Ethics and the dark side of online communities: mapping the field and a research agenda

The first systematic literature review stage required the drafting of the protocol (Fig. 1) for the following stages. The Review Protocol starts by searching the literature for the definition of the terms applicable to determining sets of data chosen to select the maximum number of studies able to respond to the research questions formulated.

The source of the relevant articles complied for this study was the online Web of Science™ Core Collection (WoS) database, the property of the company Clarivate Analytics. The data for this study were collected and compiled from the Web of Science (WoS) online database. Hence, WoS provides a comprehensive view of worldwide research production (Ferreira et al. 2021). In the subsequent phase, we defined the criteria for inclusion and exclusion (Table 1).

In order to identify the relevant articles, the search applied the following expressions whether included in the title, the abstract or the keywords:

The search for these key terms in the database selected by the review protocol resulted in 175 documents. We refined this data set to include only English language articles and obtained a set of 131. Reading and content analysis led to a set of 53 relevant studies for our research. The research took place in April 2022. Figure 2 illustrates the study selection process, based on our inclusion and exclusion criteria.

Table 2 presents a descriptive summary of the articles included in the present study. The articles result from the work of 103 authors (107 author participations), and 18 were sole authored.

The literature review applied the bibliometric methodology (e.g., Zupic and Čater 2014), thus, defining the flows of research work into a scientific mapping by bibliometric methods. Furthermore, bibliometric analyses serve to evaluate the state of research on a global scale (Mutschke et al. 2011), consisting of the application of quantitative and statistical analysis to publications such as articles and their respective citations, deployed as a means of evaluating the research performance in keeping with how the approach provides data on all the activities ongoing in a scientific field and summarising this data offers a broad perspective on the research activities and their impacts (Hawkins 1977; Osareh 1996; Thomson Reuters 2008).

Despite co-citation analysis having held a predominant position in bibliometric analysis, there has recently been the resurgence of bibliographic coupling to the detriment of this historical preference for co-citation analysis (Boyack and Klavans 2010), given how this allows for the inclusion of more recent publications to evaluate potential patterns and trends among the publications (Leydesdorff and Vaughan 2006). To understand this field in greater depth, the authors carried out the qualitative codification and thematic analysis of the sample established (n = 53). Various authors defend that, depending on the level of detail required, the number of publications needed to study the trends and evolution in a field of study may be in the region of 50–100 articles (Walsh and Renaud 2017; Walsh and Kalika 2018).

To graphically represent the bibliographic coupling analysis, the authors made recourse to network theory and determined the clusters that took place according to the methodology applied by Waltman et al. (2010). Waltman et al. (2010) developed a modularity-based clustering with a resolution parameter where small clusters can always be identified by choosing a sufficiently large value for the resolution parameter. This clustering technique is based on the most recent developments in bibliometrics and has been used and validated by other researchers in the field of bibliometrics (van Eck and Waltman 2017). We carried out all calculations through Microsoft Excel 2010 (Microsoft Corporation, Washington, USA), VOSviewer version 1.6.14 (van Eck and Waltman 2010), and Bibliometrix (Aria and Cuccurullo 2017) software programs.

Figure 3 summarises the trend in the number of articles and the citations referring to research on questions relating to the ethics associated with online communities. The average year of publication stood at 2015.7.

The research interest in the dark side of online communities essentially began in the last decade, despite the publication of the first article dating back to 2004 and with over 50% of the articles published since 2016 (no.=32, 60.4%). In terms of the number of citations, the articles incorporated into this study received an average of 67.2 citations to form an overall total of 3,563 citations. Table 3 presents the ten articles with the largest number of citations.

The articles with the largest number of citations in this set of articles come from Boyd and Crawford (2012) (2,363 citations), Posey et al. (2010) (223 citations), Hajli and Lin (2016) (124 citations), Flicker et al. (2004) (82 citations) and Barrett et al. (2016) (80 citations).

To structure the research into the ethics and dark side of online communities, we also organised the articles studied based on the countries of affiliation hosting studies of the different topics (Table 4). The United States (47 articles), the United Kingdom (28 articles) and China (18 articles) are the most active countries in terms of publications in this field.

Based on the bibliographic coupling matrix, we determined the shared reference matrix for these 53 publications to establish the networks of connection between the articles and the clusters. The clusters took place according to the methodology applied by Waltman et al. (2010). The clustering technique that we used determines the relatedness of publications based on direct citation relationships.

Figure 4 displays the article networks based on the shared reference matrix data and the publications’ respective groupings ascertained by cluster analysis (Table 5).

Flicker et al. (2004) identify how the growth in utilising the data available in online communities has brought ethical problems, especially in studies approaching Internet-based discussion forums. Therefore, directives and guidelines are needed to guarantee ethical conduct in this type of research. These authors state that three ethical problems may emerge: registering research participants, protecting participants from risk or harm, and interlinking public and private data sources. Bowker and Tuffin (2004) also defend that online data, constantly available, inevitably raises the need to reflect on ethical questions. Taking up this concern, Roberts (2015) identifies ethical questions associated with research into online communities as follows: (1) public/private spaces; (2) authorship versus research participants as human beings; (3) informed consent; (4) anonymity and pseudonyms; (5) confidential research; (6) false identities; and (7) the quality of the data obtained. Hence, before advancing with qualitative research in online communities, researchers hold the ethical responsibility to identify and evaluate possible risks and benefits for the community and its members. These ethical issues also reflect in the retrospective study by Paechter (2013) of an online community that provides support during divorce cases. In turn, Schultze and Mason (2012) conclude that an increasing number of Internet users are simply bots. Hence, the need to review the ethical directives in effect for research on the Internet.

According to Trevisan and Reilly (2014), Internet research ethics should remain permanently informed about the disciplinary perspectives of those studying online communities. Therefore, Internet researchers should ground their practices in the already established ethics in their respective subject fields to approach these questions “centred on the human being”, and these questions become still more predominant when related to healthcare issues. Gustafson and Woodworth (2014) recommend that principles of concern deriving from well-being, respect for persons and justice always require application in research projects studying the data of online communities.

According to Brady et al. (2016), online privacy represents a nebulous concept and the distinction between what individuals are comfortable with and what relates to subjective factors. This applies to such an extent that Germain et al. (2018) provides evidence of the importance of reflecting on the perceptions of participants about just what is public, preserving anonymity and protecting the participants from harm. Paris et al. (2013) also convey the relevance of these ethical questions to researching online communities, especially when the study takes place through government institutions. Guaranteeing the privacy of community members while also complying with the legal responsibilities set by the government is, above all, an act of balancing.

Kantanen and Manninen (2016) identify some fundamental questions regarding the ethics of online community research: publicness versus the privacy of the community; the definition of research into human beings; recruitment of participants; informed consent; and ethical questions associated with observing virtual communities and reporting and disseminating the research results.

This thus conveys how, despite the growing utilisation of data from online communities, there is no standard set of ethics for carrying out ethnographic studies of the platforms used by these communities (Hair and Clark 2007; Wheeler 2017). The study of the participants on these platforms prior to data collection and study is thus fundamental to ensure a higher quality of the data subsequently collected. In this sense, Stvilia et al. (2019) maintain that the fundamental question for ethics is guaranteeing the quality of information gathered in online communities. Hence, this requires deploying information and research management systems, which should involve the researchers within the scope of guaranteeing the quality of information. Within this line of thinking, Liu  (2021) identify artificial intelligence as a solution for mitigating the ethical problems related to studying the data of online communities. Yeku (2020) raises another ethical research question related to online communities’ fraud and deceptions.

Given the growth of the Internet, it is fundamental to educate people on the difficulties in determining right and wrong, which may be offset by the new technologies between online and offline environments (Griffiths and Light 2008; O’Neil, 2015). Additionally, individuals may grasp this understanding of “right and wrong” through the benefits they accrue from participating in these online communities, which adopt their specific policies guaranteeing their ethical conduct or, on the contrary, the negative returns received for their participation in communities that do not enact such policies (Gummerus et al. 2017; Jones and Giles 2022). Steinmetz (2012) defends that researchers need to adopt complete ethnographic methodologies to effectively observe these phenomena in keeping with the growth of online communities. Therefore, this author supports that three specific facets of virtual ethnography require consideration: space and time, identity and authenticity, and ethics. The Internet is a globalised and instantaneous means in which space and time collapse, identity becomes more playful, and ethics become more tenuous; incorporating these aspects is crucial to studying online communities. Kafai et al. (2019) warn of the importance of “the normalisation” of terms such as deceit and fraud introduced by individuals with an absence of ethical characteristics in online communities and that young persons with unstable moral values also tend to normalise. This, thus, refers to the cyber influence of destructive ethics (Burnett and Burnett 2019; Baeva 2020; Cheng and Yu 2022a; Cheng and Yu 2022b). “Death groups” that call on young persons to commit suicide, as well as the “aggressive communities” associated with acts of aggression and violence in institutions, now extend to cyberbullying (Breitsohl et al. 2022).

Boyd and Crawford (2012) verify how various types of academic specialists (computation scientists, physicians, economists, mathematicians, political scientists, bio-information specialists, and sociologists) increasingly request access to the enormous quantities of information produced by and about people, things and interactions. These authors arrive at the following questions: will these large scales of data help develop better tools, services and public goods? Or will it trigger a new wave of incursions into privacy and invasive marketing? Will this analysis return a better understanding of online communities and political movements? Or will this simply serve to identify demonstrators and annul their discourses? These are questions posed by Big Data to the research community and society. Within this scope is the increasingly prominent practice of “sharenting” (parents and children sharing their online representations). On this, Blum-Ross and Liningstone (2017) study how parents define the boundaries of their digital “selves” and how they justify the “story they have to tell”. The authors conclude that these bloggers deal with profound ethical dilemmas. In representing their identities as parents, they inevitably make public aspects of their children’s lives and thus introduce risks against which they are, paradoxically, responsible for protecting. Another question about data sharing stems precisely from the facility where individuals may share personal data on social networks as their structures favour and facilitate such sharing (Watanabe 2007; Rawassizadeh 2012; Kafai et al. 2019). Rawassizadeh (2012) defends a fundamental need to limit the availability of the personal data of users of this type of online community.

Regarding the creation of online identities, Caiani and Parenti (2009) studied the utilisation of the Internet among extreme right-wing organisations in Italy to create a collective identity. They then conclude that these organisations use the Internet to boost the number of extremist websites and disseminate their propaganda and fundraising for political campaigning. These considerations receive support from Arfini et al. (2019), who confirm the recourse to online communities by extremist ideology groups that, during crises, foster ethically dubious consequences.

Hajli and Lin (2016) maintain a fundamental need to study the safety of social networking sites. Social network sites have challenged ethical questions on the security and privacy of user information. This situation has led to user concerns over their privacy and their control over the information they share. These authors conclude that perceived control over sharing information among users is a significant factor. Another issue related to safety stems from the piracy of software and the security questions that such actions may cause to their users (Yu et al. 2015). Ouakouak and Ouedraogo (2019) approach the influence of commitment and organisational trust in sharing and using knowledge before concluding that both affective dedication and professional trust generate positive influences on sharing and utilising knowledge, while personal trust and commitment to continuity do not. These authors also report that ethical business practices moderate the relationship between sharing and using knowledge.

Regarding safety, the ongoing pandemic since 2020 has brought new challenges for online communities (Tseng et al. 2022). Tseng et al. (2022) demonstrate the importance of the credibility of information and the perceived privacy risk in social networks to consumption by their users. Therefore, since users culturally differ from each other, Posey et al. (2010) identify the extreme importance of approaching online communities from transcultural perspectives. The authors apply three theories in this approach: ideas of social exchange (individuals get involved in relationships when the perceived costs associated with the relationships fall below the expected benefits) and penetration theory (individuals participate out of self-revelation to foster relationships) as well as transcultural theory related with individualism and collectivism.

After researching the factors that shape the aggressive intentions among users of online communities, Xu et al. (2016) conclude that online aggression may be inhibited by moral beliefs and susceptible to strengthening through means of implementing consequences through effective community policies and mutual pressures applied by community members.

The growing dependence of society on information technologies comes coupled with a hike in the number of increasingly sophisticated cyber-attacks committed by criminals operating on the Darknet (Gatzweiler et al. 2017). Gatzweiler et al. (2017) put forward the Darknet Identification, Collection, Evaluation, with Ethics (DICE-E) tool to enable researchers to study processes ongoing on the Darknet.

In keeping with the constant growth in social network popularity, a new ethical debate emerges on marketing and digital technology (Vanacker and Heider 2012; Barrett et al. 2016; Kaylor 2019; Liao et al. 2020a; 2020b). Despite the ever-more intensive utilisation of online communities, consumers hold more significant concerns over the credibility of the information provided. Various researchers have reported that the greater the quality of the information provided in online communities, the more influential these become in formulating opinions and shaping consumer decision-making (Barrett et al. 2016; Tsao and Mau 2019; Banerjee et al. 2021).

If opening up company decisions to the opinions of consumers returns positive effects, it also has limitations as companies end up ceding control over decisions to an unknown crowd (Hajli 2018), providing the scope for that termed deviant co-creation. Hajli (2018) identifies how deviant co-creation spans two types of content: (1) illegitimate content (including five themes: humour, provocation, uniqueness, and violation of technical and social norms) and (2) deviant content (generally confusing to its consumers and may trigger visible and malicious protests or result in exchanges and ridicule in the online community, exposing the individual to reputation risks).

The sharing of “fake news” or disinformation through online communities has risen drastically recently, especially during the current COVID-19 pandemic (Benjamin et al. 2019; Stockinger 2015). Stockinger (2015) reports that in sharing news without verifying, the veracity depends on the psychological profile in which greater impulsiveness is associated with immediate sharing without any confirmation as to the source/accuracy of the contents (Fig. 5).

There have been various research projects into the general ethics of online communities as well as how to carry out research in these communities (Flicker et al. 2004; Bowker and Tuffin 2004; Roberts 2015) and on the quality of the information collected (Wheeler 2017; Stvilia et al. 2019; Yeku 2020; Liu 2021).

Irrespective, the research on the Internet retains the need for retrospective studies of particular relevance in understanding how online communities form, develop and obtain success.

Other important questions for future research stem from the issues around the protections given to humans and the Internet bots participating in online communities. The protection of the identity and data of every type of individual is fundamental. Therefore, establishing limitations on obtaining data from studies of online communities amounts to a crucial ethical question. This also needs further studies regarding government institutions’ involvement in online communities, especially when seeking to involve citizens in this type of community.

Regarding the deployment of artificial intelligence in online communities, further studies require applying qualitative analytical approaches and mixed methodologies and demonstrating how, through their utilisation, we may study online communities while simultaneously protecting their users’ data.

A highly diverse group of researchers has dedicated their attention towards the education and information held by users of online communities (Griffiths and Light 2008; Gummerus et al. 2017; Jones and Giles 2022). Furthermore, other academics have focused on the question of creating online identities and normalising anti-ethical behaviours in specific communities (Steinmetz 2012; Kafai et al. 2019; Burnett and Burnett 2019; Baeva et al. 2020).

Engaging in research that questions the suppositions, values and biases of this new wave of research bound up with Big Data is vital. It is equally important to research extremist groups’ recourse to online communities. The Internet provides a very relevant tool for these groups and the construction of (their) collective identity.

Virtual ethnography represents a methodology for consideration by future studies since it provides a powerful tool for understanding social communities and their subcultures. Researchers need to develop methods for appropriately examining these virtual interactions. Future research projects should also consider advancing mechanisms to educate users, in the best possible way, about the difficulties of determining right from wrong and how right and wrong may be seen concerning events in online communities. This is just as relevant for the owners/operators of these environments as the individuals who use them.

New studies should also verify the differences in the effectiveness of online communities across the ethical domains, especially as the ethical standards, ideologies and possession of consumer information define and shapes consumption patterns. There is also the requirement for research on which community interactions (behaviours that increase or decrease uncertainty) are the most beneficial to their members. Finally, evaluating the differences between members of different communities also accounts for an important target for future research projects.

Various researchers have dedicated their attention to studying safety in online communities (Hajli and Lin 2016), such as software piracy (Yu et al. 2015), commitment and organisational trust (Ouakouak and Ouedraogo 2019), the credibility of information (Tseng 2022), online aggression (Xu et al. 2016) and the relationship between cultural differences and the importance of safety in online communities (Posey et al. 2010).

Future research needs to make more detailed studies of the safety information prevailing in online communities and the cultural differences among community users. Future research should also look at the level of diversity in online communities and how the perceptions of which organisational characteristics enable a greater sharing of information in safety and can generate positive results for this organisation. In addition, community policies may or may not directly influence their users. Thus, it is important to verify which mechanisms make the ethical guidelines of online communities more efficient and thereby reduce the incidence of aggressive activities and behaviours, especially among younger users. There is also a great need for transcultural studies to ascertain the impact of culture on the context of online aggression.

Various researchers have concentrated their efforts on the contents of online communities, including Darknet (Gatzweiler et al. 2017), the ethics of marketing, and digital technology (Barrett et al. 2016; Kaylor 2019); the co-creation of value through analysis of the content of online communities (Barrett et al. 2016; Tsao and Mau 2019; Banerjee et al. 2021); unlawful content (Hajli 2018) and fake news (Benjamin et al. 2019; Stockinger 2015).

Future research lines need to verify online communities’ impact not only on already existing products or services but also on the level of products still under development. Hence, this may produce a support basis enabling future innovations and improving existing products and those still for launching, providing a departure point for open innovation. Future research also has to approach the level of deviant content by varied values, norms and behaviours. There is also importance in studying the normative perceptions of online community users. Evaluating this deviant content’s impact on organisations’ innovation processes represents another area for the literature to deepen.