Therefore, the effects of specific interaction partners remain masked even though such information may be relevant for (idiographic) research and personalized therapy (Stadel et al., 2023; von Klipstein et al., 2023). Although it does matter whether one is interacting with a family member versus a friend, it may matter even more whether the person is interacting with the best friend from high school or with a new friend that one met only recently at a sports class.
Second, integration of ESM and PSN can shed light onto bottom-up processes, that is, how social interactions in daily life contribute to the formation or changes of social relationships and the social network of a person, and how these, in turn, are related to personal characteristics of the respondent (e.g., a person’s overall well-being, health). For example, when a longer ESM assessment time frame is chosen (e.g., 4 weeks), it may be fruitful to capture the PSN before and after the ESM data collection. If changes in the PSN occur (e.g., the formation of a new relationship), the ESM data may be able to shed light on the process underlying them.
Third, the integration of ESM and PSN allows for investigating to whom people turn to in particular situations in daily life. Researchers can, for example, identify which specific network members the participant seeks for social support when experiencing personal issues and whether they share certain characteristics.
Fourth, by studying social interactions with specific interaction partners using ESM and PSN, researchers can get a better understanding of top-down processes, that is, how the personal characteristics of participants, such as introversion and extraversion, are associated with changes in interaction patterns. One could, for example, examine whether introverts interact more frequently with network members who are closer or whether they seek out less close network members (e.g., acquaintances) in specific situations.
In sum, the integration of method allows researching different levels of social context, such as how daily social interactions contribute to the formation or change of relationships and how interactions, relationships, and social-network structure are related to individual characteristics (e.g., personality). Figure 1 illustrates these different levels of social context and how the different assessment methods (i.e., surveys, PSN, ESM) are related to the specific levels.
The levels involved in social-context assessment.
ESM and PSN are complex methods in and of themselves (see Myin-Germeys & Kuppens, 2021; Perry et al., 2018). Their applications can become burdensome in terms of time investment for participants and researchers, and setting up a data collection with these methods involves many methodological decisions (Stadel & Stulp, 2022; Trull & Ebner-Priemer, 2009; Vachon et al., 2019). Perhaps this is why despite the potential of integrating ESM and PSN, these methods are rarely used in conjunction.
The few existing studies, however, highlight the diverse applications for such methodology. One pioneering study by Harlow and Cantor (1995) combined pen-and-paper-based social-network data collection and experience sampling to research the functions of specific types of social support in daily life and how that relates to the well-being of a group of female undergraduate students. The Berlin diary study (see Denissen et al., 2008) did not assess an entire personal network of participants but asked participants to report interactions with one family member, one friend, and their partner. Fingerman and colleagues (2020) investigated the relation between the closeness of ties (strong vs. weak ties) in relation to daily activity and mood. For this, they assessed the network of the 10 closest social contacts of their participants, who were defined as strong ties. In a subsequent ESM period, social interactions were classified as taking place with either these strong ties or someone else (i.e., a weak tie). Finally, Lin and Lachman (2021) assessed the number of individuals in the personal network of their participants and used this as a covariate in their analyses of social-interaction ESM data.
To facilitate such innovative research on participants and their social context, user-friendly software that integrates both personal-networks and experience-sampling data as well as methodological guidance are needed. Thus, in this article, we aim to encourage and guide researchers to integrate PSN and ESM to better understand the social context in people’s daily lives. The remainder of the article is structured in the following way. We first introduce existing software that—when combined—facilitates flexible integration of ESM and PSN in a user-friendly way, allowing one to combine the two methods in any order: (a) assessing participants’ PSN after an ESM period and (b) using a repeated PSN assessment before and after an ESM period. Second, we present the result of a study exploring the feasibility of these two combinations of approaches. Specifically, we examine dropout, compliance, and the subjective experience of our participants. We further evaluate differences in the obtained social-network data between two combination approaches to help researchers pick a reasonable approach for their study. Third, we reflect on several study-design considerations (see Table 1) and decisions that need to be made when combining ESM and PSN. Finally, in the discussion section, we outline the broader role of our PSN and ESM integration tool for psychological research and potential future avenues.
Design Considerations When Integrating ESM and PSN
Note: ESM = experience-sampling methodology; PSN = personal social network.
Collecting ESM and PSN With Integrated Software
We developed a user-friendly software integration that can facilitate the collection of ESM and PSN data by combining the experience-sampling application m-Path (Mestdagh et al., 2022) with the Graphical Ego-Centered Network Survey Interface (GENSI; Stark & Krosnick, 2017), which assesses personal networks.
ESM assessments of social interactions with m-Path
M-Path (https://m-path.io) is an online platform for the implementation of ESM and ecological-momentary interventions for research and application in blended care. It can be used free of charge in its basic features but requires a subscription for advanced features (e.g., collecting data for larger samples). M-Path and its functionalities were extensively described by Mestdagh and colleagues (2022).
We collaborated with the developers of m-Path to implement a PSN assessment within their platform because it fulfilled a number of important requirements. First, it is easy to use for users at both ends. Researchers or clinicians can quickly implement state-of-the-art ESM designs. Participants or clients can download the m-Path application on their Android or iOS smartphone, via which they can complete the surveys and receive interventions. Second, m-Path adheres to the standards set by the General Data Protection Regulation (GDPR), the applicable data-protection legislation in the European Union. Data processing and storage are secure, and no direct personal identifiers of participants are collected (data pseudonymization). Third, when assessing social interactions in m-Path, researchers and clinicians have immense flexibility. For example, it is possible to send prompts to participants and ask them about their current company or interactions they had since the last assessment moment (i.e., signal-contingent assessment). Furthermore, it is also possible to let participants initiate assessments themselves after each social interaction by clicking a button in the m-Path app (i.e., event-contingent assessment). In addition, there are a range of different question formats available to assess more details about a social interaction that is captured. The crucial item for the ESM/PSN integration is the one assessing the partner of the social interaction, which can be implemented in m-Path as a multiple-choice item that can be extended with new answer options by the participant throughout the study. In this item, participants can choose between existing multiple-choice options (e.g., names of interaction partners based on a network assessment) or adding new options (e.g., names of encountered interaction partners). Newly added options are saved so that they can be reselected in future assessments. This item acts as a connection between the ESM and PSN surveys and contains names of all interaction partners/network members.
PSN assessments of social relationships with GENSI
As a PSN assessment tool, we used the newest version of the open-source code of GENSI (Stark & Krosnick, 2017; for the adapted version we used, see Stulp, 2021). GENSI uses an interactive graphical representation of the participant’s social network, which is more efficient and enjoyable for respondents than asking them to list names and leads to better data quality than nongraphical assessment methods (McCarty & Govindaramanujam, 2005; Stark & Krosnick, 2017; Tubaro et al., 2014). GENSI provides a template for a name generator (see Figs. 2a–2c), different item types to ask questions about the people included in the network (e.g., their gender or closeness of the relationship; see Figs. 2d– 2g), and a user-friendly assessment of ties between network members (see Fig. 2h). The m-Path developers implemented a network assessment using the open code from GENSI. This network assessment is linked to ESM surveys via the interaction-partner item described above. The network assessment in m-Path is currently tailored directly to the purpose of the research presented in this article, and modification by users themselves is not yet possible. 3
Excerpt from a personal network assessment using the Graphical Ego-Centered Network Survey Interface. In Figure 2h, the color of the circles indicates with whom the selected social contact is connected. Connected network members appear green, and nonconnected members appear red.
Integration of PSN and ESM assessments
To allow for a flexible integration of ESM and PSN, all names filled into the partner item during an ESM assessment period are automatically accessible in the following PSN assessment. The network assessment can then be used to ask about characteristics of the people encountered during the ESM period (see Fig. 3a). In reverse, names added in a network assessment (e.g., a baseline questionnaire) are automatically accessible as answer options of the ESM interaction-partner item. This connection, developed by us, allows for repeated assessments of both ESM and PSN in any order (e.g., a pre- and post-PSN, before and after an ESM period) while always having access to an up-to-date list 4 of individuals that belong to the social environment of a participant (Fig. 3b) without requiring a manual transfer of names by the researcher. We were interested in examining the feasibility of both of these approaches because they may be useful for different research goals.
The two approaches of the experience-sampling-methodology and personal-social-network integration with the Graphical Ego-Centered Network Survey Interface and m-Path.
A First Feasibility Study
In the first feasibility study using the network module integrated into m-Path, we aimed to test how feasible a maximal design including all components of ESM and PSN assessments is. Thus, we designed a rather high-burden study to test the integration of a full PSN assessment (i.e., including multiple name generators, a number of questions about network members, and an assessment of all ties in the network) and an ESM study aiming to capture most social interactions people have in daily life. Furthermore, we explored both approaches of integrating ESM and PSN between subjects: (a) beginning with a personal-network assessment and then using the generated names as possible interaction partners (Fig. 3b) or (b) collecting names during an ESM-assessment period, which is followed by a person-network module in which characteristics of the encountered individuals are assessed (Fig. 3a).
Method
Disclosures
Preregistration
Because of the exploratory nature of this study, no preregistration was done.
Data, materials, and online resources
All assessment materials containing exact item wordings, analysis code, supplementary materials for qualitative analyses, and a detailed description of the data collection can be found on our OSF repository (https://osf.io/jqdr9/). Data are not openly available because of its sensitive nature, but parts of the data may be shared on request if in accordance with participant consent.
Reporting
In our article and supplementary materials, we report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study.
Ethical approval
Our study procedures and data processing and management were carried out in accordance with the ethical guidelines issued by the Council for Social and Behavioural Sciences in the Netherlands and legal requirements laid out by the GDPR. Our research was approved by the local ethical committee.
Sample
In October 2021, 5 we recruited 23 undergraduate students (N = 18 women; Mdn age = 21) from a Dutch University who had an Android smartphone and were willing to use this phone during the study period. We did not recruit participants with an iOS smartphone because we collaborated with researchers interested in collecting passive smartphone data with the BEHAPP app (https://www.behapp.com/), which at the time was available only for Android devices. Our sample size was guided by constraints in terms of resources available to the research team. Students were compensated either financially or with course credits. The amount of compensation was tied to the completed study parts, with a maximum of 50€ or 7.3 credits.
Procedure
For an overview of our study procedure, see Figure 4. After sign-up and filling in an online baseline questionnaire, students were divided into groups, one starting with a PSN assessment (Group 1: PSN→ESM→PSN; Fig. 3b) and the other starting with an ESM period (Group 2: ESM→PSN; Fig. 3a). We aimed to compose the groups as similar as possible in terms of sociability, with each group consisting of both very socially active and minimally socially active individuals based on self-reported social behavior in the baseline questionnaire.
Procedure of the feasibility study. The symbols indicate the mode of assessment. Laptop = online survey on a computer; phone = experience-sampling-methodology assessments on participants’ own smartphones; speech bubbles = in-person instruction/interview.
Each participant was invited for a group instruction session, during which we explained the procedure further and obtained informed consent. Afterward, the PSN→ESM→PSN group completed an online personal-network assessment. Both groups started a 28-day ESM period on the day after the instruction session. Directly after the ESM period, both groups were asked to complete an online PSN assessment, with the PSN→ESM→PSN group supplementing the previous assessment and the ESM→PSN group doing the complete assessment for the first time. For both groups, all social-interaction partners added via the ESM questionnaires were piped into the network survey (e.g., if Anna was named in the ESM survey but not in the pre-PSN assessment, the name “Anna” would automatically appear in the post-ESM PSN assessment). After completing this PSN assessment, participants also reported on their study experience during a brief online questionnaire.
Two weeks later, all participants completed a personal-network assessment via a one-to-one in-person interview. The network data obtained from this interview serve as a benchmark for the data obtained from our integration procedure: To evaluate the data quality of the online PSN assessments, we compared these networks with the one obtained during the interview—which is currently the “gold standard” for assessing personal networks (Perry et al., 2018). Toward the end of the network interview, we showed the participants the network obtained from the online surveys and asked them to clarify discrepancies between the online network and the interview network, such as missing names or notable differences in relationship-quality ratings.
Assessment materials
PSN survey
Each network survey (pre-PSN, post-PSN, and interview) consisted of three stages: name generation, questions assessing characteristics of network members, and the assessment of ties between network members.
Name generation
For our feasibility study, we decided to capture the social context of our participants as comprehensively as possible by using three questions to generate names for the social network (McCarty et al., 2019). Initially, participants were asked to come up with names of people they “interact with during daily life.” Then, we asked for names of individuals with whom they “have less frequent contact with, but who are still important contacts they could reach out to if needed.” In the last step, we asked participants to check the GENSI visualization of their network and to add people that they felt would still belong to their social network. Participants were encouraged to use their smartphones or any other resource to help them remember individuals that are part of their personal social network.
Characteristics of network members
We assessed demographic characteristics (e.g., age and gender) of network members and several ratings of the relationship between the participant and the network members. For examples of personal-network items, see Figure 2b. To facilitate the integration of ESM and PSN, we chose items that connect to each other (e.g., PSN item: “This person gives me energy”; ESM item: “This interaction did give me energy”). A full list of items used can be found on OSF (https://osf.io/m9y7h).
Ties between network members
Given that the goal of our study was testing the feasibility of a design that includes all possibilities that ESM and PSN offer, we opted for assessing ties between all the network members as shown in Figure 2h.
ESM period
Sampling schedule
The ESM period lasted 28 days. We used a combination of two semifixed assessment moments (i.e., set moments when the morning and evening questionnaires became available), four equally spaced semirandom measurement prompts throughout the day, and event-contingent recording of social interactions (i.e., participants initiate an assessment themselves after the occurrence of a social interaction).
The morning questionnaire became available at 6:00 a.m. and stayed accessible until 10:30 a.m. We instructed participants to complete the questionnaire as soon as they got out of bed and were ready to start their day. In case participants did not respond to it until 10:00 a.m., we sent a reminder. The following four questionnaires were sent at semirandom times covering the whole day (between 10:30 a.m. and 12:30 p.m., between 1:00 p.m. and 3:00 p.m., between 3:30 p.m. and 5:30 p.m., and between 6:00 p.m. and 8:00 p.m.). Participants had 30 min to complete these questionnaires; after 15 min, we sent a reminder. At 9:00 p.m., the evening questionnaire was sent. Participants were instructed not to respond to this prompt immediately but only shortly before going to bed. The questionnaire stayed available until 4:00 a.m. A reminder was sent at 11:00 p.m.
For the event-contingent assessments, participants were instructed to report any face-to-face, video, or phone-call interaction they had that lasted longer than 5 minutes. 6 We were interested in all one-on-one and group interactions with verbal exchange. For cases in which participants spent a longer time frame with someone and there was a break of at least 15 min in between verbal interactions, we instructed participants to log each interaction separately. When breaks were shorter or participants had a continuous verbal exchange, the whole period should be logged as one interaction. The signal-contingent prompts included a reminder about reporting social interactions and allowed participants to do so immediately. Participants were, however, instructed to primarily report event-based assessments. With this combination of event- and signal-contingent assessments, we aimed to capture most daily social interactions while assessing affect and other variables also in moments outside of social interactions.
Questionnaire content
The signal-contingent questionnaires assessed momentary positive and negative affect (Barrett & Russell, 1998), activities since the last beep, whether the participant is currently alone, and how being alone/in company is perceived. Then, the reminder for reporting social interactions followed (see above). The morning and evening questionnaires included a few additional items (e.g., about sleep quality). The event-contingent social-interaction questionnaire assessed momentary affect and detailed information about the social interactions (interaction partner, 7 mode, timing, duration, type, content, and the participant’s perception of the interaction). For the exact items and response formats, see OSF (https://osf.io/58wcx).
Compliance feedback
Participants received an email each week containing their compliance rate and the number of recorded social interactions. In case no interactions were reported, we asked participants whether they experienced any technical difficulties. To receive full compensation for the ESM period, at least 75% of all signal-contingent assessments needed to be filled in.
Personal-network interview
For the in-person one-on-one personal-network interview, the name generators and attribute questions were identical to the ones used in the online network survey but were implemented in the software NetworkCanvas (Birkett et al., 2021), an established state-of-the-art personal-network data-collection software, and supplemented with more detailed instructions by the interviewer. When checking whether the names of all social contacts were reported, the interviewer read out a list of possible social roles to the participant and still added forgotten names. The assessment of ties between all the network members was designed more interactively, and the interviewer and participant mapped out the social network together until the participant was satisfied with the placement of each network member.
Immediately after this process was completed, the interviewer ran an R script comparing the data from the post-PSN assessment with the personal network obtained during the interview. Then, the participant and interviewer would discuss identified differences together and note down explanations provided by the participant. This was done to be able to determine the quality and validity of the post-PSN data.
Experience questionnaires
After the ESM/PSN integration and after the interview, participants were presented with a questionnaire assessing their perception of the study procedures. Besides positive and negative feedback, we asked about technical problems and improvement suggestions. The questionnaires included a mix of items rated on a 11-point Likert scale ranging from not at all (0) to very much (10; e.g., “I feel like the diary in combination with the network captured my social environment adequately”) and open questions (e.g., “What did you like about the diary in combination with the network?”). A complete list of questions can be found on OSF (https://osf.io/58wcx).
Analysis
Data quantity and quality
To evaluate the feasibility of combining ESM and PSN, we examined the quantity and quality of the data we collected. Specifically, we determined dropout and ESM compliance rates and absolute numbers of social interaction reports and reported network members/interaction partners. In addition, we examined the time it took participants to report social interactions using the event-contingent ESM questionnaire. For the duration calculation, we included only observations with a minimum duration of 20 s (time needed to just click through the full questionnaire) and a maximum of 10 min (we assume that in such a case, participants put their phone away while leaving the assessment open).
Furthermore, we explored the differences between the two integration approaches (see Fig. 3) in terms of the number of reported network members at each network-assessment moment. Given our small sample size and the exploratory nature of our study, we do not report on inferential statistics comparing the two study groups but, rather, describe effect sizes in terms of median differences.
Finally, we examined the results of the qualitative comparison between the post-PSN and interview network. Specifically, we determined the number of differences and summarized the reasons participants indicated for discrepancies.
Participant experience
To gain more insight into how participants experienced the study participation, we summarize the responses to the experience questionnaires descriptively. For the quantitative ratings, we examined median differences between the two integration approaches, and all qualitative responses were inductively grouped into themes by one researcher. Because the data snippets were relatively short, no formal qualitative analysis approach was used. In addition, we explored the relation between the quantitative experience ratings and the number of network members, the number of reported interactions, and ESM compliance by calculating Spearman’s rank correlations and visually examining the relationship.
Results
Data quantity and quality
Dropout
Our aim was to recruit 24 participants who would take part in the full study procedure. However, a number of participants were not willing to continue participation after the brief online baseline questionnaire. In total, we needed to recruit 34 participants to reach our target of 24 who committed to the lengthier study procedure including ESM and PSN assessments. Once participants committed to full participation, they remained engaged: All but one participant from the instruction session—who decided not to participate because of hesitance to use the smartphone frequently—continued with the rest of the study and completed all parts.
ESM data
Compliance with scheduled assessments
Overall compliance was high, with an average of 82.7% (minimum = 47.6%, maximum = 95.8%; see also Fig. 5), which corresponds to 134 out of 186 signal-contingent ESM assessments being filled in. Four participants did not meet the minimum compliance rate of 75% by the end of the study. Setting a minimum compliance rate for full compensation appeared to be useful because some participants who barely missed the requirement at their Week 3 feedback increased their compliance during Week 4.
Compliance rates per participant across the study. N = 24. Each thin line represents one participant. The thick blue line represents the mean compliance, and the gray dashed line indicates the required compliance rate for full compensation.
Reported social interactions
We captured a total of 1,714 social interactions across 23 participants and 28 days. On average, a participant reported 75 social interactions (Mdn = 69, SD = 34), and half of these were logged in an event-contingent way (M = 38, Mdn = 25, SD = 36); the other half was reported in response to the signal-contingent reminder prompts (M = 36, Mdn = 31, SD = 22). Reporting an interaction took participants on average 83 s (SD = 57, minimum = 21, maximum = 499).
When examining the differences between the two combination approaches of ESM and PSN, the two study groups were very similar in the total numbers of reported interactions (Mdn = 69 vs. Mdn = 71) and the duration of a social-interaction assessment (Group 1: M = 85.06; Group 2: M = 80.23). However, there are differences when splitting interactions by reporting mode, with the PSN→ESM→PSN group reporting interactions more often via signal-contingent reminder prompts (Mdn = 35 vs. Mdn = 25) and the ESM→PSN group reporting interactions more often via event-contingent assessments (Mdn = 20 vs. Mdn = 36). In addition, the PSN→ESM→PSN group reported more unique interaction partners compared with the ESM→PSN group (Mdn = 53 vs. Mdn = 32); see also Figure 6.
Numbers of reported interactions, unique interaction partners, and network members split by group. The total number of interaction partners is the count of unique names added at any point during the study; the numbers of names per network assessment include only the names specifically added or retained at that assessment moment.
When examining the data per participant, it becomes apparent that this difference is driven by between-subjects preferences in reporting mode—some participants predominantly chose to report event-contingently, whereas others predominantly used the signal-contingent reminder prompts despite being instructed not to do so. 8
Personal-network data
PSN→ESM→PSN versus ESM→PSN
Figure 6 shows the number of reported network members for each network assessment split by group. The pre-PSN assessment was by design completed only by the PSN→ESM→PSN group. During this assessment, participants included a median number of 27 names. Both during the post-PSN and the interview network assessment, this group reported higher median numbers of network members than the ESM→PSN group (Mdn = 44 vs. Mdn = 25 and Mdn = 36 vs. Mdn = 26, respectively), indicating the added value of a pre-PSN.
Post-ESM networks versus interview networks
Averaged across the 23 participants, the post-ESM network contained a higher number of network members than networks obtained through a face-to-face interview 2 weeks later (Mdn = 37 vs. Mdn = 30). When we compared these two networks with the input of the participant during the interview, it became apparent that the network members considered as the closest by participants usually overlap. In a few instances, missing individuals in either network were simply forgotten to be added during the assessment. In addition, participants confirmed a few actual changes in their social environment that were reflected in the network differences (e.g., meeting new fellow students because of starting a new course). According to the participants, the largest difference, however, seems to be rooted in the connection of the post-ESM network to the ESM assessments. The post-ESM network included more network members for almost all participants, and almost all of the additional names were added based on social interactions. At the same time, the post-ESM network omitted a few network members that participants did not have recent contact with (e.g., friends from their home country) despite them still being important to the participant. Overall, however, the post-ESM network assessment is casting a wider net and captures more distant social contacts encountered during social interactions. For a detailed overview of the network differences per participant categorized by the reason the participant gave for each difference, see the supplementary materials on OSF (https://osf.io/cazke).
Participant experience
Participants overall considered the ESM and PSN completion as positive (Mdn = 7 for both ratings) and enjoyed it (PSN: Mdn = 7; ESM: Mdn = 6). The PSN→ESM→PSN group gave somewhat lower ratings for the overall ESM experience (Mdn = 7 vs. Mdn = 8), the PSN experience (Mdn = 7 vs. Mdn = 8), and the PSN enjoyment (Mdn = 6 vs. Mdn = 7) than the ESM→PSN group (see Fig. 7). The ESM was enjoyed slightly less by the ESM→PSN group (Mdn = 7 vs. Mdn = 6). Both groups considered their social environment equally well captured (Mdn = 8) and regarded the ESM assessments equally insightful (Mdn = 8). The PSN→ESM→PSN group considered the network assessments slightly more insightful (Mdn = 8 vs. Mdn = 7.5). Overall, the ESM assessments somewhat obstructed participants daily life (Mdn = 6) but mainly in the PSN→ESM→PSN group rather than the ESM→PSN group (Mdn = 6 vs. Mdn = 3.5).
Quantitative responses experience questionnaire. The response scale ranged from 0 to 10. For exact item wording, see supplementary material on OSF (https://osf.io/58wcx).
To see whether the reported experience is connected to characteristics of the social environment of participants, we explored the relation between the quantitative experience ratings and the number of network members, the number of reported interactions, and ESM compliance by calculating Spearman’s rank correlations and visually examining the relationship. 9 We observed an overall positive relation between the number of reported interactions and how much daily life was obstructed by the ESM assessments, r(21) = .29. In line with that, the number of reported interactions and the overall ESM experience, r(21) = –.17, and enjoyment, r(21) = –.30, are negatively related; two are positively related to compliance: overall experience, r(21) = .46, and enjoyment, r(21) = .35. Likewise, the network assessment was enjoyed less, r(21) = –.33, and experienced less positively, r(21) = –.44, by participants reporting more network members.
In the responses to the open questions, most participants positively noted gaining insight into their social life and enjoying the opportunities to reflect. Furthermore, participants complimented the easy-to-use interface. As points for improvement, three participants suggested decreasing the frequency of ESM questionnaires. Moreover, three topics that were difficult to deal with during the social-interaction assessments came up: (a) determining what counts as an interaction, (b) describing the whole content of an interaction, and (c) dealing with group interactions that involve many different partners. Three participants noted negative consequences, such as using their phones more and feeling discouraged from social interaction because of having to report on them. For an overview of how often each theme was mentioned, see supplementary information on OSF (https://osf.io/kn5j8).
Design Considerations When Integrating ESM and PSN
While we designed and conducted the first feasibility study testing the ESM/PSN integration, we were confronted with several design considerations. Table 1 provides an overview of the seven main considerations we had to make. In the following, we elaborate on these and provide guidance to researchers planning a study using an integration of ESM and PSN.
Consideration 1: the kind of personal network
A personal-network assessment can concern a small, well-defined group (e.g., one’s five closest friends) or capture the wider environment also including weaker ties (e.g., all individuals one interacts with in daily life). Researchers need to decide on the type and the number of social contacts they want to capture with their network assessment (Bidart & Charbonneau, 2011; Hogan et al., 2007; Marin & Hampton, 2007; Vehovar et al., 2008). The personal-network literature provides different validated options for so-called name-generator items (see Perry et al., 2018). Choices will be mainly guided by the research question at hand, but especially when choosing the number of names, participant burden also needs to be considered. When increasing the number of names, the remainder of the survey becomes increasingly longer because all further questions need to be answered about each name (see Golinelli et al., 2010; McCarty et al., 2007). Concrete estimations of the burden of a GENSI network assessment in terms of time for different network sizes were provided by Stulp (2021) and Stadel and Stulp (2022).
Consideration 2: items in PSN and ESM
In both PSN and ESM, item choices will be tailored to the constructs a researcher is interested in. When choosing ESM items, the ESM item repository can be a good starting point (https://esmitemrepositoryinfo.com/) to facilitate replicability and standardization. Using the same constructs in PSN and ESM allows for comparison between the view of a participant in a baseline personal network and the recorded social interactions in daily life. Assessment of the same construct on both levels also enables studying the same social process at different timescales (within hours and days with the ESM item and within weeks or even months with the PSN item).
Consideration 3: ties between network members
In personal-network assessments, ties between social contacts that were listed can be assessed (Lackaff, 2012). Ties can, for example, be defined as social contacts either knowing one another or interacting with each other in daily life. Based on the ties between the social contacts, structural characteristics of the network, such as density or the centrality of specific network members, can be determined (see e.g., Wasserman & Faust, 1994). However, this part of the data collection can be particularly burdensome. Different approaches to minimize the burden while maximizing data quality have been proposed (see Golinelli et al., 2010; McCarty et al., 2007; Stadel & Stulp, 2022). Using an interactive graphical way to assess ties such as the one implemented in GENSI (see Fig. 2h) or Network Canvas (Birkett et al., 2021) is recommended. Still, researchers should carefully consider whether a tie assessment is necessary to answer the research question and whether the burden to the respondent is acceptable.
Consideration 4: defining social interactions
When capturing the daily social life of a participant using ESM, a central goal is to assess social interactions. Thus, it is vital to define what type of social interaction should be included so that interactions can be compared across and between individuals. Mostly, ESM research focuses on verbal-communication episodes (Goldsmith & Baxter, 1996) as opposed to nonverbal social interaction. A common definition of social interaction is a (verbal) exchange between the participant and at least one other person lasting longer than 5 min (see e.g., Himmelstein et al., 2019). Note that this includes group-based interactions. At the same time, researchers can also decide to focus solely on one-on-one encounters. Another consideration is whether to include interactions with strangers or purely role-based interactions (e.g., with the barista at a coffee shop; see Hall, 2018). In addition, social interactions can take place face-to-face or via digital tools in the form of audio/video calls, texts, or voice messages. It is important to define which modes of interactions participants should report on—the quality of interactions may vary across modes (Achterhof et al., 2022).
Consideration 5: experience-sampling design
When aiming to capture social interactions with ESM, there are two possible sampling designs: (a) prompting participants multiple times per day, assessing interactions since the last prompt or at the current moment (i.e., a signal-contingent design), or (b) asking participants to record social interactions directly after they occurred (i.e., an event-contingent design; see Myin-Germeys & Kuppens, 2021). First studies indicate that the choice of sampling design matters for the number of total social interactions obtained (Himmelstein et al., 2019) and the number and quality of social interactions involving less frequent interaction partners (Stadel et al., 2022). Affect ratings and overall interaction quality appear, based on the limited number of studies available, not to be affected by sampling-design choice (Himmelstein et al., 2019; Stadel et al., 2022).
Especially when deciding on a signal-contingent design, the time period covered will influence the number and type of captured social interactions (Stadel et al., 2022). The timing of ESM prompts should be aligned with the expected time in which social interactions happen in a particular population, or participants could be allowed to choose their time interval aligning with their personal daily rhythm. In a signal-contingent design, the number of prompts should be determined according to the expected frequency of social interaction.
In a signal-contingent framework, it is common practice to ask for the most important (instead of the most recent or all) social interactions since the last beep. Often, it is not clearly operationalized what “important” in this context means. If assessing a subsample of social interactions, we would recommend determining clear criteria for this subsample (e.g., only interactions with the three predefined closest network members). In any case, we would advise against setting a minimum requirement for the number of reported social interactions because there is great between-persons variability in social activity; thus, setting a minimum requirement can be confrontational. One potential future development is that ESM prompts may be triggered by specific situations a participant is in (e.g., particular GPS locations or speech detection via the smartphone’s microphone; Bachmann, 2015; Bachmann et al., 2015; Mehl, 2017).
Consideration 6: overall integration design
When integrating PSN and ESM, researchers have to decide whether to start with PSN or ESM assessments. Starting with a PSN or an ESM assessment both have benefits and challenges. Suppose participants first complete a personal-network survey. In that case, they have a list of names to select their interaction partners from during an ESM period, which saves time in the moments of assessing interactions. The researcher could also elicit specific types of individuals in the network assessment (e.g., family members or individuals who provide social support) and ask participants to report social interactions with only these people (see Denissen & Penke, 2008; Harlow & Cantor, 1995). Moreover, when the network is assessed before an ESM period, it represents participants’ intuitive view of their social environment. In contrast, network assessment after completing an ESM period is likely influenced by social interactions during that period. When starting with a PSN assessment, it may be that participants forget important individuals who they still encounter in daily life, which would be captured with ESM. Hence, the researcher needs to make a decision whether to allow adding individuals to the list based on ESM-captured interactions.
Suppose participants start with an ESM assessment phase. In that case, this phase can serve as a way to gather a list of relevant names for a network assessment, which is particularly useful when the researcher is interested in individuals with whom the participants have daily contact. Using this strategy, researchers also have the option to already filter interaction partners before a network assessment (e.g., based on interaction frequency or ratings on specific items).
Finally, the personal network can also be assessed repeatedly (e.g., before and after ESM), which is useful for capturing changes and how these connect to social interactions or anything else happening in daily life. This method combination can deliver the richest information and also allows assessing effects of interventions during the ESM period on the social network.
Consideration 7: ethics and data management
When collecting data about social relationships and social interactions in daily life, depending on the questions asked, researchers obtain detailed and potentially sensitive information about a participant (e.g., sexual or political orientation or sexual relationship and activity; Hogan, 2021). Inviting participants to an instruction session for a detailed explanation is essential. In these sessions, it should be stressed that full anonymization of the data is not possible (Tubaro et al., 2021), but pseudonymization techniques can allow for data sharing. Participants should be fully informed about the steps taken to anonymize the data. We recommend using visual representation and practical walkthroughs of the assessments and data-processing steps. We provide our information form and instruction materials in the supplementary materials on OSF (https://osf.io/jqdr9/files/osfstorage).
One particularly sensitive issue is asking for the names of social contacts, which represent individuals who did not consent themselves to participate in the study. For the assessment, it is only important that the participants recognize names and that different interaction partners can be distinguished. Participants should thus be asked to name nicknames or first names only—instead of full names. Asking for initials often leads to overlap (Jane Doe and Jon Deen both have JD) and is unintuitive for participants to use during data collection. It requires further cognitive effort to remember who was meant and increases the participant burden. After collection, the names should be exchanged for IDs during data preprocessing. This also holds for names listed in the qualitative comments during ESM assessments.
Discussion
In this article, we advocated for an integration of ESM and PSNs to study different levels of participants’ daily social context. To facilitate studies combining the method, we introduced a flexible software that can be used for collecting ESM and PSN data in conjunction using the ESM software m-Path. The software allows one to automatically transfer names of interaction partners/network members between ESM and PSN surveys.
We presented a first study with 23 student participants illustrating the feasibility of the methodology and software. For each participant, we collected up to two personal networks and 28 days of affect and daily activities via six signal-contingent ESM surveys per day and social interactions via event-contingent reporting. Even though our design was intensive, our results showed that student participants did not experience the study as too burdensome. On the contrary, the majority of participants considered their participation insightful. All participants were retained during the data-collection procedure, and average compliance rates were high. Thus, we demonstrated that a rather high-burden study is feasible and that the time investment from researchers and participants delivers a rich data set: Across 23 participants, we recorded 1,714 social interactions and 1,004 network members/interaction partners. During the 4-week data-collection period, a participant reported, on average, 75 social interactions (i.e., approximately 19 interactions per week). This number seems adequate compared with previous research, such as a study that examined event- and signal-contingent sampling of social interactions by Himmelstein and colleagues (2019). During this study, participants reported, on average, 19 social interactions during 1 week when assessed in a signal-contingent manner. When assessed event-contingently, participants of Himmelstein and colleagues reported, on average, 28 interactions across 1 week. Our design assessed social interactions in an event-contingent design but also included signal-contingent questionnaires assessing mood and activities and reminding participants about reporting interactions. Given the higher burden during our data collection and the fact that it took place during the COVID-19 pandemic, with restrictions on daily social life, the lower interaction frequency makes sense.
Half of our participants were asked to provide two personal networks, before and after an ESM period, and the other half completed only one network assessment after the ESM period. This design allowed us to examine two different approaches of integrating ESM and PSN: (a) starting with a PSN assessment that maps important social network members/potential interaction partners, followed by a 28-day ESM period and concluding with a follow-up network assessment, which allows to capture change in the network; and (b) starting with a 28-day ESM period that captures names of interaction partners/network members via the reporting of social interactions followed by one PSN assessment. The social environment of participants seemed to be more comprehensively captured with the first approach. This approach delivered larger social networks, and participants felt like their social network was represented somewhat more accurately. This added accuracy came at the cost of a higher participant burden, which was, however, still considered acceptable. The second approach delivered somewhat smaller networks but still captured the social contacts closest to the participant. Thus, the approach of assessing a network only after an ESM period is a suitable alternative when no comprehensive network is required and the participant burden needs to be reduced. Yet this approach does not allow for investigating changes and effects of interventions in the social network of participants.
On the basis of our feasibility study, we also reflected on different design considerations when integrating PSN and ESM. Our reflections sketch the wide array of possibilities that the method combination provides: Many variants of ESM, PSN, and the combination of both are possible and come with various degrees of participant and researcher burden. We encourage future research to explore this space.
After the feasibility study, we continued to develop the methodology by implementing a few changes, including (a) a less demanding ESM schedule (five assessments instead of six); (b) the possibility of directly adding multiple new interaction partners to the interaction-partner item via clicking a little plus sign; (c) the interaction-partner item containing a scrollable, alphabetically sorted list of all possible interaction partners; and (d) the option to delete social-network members in the post-ESM network assessment. These changes appeared to be useful (see Stadel et al., 2023 and supplementary material on our OSF repository [https://osf.io/jqdr9/]). However, particularly the assessment of social interactions in daily life using ESM remains challenging, and many methodological questions, such as which sampling scheme and design should be used, are still understudied and would benefit from further research. With our materials, software, and guidance about design considerations, we hope to aid future research and eventual practical applications integrating ESM and PSN to gain more insight into daily social dynamics.
Limitations
When interpreting our results, some limitations need to be addressed. First, during (part of) our data collection, COVID-19 restrictions, such as a 1.5-m distance and wearing face masks in public buildings, were applicable. Toward the end of the data collection, after the final interviews took place, a second lockdown started. The social-distancing measures likely influenced the daily social life of our participants; thus, the data we collected may not be applicable to circumstances in which no restrictions are in place. When asked during the interview, students indicated that they tried to live their “regular” lives as much as they could and tried to stay socially active.
Second, for this feasibility study, we tested the methodology with a young, mostly female, highly educated, and likely highly digitally skilled sample. Our approach is probably more demanding and time-consuming for people with fewer digital skills. For such populations, it may be necessary to use a less intense ESM sampling scheme or restrict the number of social-network members/interaction partners that should be named. Thus, future applications should keep characteristics of their sample in mind and adapt the study procedure accordingly.
Third, given that participants completed their last PSN assessments only 2 weeks before the interview-network assessment, it is likely that the latter was influenced by the former, which may lead to the personal network of participants being biased toward individuals that they meet frequently.
Future application in (clinical) practice and research
The integration of the methods of ESM and PSN allows researching multiple levels of participants’ daily social context, social interactions, and social relationships in a dynamic fashion. Although potential applications of such methodology in the social sciences are diverse, we highlight one particular future avenue in a clinical context. Increasingly, ESM is explored as a clinical tool aiding personalized assessment and treatment (see e.g., Piot et al., 2022; von Klipstein et al., 2023). In such applications, specific information on social relationships and daily social interactions of a patient is relevant. An ESM/PSN integration allows zooming into the effects of particular network members, providing detailed insight into daily social functioning of patients with different individuals. The collected data of a patient can be summarized in a personalized feedback report that may aid communication between patient and therapist and support finding fruitful (interpersonal) intervention targets (Stadel et al., 2023). Such feedback reports can also serve as a form of compensation, motivating participation in assessments (Rimpler et al., 2024). Thus, the methodology proposed in this article can help to understand how social relationships and daily interactions relate to poor (mental) health and what kind of (interpersonal) intervention is needed. In addition, effects of interventions (e.g., interpersonal therapy for depression; Klerman & Weissman, 1994) on social relationships and daily interactions with specific partners can be examined more closely. This may help researchers to identify specific mechanisms through which they are effective.
While in clinical practice idiographic assessment and analyses become increasingly popular, research remains interested in nomothetic goals. When collecting ESM and PSN data for such purposes, data processing and analysis are not trivial. One possibility of analyzing the resulting data is using multilevel models with social interactions being modeled at the lowest level, interaction partners/network members at the second level, and participants at the third. For an illustration of such an analysis approach, see Stadel and colleagues (2022).
Conclusion
In this article, we described how researchers can combine ESM and PSNs to collect data on different levels of participants’ daily social context. Such data have various applications in research and practice. Our feasibility study presents a first step in exploring the potential of this combination. With the presented software and design considerations, we hope to encourage future research exploring and developing this methodology.
Footnotes
Acknowledgements
We thank Anna Langener for her collaboration during the data collection for this project.
Transparency
Action Editor: Rogier Kievit
Editor: David A. Sbarra
Author Contributions
