Virtual Reality as a Window into Crime

Despite several decades of research, the exact triggers for bystander intervention in violent incidents remain elusive. We test the hypothesis that the decision of bystanders to intervene in a violent incident depends on their belief about whether the actors involved know each other. We predict that when bystanders assume that the actors in a violent incident have a pre-existing social relationship, they are less likely to intervene compared to when they are believed to be strangers.

In this preregistered study, we test this hypothesis in an immersive 360° video experiment. Participants witness a situation of sexual harassment in a barroom setting and are asked to indicate the likelihood that they would intervene. We compare conditions that vary whether the involved parties are suggested to know each other prior to the incident. We also explore additional motivations to intervene or abstain, including perceived danger, anger, acceptability of violence, and demographics.

The study mimics a real-life event as closely as possible by using a professionally developed immersive 360° video recording and the barroom where the scenario was recorded as the testing location. This “bar-in-a-bar” setting lets participants experience the scenario from the exact same location in which it was originally recorded, maximizing ecological validity.

How do bystanders decide whether—and how—to intervene when witnessing crime in progress? We propose and test a trait–state model of reactive guardianship in which stable personality traits influence bystander intervention decisions through situationally activated affective and cognitive states. Drawing on the HEXACO model, we conceptualize traits—particularly Honesty-Humility and Emotionality—as predispositions that shape momentary moral appraisals and risk perceptions, which in turn predict intervention pathways.

Using an immersive 360° video virtual reality scenario, 247 male participants experienced a bystander situation from a first-person perspective and reported their intended responses. We examined violent and non-violent intervention, such as seeking help. Structural equation modeling indicated that violent intervention was shaped primarily by inhibitory mechanisms. Moral appraisals—perceived unacceptability of violence and anticipated guilt or shame—were central pathways linking Honesty-Humility to reduced violent intervention. Perceived risk, linked to Emotionality, reduced violent intervention but redirected responses toward indirect intervention.

Together, these states explained 26.2% of the variance in violent intervention. The research extends trait–state models beyond offending to guardianship and offers a process-oriented account of bystander decision-making in which behavior is shaped by situationally activated psychological states. It further demonstrates that guardianship is multidimensional, unfolding through distinct intervention pathways.

Introduction. We assess the interplay of perceived disorder and HEXACO personality traits in the emergence of fear of crime through two studies conducted in both real life and a novel virtual reality environment.

Methods and design. We first conducted an observational study embedded in an experimental design. Data were collected in a real-life alleyway at night and in a virtual copy of that environment. We then conducted a 2 × 2 randomized control trial, manipulating the presence of disorder and the quality of the virtual environment. Ordinary least squares regression models assessed the interaction of perceived disorder and HEXACO Emotionality and benchmarked it against other individual-level variables.

Results and discussion. Preliminary results indicate that the established association between perceived disorder and fear of crime replicates, but that the effect is reduced when HEXACO Emotionality is considered. Further individual differences, including gender and prior victimization, improve model fit. The findings combine established mechanisms underlying fear-of-crime experiences that have so far been analyzed separately, while pushing the boundaries of virtual reality in fear-of-crime research.

Fear of crime shapes how individuals evaluate and navigate urban spaces, yet the nature of and rationale behind these behavioural adaptations remain underexplored in realistic scenarios. This study is the first in a set investigating fear of crime in real and virtual urban environments. Within a larger ongoing data collection, it uses a mixed-methods approach to examine how individuals experience and make sense of fear and how they might adjust their behaviour because of it.

Participants experienced and evaluated an urban environment containing signals of disorder. Their trait fear of crime, situational fear of crime, and behavioural intentions were assessed alongside demographic variables. They also completed a retrospective think-out-loud protocol, recalling their thoughts and experiences in the disordered environment, before being asked specifically about behavioural adaptations due to fear of crime.

Analyses identify themes in the qualitative data, triangulate them with quantitative self-report data, and explore where the methods and results diverge or complement each other. Preliminary results reveal spontaneous observations consistent with criminological theories such as prospect-refuge and CPTED. Reported behavioural adaptations vary strongly between individuals, with gender as a clear predictor. The range of verbally reported behaviours also differs from standard behavioural items used to measure fear of crime, suggesting a potential mismatch between research practices and lived experience. The study aims to formalize a mixed-methods approach and generate hypotheses for a follow-up experiment.

Situational theories posit that environmental features shape offender decision-making by altering perceived risks and rewards. Yet traditional self-report data often suffer from post-hoc rationalization. This study tests explicit survey data against implicit eye-tracking metrics during a VR burglary task. Incarcerated men with varying burglary experience explored virtual neighborhoods containing manipulated cues, including vehicles and “Second Amendment” signs. Divergence analysis identified participants with high visual fixation but low reported conscious impact, while correlations between time to first fixation and total dwell time mapped how criminal proficiency modifies visual-search schemas.

Data revealed a systematic disconnect between visual attention and conscious judgment. High implicit-explicit divergence occurred for the gun sign (32.5%) and car (47.5%): participants focused heavily on these cues but rated them as low risk in surveys. The car functioned as a universal “attentional anchor,” with faster detection predicting longer processing. The gun sign produced a “delayed scrutiny” effect exclusively in proficient offenders: late detection prompted prolonged evaluation, a structured strategy absent among novices.

The findings challenge the assumption that visual prioritization requires a corresponding conscious interpretation. Participants often dedicated significant cognitive load to theoretical threat cues, then rationalized and dismissed the threat in explicit questionnaires. Eye-tracking captures assessments that explicit surveys may erase and isolates structural mechanics of criminal expertise. Understanding these hidden perceptual mechanisms is important for situational crime-prevention strategies that account for how offenders actually see the environment, rather than only how they say they do.