Scroll on or Step in – The Digital Bystander Dilemma

Previous CSR articles highlight the critical role of bystanders in the detection and prevention of mass violence. That is, friends, family members, acquaintances, and strangers who may notice unusual or concerning behaviours in the lead up to an act of violence. Despite this, many bystanders do not intervene or report concerns (i.e., become ‘upstanders’) due to a range of contextual, personal, and social barriers. A common barrier is the sense that their concerns will not be taken seriously, or ambiguity around where and how to report concerns. 

Do the same barriers exist for Digital Bystanders? And perhaps most importantly, can digital environments be designed to incentivise bystander intervention? 

This article extends this perspective and considers the motivations and enablers of bystanders within digital environments. Do the same barriers exist for Digital Bystanders? And perhaps most importantly, can digital environments be designed to incentivise bystander intervention? My research examines these fundamental questions within the context of online harassment amongst young adults within the UK. This article considers the possible implications and next steps from a defence, security, and policing perspective. We unpick the reasons why bystanders do (and do not) intervene within online settings and provide some practical recommendations. 

What is a Digital Bystander?

Digital bystanders are those that witness harmful or potentially concerning behaviour in digital settings with the ability to intervene in some capacity. They may be close friends, family members, online contacts, or strangers. Digital bystander intervention can take many forms – from reporting through official channels through to DM-ing users. 

Our qualitative research identified a range of subtle, indirect, and often creative intervention strategies used by young adults. For example, the use of screenshot as evidence, ‘like-bombing’ victim accounts and crowdsourcing reporting. 

Traditionally, bystander intervention is conceptualised in terms of whether someone physically confronts a perpetrator or passively ‘stands by’ (i.e., action versus inaction). We propose that this binary notion of bystander intervention is outdated. Digital bystanding exists along a spectrum. That is, depending on the platform, digital bystanders use a range of micro-interventions or behavioural strategies to intervene in harmful digital behaviour. 

Intervention options are largely afforded by the specific digital environment that the bystander is in (e.g., Reddit’s ‘downlinking’ feature is a visible yet low effort ‘micro intervention’). Technological developments have facilitated new forms of bystander intervention varying in visibility (i.e., one to one encrypted communication versus one-to-many public postings on a comments page), effort (i.e., clicking a ‘red flag’ versus complex reporting forms) and anonymity (i.e., own name account versus sock puppet account). Critically, in online settings abuse/harm can be reduced before it occurs (e.g., through regulation), at the point of posting, or after the event (e.g., through reporting or removal).

Digital Bystanders and National Security

In defence and security contexts, digital bystanders can identify and report online behaviours of concern, leading to early detection. For example, from an insider threat perspective, colleagues can be ideally placed to identify concerning behaviour of their co-workers both offline (i.e., in the office) and online (i.e., social media, corporate systems). Furthermore, the role of the general public in detecting terrorism and crime is increasingly recognised and integrated into counter-terrorism campaigns (e.g., ‘See it, Say it, Sorted'). 

Bystanders may have pre-existing relationships with the perpetrator, putting them in an ideal position to witness pre-incident indicators (i.e., online behavioural changes).
Rowe’s (2020) research has consistently highlighted the role of “intimates”; close family and friends with key insight into the intent and motivations of potential attackers. 

As noted by Gill et al. (2014), ‘lone actor’ terrorists often share intentions and grievances in the lead up to attack planning. In an increasingly digital world, there is arguably more opportunity for bystanders - strangers, online associates, through to close family members - to detect behaviour of concern within online environments. Somewhat surprisingly, comparatively little research has been dedicated to the topic of digital bystanding. Even less research has been conducted on whether (and how) online environments can be shaped to increase bystander intervention. 

Digital bystanding exists along a spectrum. That is, depending on the platform, digital bystanders use a range of micro-interventions or behavioural strategies to intervene in harmful digital behaviour.

Rewinding to traditional bystander research

To understand digital bystanding, we revisit traditional bystander research. Interest in bystanders exploded following the tragic and brutal murder of Kitty Genovese outside of her apartment block in New York in 1964. At the time the world was horrified to learn of the apparent inaction of 37 witnesses during three separate attacks in public view (hereafter referred to as ‘bystanders’).  Media reports claimed the bystanders heard or even saw the attack from nearby apartments but turned a blind eye. This led the public outrage at the time; how could they simply stand by during such a brutal attack? What does this say about human nature? 

The details of the case and the so called ‘apathy’ of the 37 bystanders has been challenged and refuted in recent years (Manning et al, 2007). In reality, the majority of bystanders did not realise an emergency was taking place. Many interpreted the event as a domestic argument rather than a fatal attack. In fact, bystanders did indeed investigate the scene (therefore risking their own physical safety) and came to the victim’s aid. Among them, a 70-year-old woman who stayed with the victim until the emergency services arrived, and many others reported the incident to authorities. 

Nevertheless, the case triggered an urgent interest in what social psychologists refer to as the bystander effect (Darley & Latane, 1968). For a long time, this research was dominated by the idea that the more bystanders are present during an ‘emergency’, the less likely that anyone will step in. In reality, research has established a range of reasons why people choose not to step in; from legitimate concerns about physical safety, fear that they may misjudge the situation (evaluation apprehension), and often a lack of skills and confidence (self-efficacy). Researchers set about trying to answer the question of what motivates (or more importantly perhaps prevents) people from helping in emergency situations. There is now a large body of research into bystander intervention in a range of contexts (i.e., sexual violence, physical emergencies, harassment, and stalking). The topic of bystander intervention within digital settings is relatively underexplored in comparison. 

Digital Bystanding – same, same or different?

Our research into digital bystanders has revealed some fundamental differences that can shape the experiences (and ultimately the behaviours of) bystanders in online settings:

1. Diffusion of responsibility between users and platforms: In traditional bystanding incidents, such as a bar fight for example, bystanders often feel that another person should (or will) step in. This contributes to what is referred to as the diffusion of responsibility (whereby nobody steps in). With Digital Bystanding, a similar process appears to occur except for one fundamental difference. Users feel that platforms themselves will step in. Bystanders often outsource this responsibility onto platforms (or AI), leading to a similar ‘diffusion of responsibility’ effect. 
2. Ambiguity (amplified in online settings): There is arguably increased ambiguity online; leading bystanders to question whether an intervention is necessary - or indeed, whether they even notice in the first place. Taking social media posts for example, older posts introduce uncertainty in terms of whether violent and / or concerning material has been ‘dealt with’ (by the platform or other bystanders). It is also not always clear whether postings are an inside joke or ironic post rather than a credible threat to harm. Getting the timing ‘just right’ can be a challenge for digital bystanders. Furthermore, bystanders often lack the critical information they need (e.g., body language, tone of voice) to help interpret it as an incident requiring action. Taken together, these factors can amplify evaluation apprehension within online settings. 
3. Social risk amplification: Young adults (defined in our research as 18-25) are extremely mindful of social and reputational risks. They have, after all, grown up in a digital world where their digital behaviours directly impact upon their offline behaviours (i.e., from job prospects through to their status within social circles). Our research found that young adults therefore carefully balance the pros and cons of intervening online. Key barriers included the fear of becoming a victim themselves or making things worse for the victim by further ‘fuelling the fire’. Further concerns related to being socially outcast or scrutinised by their peers. However, it wasn’t all negative. The relative anonymity afforded by some online environments facilitated intervention by reducing some of the psychological and social barriers. Furthermore, our research illuminated several intervention methods that are not available in-person. 

Key barriers included the fear of becoming a victim themselves or making things worse for the victim by further ‘fuelling the fire’.

Bystanding by design

Our research suggested the following practical recommendations to lower barriers and increase intervention: 

1. Bystanding by design: Firstly, reporting and intervention needs to ‘designed into’ platforms from the offset rather than an afterthought. In our studies, simply increasing the range of intervention options available on the platform facilitated bystander intervention. Through simulation studies (e.g., mock social media), we’ve tested key design features. Subtle and straightforward changes can be implemented to remind, normalise, and encourage bystander intervention (e.g., pop-ups, advertising, community rules). 
2. Minimise (or reframe) social risks: For some users, digital settings can reduce the barriers experienced in traditional bystander scenarios such as a fear of physical retaliation. However, digital settings can also amplify social and reputational risks due to the high visibility and relative ‘permanence’ of actions online. Providing a range of intervention options varying in visibility and effort is therefore key, along with normalising and celebrating bystander intervention where successful. 
3. Reduce ambiguity through decision-making support: In our final study, we examined decision-making in detail using a novel experimental platform. We found that the type of information that bystanders seek influences and guides subsequent actions (i.e., deciding when and how to report). For example, seeking more information on the perpetrator led to more active intervention. With the rise of AI assistants, there is an opportunity to consider how human users and AI can work together with the shared aim of identifying, reporting, and intervening in online behaviours of concern. 
4. Transparency and feedback: Our research found that reporting via social media often feels ‘empty’ to young adults. Participants lacked trust in the reporting process overall and asked for greater transparency. They called for clear and concrete feedback on how their reporting would be dealt with. Even better would be more awareness of how reporting and intervention helped. 

Finally, and perhaps most importantly, bystanders learn by doing. Our research consistently highlighted the role of self-efficacy (developing the skills, confidence, and abilities to intervene). Young adults may be particularly encouraged by witnessing their peers intervene (i.e., ‘social modelling’) and through role-play or immersive games to develop skills and assertiveness (i.e., ‘mastery’). Seeing peers intervene also increases a sense that bystander intervention is normative, that is, it’s valued by the online community that they are part of. There are many promising educational initiatives in relation to digital literacy, assertiveness, and empathy to increase digital bystanding. However, these initiatives need to be grounded in an understanding of online behaviour specifically. We cannot simply ‘lift’ traditional models and apply them without first testing them in digital settings. 

In short, whilst there are many parallels between traditional and digital bystanding, there are fundamental differences to acknowledge and ultimately ‘design into’ digital settings. Our research revealed just how important it is to engage with young adults themselves in designing these features.