Given that social networks research is fundamentally about relationships between entities, not (solely) about those entities themselves, the first principle of informed consent is inherently problematic. As noted in Chapter 2, it is difficult to conceptualize whether we should consider network data as “belonging” to both/either/neither of the nodes involved in a relationship. It’s similarly not immediately clear how permissions and protocols that have been developed with a focus on individual autonomy should be translated into relational contexts (Schönhuth 2017).
Informed consent is typically interpreted to require researchers to develop procedures that ensure voluntary, informed participation by anyone included in human subjects research. Depending on the level of potential risks to human subjects within any given study, IRBs have developed considerably different standards and procedures—e.g., the substantive focus of a project can lead to the difference between requiring passive vs. active consent from participants. Typically, informed consent procedures assume that consent is provided explicitly.76 In the case of network data collection, the question of whether alters should be treated as human subjects is not immediately apparent.
Not surprisingly, a variety of opinions have arisen in answer to this question. It is relatively straightforward to see how some may contend that any information provided about an individual—even, or especially, if coming from third parties—is data about those individuals, and therefore qualifies them as human subjects (Botkin 2001). The strictest application of informed consent would require explicit consent from each named alter before gathering any information about them.77 This standard is more easily adapted for complete network research designs than other sampling strategies. If we know everyone in the population to be studied, we can administer informed consent before gathering any potentially personally identifying information, and simply exclude any potential subjects who refuse to participate (Borgatti and Molina 2003).78 This has variously been deployed through active or passive consent protocols (Bearman, Jones, and Udry 1997). Active protocols require positive consent for participation (i.e., applies an “opt-in” standard) to be included among a roster. Passive protocols inform everyone in the population about the design, and allows anyone to opt-out of inclusion—both as a participant, and for inclusion of their information as provided by others.
For ego or partial network designs, an active informed consent standard would create a sort of catch-22 situation, in which we cannot gather information about anyone from whom we don’t already have consent, but we do not know to request consent from those individuals until information has been provided about their relationships. Some researchers have successfully argued that active consent is an undue burden on researchers when studies involve minimal risk, leading IRBs to allow passive consent—that is notifying the entire population of the aims and potential risks of the study, then only excluding information from being collected on those who opt out (Lorant et al. 2015). Additionally, a common basis for requesting exemption from certain elements of the Common Rule arises when research could not be conducted any other way, and as long as sufficient protections in all other domains are sufficient. This is particularly applicable to some social network research designs.
Alden Klovdahl helped to pioneer important strategies for link-tracing approaches in research on infectious disease outbreaks, particularly HIV [e.g., Klovdahl:1985aa]. He pointed out that some argue that reports on secondary subjects are not always perceived as information about human subjects.79 A primary reason for this contention is that the data provided are the primary respondent’s perceptions of secondary subject’s information, and not necessarily their actual data (Klovdahl 2005). Moreover, questions arise about how readily–and accurately–identifiable secondary subjects are from such information.80 Klovdahl further provided an outline of procedures for IRBs to follow in adapting strategies that conform to the aims of informed consent and minimized risk, in the context of potentially highly sensitive information surrounding the behaviors and status pertinent to HIV research (Klovdahl 2005).
Social network researchers have also incorporated several other strategies of consent, including the notion of indirect consent.81 Essentially, indirect consent is consent that is provided though a third party. Hogan (2017a) illustrates this with a prominent example stemming from the “Suggested Friends” algorithm on Facebook, whereby in becoming friends with someone on the platform, you are permitting Facebook to reveal that friendship to third parties (whether those who have access to Facebook’s API, or other potential friends). This sort of indirect consent is actually the foundation of a number of common social scientific research strategies. Ethnographers who focus on social interactions in publicly observable settings explicitly focus on gathering information about multiple parties simultaneously, and rarely are expected to administer individual informed consent protocols with each of those subjects (Sönmez et al. 2016), even when those interactions involve members outside the focal population for whom general consent may exist. This standard arises from considering what a person can reasonably assume regarding “expectations of privacy,” where if others are able to observe those interactions, there is no reasonable assumption that the interactions are private. Researchers have further adapted application of this principle to develop standards for harvesting digital trace data (Diesner and Chin 2015).
In a similar way, the household roster has been a vital component of demographic research for several decades (Sweet, Bumpass, and Call 1988). This method asks a single respondent to name and provide a variety of characteristics about each member of their household, and it generally requires participation by only one household member on behalf of each of the others (Kohler et al. 2015). In other words, while social network scholars have occasionally encountered difficulties in convincing IRBs that gathering information on secondary subjects is appropriate, there are ample resources from across the social sciences available to draw on in making that case (Hogan 2017a).