Ambassador,you’re on Mute: Methods for Co-interviewing Elites Online

What we know and do not know about online interviews

Interviews conducted online have become more common, particularly in recent years. Videoconferencing tools gained great prominence during the COVID-19 pandemic, and in many places, reliable Internet access is no longer an issue, especially for elites. Two decades before COVID-19, however, an early literature on online interviews was already emerging, in line with the technological developments of the time. In this first wave of literature, researchers often focused on establishing the validity and legitimacy of such approaches and extolling the potential advantages of accessibility, lower cost, and time saving, while conceding that online approaches might not fully substitute for in-person interviews (Anderson & Kanuka, 2003; James & Busher, 2006, 2009; Madge & O’Connor, 2002; Mann & Stewart, 2000; O’Connor et al., 2008). To a large extent, however, online interviews were conceived of as a facsimile of the in-person, face-to-face interview. Even a decade after some of these initial works, some methodologists still spoke in terms of the ‘translation’ of face-to-face methods to the online environment (Hooley et al., 2012, p. 66), rather than conceiving of the online interview as a form of qualitative research with distinct attributes. Meanwhile, some skeptics questioned the reliability of online methods (Beaulieu, 2004) or argued that the authenticity of online respondents’ views might be problematic (Markham, 2008).

As online interviews became more common, the second wave of literature on online methods emerged. This work engaged more deliberately with the methods of both asynchronous and synchronous online interviews. Much of this literature discusses mitigating the shortcomings of online interviews, such as the technological limitations of both participants and various web and software platforms (Deakin & Wakefield, 2014; Hanna, 2012; Janghorban et al., 2014; Lo Iacono et al., 2016). These studies highlight a range of technical challenges, including the potential for glitches to interrupt the flow of the interview and the quality of data collection (Deakin & Wakefield, 2014; Lo Iacono et al., 2016), as well as the effects of differing levels of technological competence and accessibility on the representativeness of the respondent pool (Deakin & Wakefield, 2014). Common approaches to mitigation of technical challenges include allowing respondents more time to answer, being prepared to repeat or simplify questions, and being as clear and concise as possible in research encounters. Beyond technical limitations, some authors offered a more critical perspective on the implications of conducting online interviews. Salmons (2015) asked researchers to reflect critically on their motivations for conducting interviews online, rather than via conventional means. Weller (2017, p. 622) argued that some of her research participants seemed to treat online interviews as more ‘informal encounters’, making the boundaries between researcher and participant more blurred. Scholars differed in their assessments of the advantages and disadvantages of online research for discussing sensitive subjects (Deakin & Wakefield, 2014; Seitz, 2016, p. 232) or for engaging hard to reach populations (Jenner & Myers, 2019; Wilkerson et al., 2014).

What we term the third wave of literature on online methods developed as a result of COVID-19. Beginning in 2020, work emerged on how to continue qualitative data research despite social distancing, travel, and other public health restrictions affecting both researchers and research participants (Keen et al., 2022; Lawrence, 2020; Lobe et al., 2020; Lupton, 2020; Saarijärvi & Bratt, 2021; Self, 2021; Sy et al., 2020). With the normalization of practices of working from home and at distance both because of pandemic-era restrictions as well as shifting societal norms, there is a clear temporal demarcation between earlier methods literature on online interviewing and this more recent wave of scholarship. What characterizes the third wave of literature is a growing recognition that online interviews need not be considered second-rate compared to in-person interviews, and may in some circumstances even be superior (Archibald et al., 2019; Hyde & Rouse, 2022; Zadkowska et al., 2022, p. 2). For example, Oliffe et al. (2021, p. 3), recounted online research participants as being more comfortable speaking from their homes, thus leading to richer conversations. Similarly, Wahl-Jorgensen (2021, p. 375) identified ‘novel intimacies’ in speaking to respondents via Zoom (see also Maulana, 2022, p. 397). Moreover, several scholars argue that discussing sensitive subjects was often easier in an online setting (Opara, Spangsdorf, & Ryan, 2023, p. 567; Sipes, Roberts, & Mullan, 2019). While some scholars voice scepticism or concerns with online interviews (e.g., Howlett, 2021; Teti et al., 2021, p. 3), in the current wave of literature, there appears to be a growing understanding that online interviews are methodologically different than in-person interviews, and have their own logic and conventions (Gray et al., 2020; Khan & MacEachen, 2022; Roberts et al., 2021; Thunberg & Arnell, 2022).

One conspicuous absence from the literature on online interviewing is a focus on co-interviewing, and in particular the interaction between and strategies of interviewers in co-interview settings. Although co-interviewing – when two researchers actively take part in the same interview – is a common technique, the literature on online co-interviewing is limited (Velardo & Elliott, 2021, p. 1). Nonetheless, the advantages of this type of interview, sometimes called the tandem interview (Groholt & Higley, 1972; Kincaid & Bright, 1957a, 1957b) or the ‘two-to-one interview’ (Monforte & Úbeda-Colomer, 2021) have long been recognized (see also Bechhofer et al., 1984).

Advantages of co-interviewing include the increased efficiency of data collection processes (Kincaid & Bright, 1957a), the application of interviewers’ varying perspectives to be brought to a single interview (Kincaid & Bright, 1957a, p. 306; Monforte & Úbeda-Colomer, 2021, p. 6; Rosenblatt, 2012, p. 103), better follow-up during the interview on unanticipated lines of inquiry (Kincaid & Bright, 1957b, p. 308), more gender sensitive and culturally situated research (Redman-MacLaren et al., 2014) and improved rapport between interviewers and respondents (Velardo & Elliott, 2021, p. 4), as well as making interviews more conversational (Bechhofer et al., 1984, p. 98). Inclusion of specific roles or persons, such as more senior researchers, may increase the willingness of respondents to participate in an interview (Redman-MacLaren et al., 2014). However, unequal power distribution between interviewers may lead to dominance by one interviewer (Velardo & Elliott, 2021, p. 5). Co-interviews also bring some logistical challenges, such as higher demands for coordination and preparation, which should be considered in the design and planning of an interview (Monforte & Úbeda-Colomer, 2021; Velardo & Elliott, 2021).

With the research on co-interviewing focusing squarely on in-person interviews, we argue that outlining strategies and tactics for co-interviewing in online settings is valuable for two key reasons. First, online settings enable researchers to collaborate regardless of their geographical location, making co-interviewing more feasible and, we anticipate, potentially more common. Thinking through how to conduct co-interviews online is thus an important step in developing the online interview methodology, and could help researchers to think more purposively about both how to work with their collaborators, and to prepare for interviews, beyond substantive content. Second, interpreting non-verbal communications can be more challenging in online settings. While the impact on interviewer-respondent dynamics is well-documented (Salmons, 2015; Seitz, 2016), this challenge also extends to interactions between interviewers in co-interview settings. Conducting co-interviews online therefore requires specific strategic and tactical approaches that take advantage of the opportunities while mitigating the challenges posed by this setting. Based on our own experiences conducting online co-interviews with elites and informed by online interview research and the limited in-person co-interviewing literature, we now outline our research study, before turning to explain our co-interview strategies and tactics.

Strategy 1: Insider-Outsider

The insider-outsider distinction relates to the perceived identity of the researchers from the perspective of the interviewees. In our case, the main insider-outsider dimension was between being peace practitioners and academic researchers in peace and conflict studies—roles that share an interest in understanding conflicts and their resolution, but pursue quite different activities, goals, and priorities. We anticipated that respondents would mostly treat us as outsiders, as we assumed the role of academic researchers in our interview requests and were not visibly practice orientated. As others have recounted, being an outsider in a qualitative research interview is not necessarily problematic (Lawrence, 2020, p. 156; Maksutova & Janoff, 2022; Martinez-Merino et al., 2023), but in our experience, some respondents who perceived us as outsiders would assume that we were essentially ignorant of basic dynamics and constraints of the OSCE’s SMM operation, or would be unable to understand more subtle details of their experience. Accordingly, some responses were quite general, defensive, or overly simplified.

At the same time, we agree with those scholars that reject conceptions of insider or outsider as clearly binary (Aguinis & Solarino, 2019; Dhillon & Thomas, 2019; Hampshire et al., 2014; Herod, 1999; Markiewicz, 2023; Mikecz, 2012); we agree that researchers can and do position themselves along a continuum of insider-outsider perspectives. In our study, establishing an insider positionality was largely based on our prior experience as practitioners, having worked for international organizations, and, in one of our cases, for a different OSCE programme in Ukraine. Insider interviewing, while requiring negotiation and reflexivity, has long been established as an effective interview strategy (Court & Abbas, 2013; Salmons, 2015). Since we were two interviewers, however, it was not necessary that both of us consistently assumed or attempted to adopt insider positionalities to elicit information or steer the interview. Instead, being co-interviewers allowed us to assume insider or outsider roles more purposively and individually, as the situation befit, and even switch roles on occasion in the same interview. At the outset, we tended to emphasize our practitioner and research experience to different degrees. If we felt that our interviewee perceived our positionality as academics as inhibiting information sharing, we brought up relevant practitioner experiences, moving dynamically alongside the insider-outsider spectrum.

While we identified the practitioner-academic distinction and the primary insider-outsider dimension for our study, it is important to note that other dimensions are likely more relevant in other research contexts. Certain researcher characteristics, like gender, are often viewed as fixed. Other attributes, such as cultural or socio-economic background, are certainly more fluid, and perhaps less obvious online. However, what an online co-interview provides is additional possibility for researchers to emphasize, or-de-emphasize, the relevance of these characteristics to varying degrees. For example, they may highlight their personal experiences related to being a woman, a man, or having a non-binary gender identity. The point we would make is that co-interviewers can anticipate such possibilities in advance, and plan their interview approach accordingly, rather than only be reactive, while recognizing that there will still be a need to adapt to the specific context of the interview and responsiveness and attitude of the interviewee (Hampshire et al., 2014; Hathaway et al., 2020, p. 110; Raheim et al., 2016).

Strategy 2: Good Cop, Bad Cop

In this strategy, we explicitly recognized that the power differential between researchers and respondents could lead to a frustrating interview. In line with other researchers’ experiences, we anticipated that interviewees might not fully engage with our questions, adhere to well-rehearsed lines, refuse to accept the premise of our questions, demonstrate limited reflexivity in their responses, or demonstrate a combination of some, or all, of these behaviours. In our study, a common example of this arose when a respondent was asked, if, with the benefit of hindsight, anything in the SMM could have been done better, and the response was that everything was done as well as it could have been. While this in the abstract could well be a legitimate response, it was our impression that at least some elite respondents were reluctant to be self-critical, or critical at all, even constructively, of the organization for which they had worked. This belief was often substantiated when the respondent was prompted further, and modified or made conditional their earlier answer.

Consequently, and particularly where there was a hint of an interesting insight that had yet to be fully developed in a respondent’s reply, we sometimes employed an approach we call good cop, bad cop (see Abbe & Brandon, 2014; Hathaway & Atkinson, 2003). The good cop interviewer would adopt a more sympathetic persona; the bad cop interviewer would be more openly critical of the interviewee’s responses (Berreman, 1962; Hathaway & Atkinson, 2003). Hathaway and Atkinson (2003, p. 164) argued that social science researchers can learn from the techniques of other professionals, such as journalists and the police, and suggested that the strategy is suited to a sequential approach, where the researcher is first more friendly, but becomes more skeptical or critical over the course of the interview. Kvale (2006, p. 487) posits that for the interviewer to provoke conflict, identify contradictions or weaknesses in assertions by respondents, and emphasize divergence, can be productive. We argue that this approach is particularly effective when employed in a co-interview setting, and that the clear sequence of critical escalation identified by Hathaway and Atkinson (2003) becomes less necessary when two interviewers are present. An explicitly challenging approach by a solo researcher might have risks: ultimately, even a respondent who is prepared to engage may withdraw if they perceive critiques to be too strident. Such risks are moderated when the other interviewer is, in comparison, seen to be much less abrasive, or indeed sympathetic to a respondent (Kincaid & Bright, 1957a, p. 307). Using the good cop, bad cop strategy online also mediates the more negative aspects of nonverbal communication that might be perceived if such an interaction was occurring in person, particularly with respect to the bad cop. As with our first strategy, this approach is dynamic, and roles can be switched during an interview. Playing bad cop does not imply any aggressive or abusive behavior, but rather asking more challenging and probing questions, repeating questions more stridently if initial answers were evasive or openly disagreeing with an interviewee’s responses if they appear to be dishonest or disingenuous. Elites tend to be accustomed to these types of challenges (Hathaway & Atkinson, 2003, p. 179; Ntienjom Mbohou & Tomkinson, 2022), facilitating this strategy. Conversely, playing the good cop role does not imply accommodating all views of the respondent or accepting statements at face value. It does, however, require a positive demeanour, which manifests in two ways. First, directly toward the respondent: the good cop interviewer creates a welcoming atmosphere by asking open-ended questions and providing affirmative signals, including through nodding or empathetic language. Second, in interacting with the co-interviewer: the good cop may periodically soften tough questions or elaborate with examples and qualifications, signalling to the respondent that a broad spectrum of responses is welcome. We point to Bechhofer et al.’s (1984, p. 97) experience, noting that a debate between interviewers can also be a productive way of engaging a respondent to react. In our view, the online setting provides the means for the co-construction of such an interview, while still providing safeguards and assurances to the respondent.

Strategy 3: Knowledge Ambiguity to Elicit Information

Since neither of us had worked in the SMM and were not personally known to most of our respondents, our interviewees did not know what we knew about the mission and its evolution over time. We did not want to disclose too much of our own knowledge at the outset in case this risked pre-empting lines of response. We wanted to ensure that respondents had the chance to present their own views of what they deemed important. We noticed that some respondents enjoyed sharing information they believed was new to us, as they wanted to demonstrate their competence or help us understand the conflict. Some, however, would dwell on basic facts and focus on recounting well known descriptive details such as the SMM’s mandate and role, the composition of the OSCE, or superficial information about Ukraine. This limited the time available for us to delve more deeply into their experiences of the SMM, or probe beyond what was largely public information. Yet other respondents appeared reluctant to share more profound insights if they felt the researchers were not competent in the basics.

Again, we adopted a flexible approach to mitigate this challenge. We showed interest without demonstrating too much knowledge early in the interview, and started exhibiting knowledge if the responses were too basic or general. However, with fewer non-verbal cues online, we could not always be sure if our signalling of information was effective in such cases. To elicit information, one way we demonstrated prior knowledge was to explicitly mention or repeat insights about the conflict or the mission, sometimes explaining that other interviewees had raised these points. For example, in asking what happened to information monitors gathered, we might first ask a general question about reporting procedures. If the discussion remained overly general, we might eventually ask how the regional hubs or national headquarters processed reporting information, mentioning the precise names of the relevant organizational units and senior officials. Providing this detail gradually, rather than at the outset, tended to prompt more detailed reflections. Acting as collaborative co-interviewers allowed us to reinforce each other’s points. We would then ask how respondents related to these insights. This gave us the opportunity to steer the conversation and ensure greater depth to responses, while maintaining an open attitude to respondents’ views on a particular issue.

Socratic irony can also be applied to an interview context. As McDowell has pointed out (1998, p. 2138), shifting an interviewer’s perceived competence can be effective: an interviewer might understate or downplay their knowledge with older patriarchal figures, for example, while showing that they are better informed with other respondent populations. What can enable this strategy online is the ability to have a dynamic exchange of information between co-interviewers without alerting the respondent. This we could do through the online messaging system of the interview platform, or via a parallel communication means (usually WhatsApp messages). To preserve the integrity of (and focus on) the interviewer-to-interviewee dynamic, this approach requires coordination between interviewers. Co-interviewers should have a good understanding of each other’s knowledge, and judge when to demonstrate different degrees of competence during the interview, using parallel messaging only sparingly to deal with remaining challenges.

Adapting Strategies and Preconditions

The three strategies are dynamic and not mutually exclusive; they can be employed in combination or consecutively, with the emphasis and application adapted to the interview dynamics. In our case, in some interviews, the flow was natural, while others posed more challenges, necessitating a focused application of a particular strategy or adapting along a spectrum throughout the interview. While there are no definitive triggers for selecting a specific strategy, certain challenges indicate the need for a shift in gears. For instance, seemingly rehearsed responses to questions may call for the ‘good cop, bad cop’ approach. Conversely, overly general, or basic answers may warrant a progression along the knowledge ambiguity spectrum. Effectively implementing the three strategies requires practice, preparation, flexibility during the interview, and debriefing for future learning. Awareness of the strategies can guide researchers through all these steps, enhancing their reflections and deliberate choices. This is particularly important when co-interviewers work online, where decisions often require more deliberate consideration than in in-person settings.

Besides preparation and awareness, there are conditions to effectively employ each of the strategies. Clearly, interviewers need the relevant background to assume a (partial) insider role, as they cannot simply manufacture experience, which might limit the application of this strategy to all qualitative researchers. However, in co-interviews, researchers who lack a specific personal or professional experience can deliberately pair up with researchers who complement their own background, taking advantage of the fact that online settings no longer require a shared locality. Regarding the ‘good cop, bad cop’ strategy, not all researchers may feel comfortable adopting a bad cop persona. Some co-interviewers may be better prepared to assume or take on that role. Alternatively, understanding that this does not involve abusive behaviour, but rather a more challenging and probing attitude that draws a contrast with the co-interviewer, and that the two-cop personae can be along a spectrum rather than fixed and extreme positions, may help researchers make use of this strategy. Regarding knowledge ambiguity, researchers can employ this strategy more effectively if they are well prepared for the content of the interview, and if they have complementary expertise. Researchers may also find that they can employ this strategy more effectively over time, as they accumulate knowledge and understanding of different perspectives once they have conducted multiple interviews on a specific subject.

Tactic 1: In-Interview Signalling

In co-interviews, it is crucial that interactions between interviewers are effective in establishing the foundation for a successful interview. The importance of nonverbal cues in interviews is well established (Hathaway & Atkinson, 2003, p. 180), and extends to relations between interviewers. In a physical setting, through subtle gestures and expressions, interviewers can signal to each other when they want to interrupt, if they want the other to continue their thought, or even more complex messages such as which part of a question may require a follow-up. In the context of online interviews, however, these nonverbal cues are much more difficult to send and interpret, as all participants stare straight into the camera and interviewers cannot address a signal specifically to the co-interviewer.

Occasionally, we used written communication on a different software system, such as an instant messaging platform, to check if the other interviewer had any pressing issue, or to indicate that we should move forward. However, to keep the focus on the interview, we sought to routinize in-interview signalling as a means to manage interviewer interactions. For example, we would toggle the mute function on the videoconferencing software to signal when we wanted to interject, even if the respondent was still speaking. We knew that the other interviewer would use this as a signal to ask the next question and came to look out for this signal between ourselves. We would often ask the other interviewer prompting questions in front of the interviewee, acknowledging that the online setting required more explicit and verbal coordination between researchers, while also showing to the respondent our active collaboration.

Tactic 2: The Tactical Interruption

Like communication between interviewers, nonverbal cues also serve as an important communication tool between interviewers and interviewees. Interviewers can signal to the interviewee that they should continue their thought, or that the response is sufficient. Interviewees can actively look for such signals or send their own signals to check if they are on track. Even in physical settings, this can be difficult, and some interviewers and interviewees are more effective in using nonverbal cues than others. In online settings however, these challenges are accentuated, and there is a risk that an interview might lose focus if the interviewee seeks to set the agenda or dominate and the conversation shifts away from the main focus. While we did sometimes use subtle ways of signalling—such as using the unmute button discussed above—we would also at times interrupt more assertively, by simply interrupting while the respondent was talking. We used this when interviewees decided to ignore more subtle signs, and especially in situations where they discussed something tangential to our research interests in overabundant detail. Here, we drew on the social expectation that subtle interruption in video-based settings is more challenging, as mitigation to interject in a way that may have been seen as impolite or even hostile in an in-person, face to face setting.

Tactic 3: Turn Taking

Finally, and implied in the descriptions above, we used turn taking between co-interviewers as a tactic to steer interviews. While there was no hierarchy between researchers and, while we both collected evidence for all steps of the process tracing, our interests and questions varied, leading to a natural role division. Before each interview, we would decide on a rough order of questions, and who would take the lead for that interview. The selection of questions and order in which these were posed was determined by the function and seniority of the interviewee, as respondents had different responsibilities and different exposures to actors whose activities were of interest to us. Because we used informal Bayesian updating (Beach & Pedersen, 2016, pp. 154–226; Fairfield & Charman, 2017, p. 364) in our evidence collection, we prioritized questions about activities and causality to which we had less information relative to others.

Turn taking did not imply strict alternation between interviewers for every question or equality in the number of questions asked by each person in each interview. One person would often ask a few questions and follow up where necessary, with the other taking a backseat if the interview was seeming to be insightful on a specific issue. However, if either of us felt that the conversation was not moving forward, or that the interviewee stuck to well-rehearsed lines or meandered in their thoughts, we would indicate that we would like to move on to ask the next question (see in-interview signalling). Alternatively, the active questioner would explicitly pass the floor to the other interviewer. This often allowed us to bring the conversation back in focus, and shift from the interview’s impasse. It also provided for more active participation by everyone, compared to a situation where only one person is asking questions, or where the entire sequence of question-asking is predetermined and rigid.