Digital Bystanding: A Mixed-Methods Investigation of Barriers, Enablers, and Decision-Making Processes Among Young Adults

: (Alternative Format Thesis)

Abstract

Online harassment is linked with potentially devastating consequences for individual victims and their families. ‘Digital Bystanders’ are defined as those that: 1) observe a harmful act taking place online (e.g. on social media, the particular focus of this research) and 2) can intervene in some capacity (e.g. reporting, flagging, direct messaging, resharing). Digital Bystanders are uniquely placed to identify (and challenge) harassment and play a role in reducing the potential impact of online harassment (e.g. by providing emotional support to victims). Nevertheless, a consistent finding over the past decade is that bystanders prefer not to intervene. This thesis unpicks the factors that constrain and enable digital bystanders, with the aim of increasing bystander intervention. Existing research has largely applied traditional psychological models (‘Bystander Intervention Model’, Darley & Latané, 1968) developed for traditional ‘emergencies’ (e.g. physical violence, sexual assault) to explain digital bystander behaviour. Given that people can act differently online according to the affordances of platforms, there may be key differences between traditional in-person bystanding and digital bystanding. In this thesis, we explore this fundamental question by reviewing traditional models and research alongside cyber bystanding research (e.g. cyber-bullying). The particular focus of the thesis is young adults (18-24 years) since they have been identified as having a higher risk of experiencing online harassment compared to older adults (Pew Research Centre, 2021).

In study 1 (a qualitative study) we explored the unique barriers and enablers of bystanding from the perspective of young adults using a focus group methodology. Through thematic analysis, five key themes were constructed. From this, a new theoretical framework was proposed (Bystander Decision Response Model, BDRM) bringing together ‘in-person’ and digital models. Using the BDRM as a theoretical foundation, in Study 2, we tested the contribution of self-efficacy and personal responsibility through a novel simulation (using ‘Truman’, a mock social media environment) by exposing participants to instances of online harassment. Using design nudges (e.g. pop ups and community guidance), we found that self-efficacy and personal responsibility alone does not influence bystander intervention. Interestingly, the interaction between both factors (specifically high responsibility and low efficacy) led to an increase in indirect interventions (i.e. reporting, DM’s). Results are discussed in relation to an amplification of personal responsibility when the situation is ambiguous and the formal reporting mechanisms are not clear (i.e. low efficacy). Furthermore, qualitative findings provide several recommendations for practice including targeting bystander motivation to intervene and demonstrating clear benefits of intervention (i.e. through success stories).

Finally, in study 3 and 4, we examined whether key contextual factors informed by the BDRM (Study 3: relationship closeness, Study 4: anonymity) influence the decision-making process. Using TWINE (a ‘choose your own adventure game’), participants were taken through a series of decision junctions as a story unfolds. Study 1 found that a closer relationship increases bystander intervention and impacts the type of intervention across the game. Transitional probability analysis clearly showed that unique decision ‘paths’ are taken when there is a closer relationship with the victim. Interestingly, state-based personality variables (e.g. social risk taking, self-monitoring) were not significant drivers of intervention. Close friends sought perpetrator-focused information, whereas strangers sought social-focused information. Regardless of condition, social information facilitated disengagement whereas seeking perpetrator-focused information facilitated indirect interventions (DM’s, reporting). In Study 2, anonymity had limited impact on intervention and did not lead to distinct decision-making trajectories. Personal experience of harassment and gender impacted intervention behaviour, although this reached statistical significance in Study 2 only. In both studies, higher bystander self-efficacy led to increases in intervention (for all types of intervention; DM’s, private reporting, and public confrontation). Implications are discussed regarding the design of SNS’s and messaging campaigns since bystander self-efficacy can be developed in young people. For example, social modelling (witnessing peers intervene) and mastery (providing opportunities to ‘try out’ different interventions and outcomes) could be easily incorporated into training and education.

Taken together, these findings suggest that digital bystanders are constrained by the social and reputational risks of bystanding. To counter this, bystanders tend to intervene indirectly through a range of micro-interventions. Contextual drivers such as relational closeness are key drivers of bystander intervention and influence both the likelihood and type of intervention. Our findings provide support for an updated and reconceptualised model (BDRM) which incorporates crucial technical factors. This model suggests practical evidence-based recommendations. Namely, we suggest that SNS could equip bystanders with more information on the perpetrator to reduce evaluation apprehension and social cues (i.e. successful peer interventions) to enhance self-efficacy.

Date of Award	24 Jun 2026
Original language	English
Awarding Institution	University of Bath
Supervisor	Adam Joinson (Supervisor) & Catherine Hamilton-Giachritsis (Supervisor)