Online vs Offline Harassment: Perceived Severity and Behavioural Intentions in a Vignette Study

Participants will be shown four randomly assigned vignettes describing social interactions involving incidents of sexual harassment initiated by a stranger or unknown user. Each vignette systematically varies along three dimensions:
1. Setting: Online vs. in-person
2. Victim Gender: Man, woman, or non-binary
3. Harassment Type: Verbal (e.g., unwanted comments) vs. image-based (e.g., unsolicited explicit content)
To reduce demand effects and obscure the primary focus of the study, each participant will also be shown one filler vignette describing unrelated but socially ambiguous or neutral interactions.

2025-09-16

2025-12-31

Our primary outcomes capture participants’ evaluations of harassment scenarios. These include:
1. Perceived Severity (self-reported): Participant rating of how serious the behaviour described in each vignette is, on a 7-point scale from “Not serious at all” to “Very serious.” Measured at the vignette level.
2. Perceived Social Norms (incentivised belief): Participants indicate how socially appropriate they believe the behaviour in the vignette is, in an incentivised coordination game designed to match the modal response (following Krupka & Weber, 2013). Responses are on a 5-point ordinal scale from "Very socially appropriate" to "Very socially inappropriate." Measured at the vignette level.
3. Composite Attitudinal Index

Composite Attitudinal Index: A principal components factor analysis (PCA) will be conducted on the attitudinal responses collected at the vignette level (perceived severity, perpetrator and victim responsibility, anger, likelihood of intervention, punishability) to identify underlying dimensions such as moral judgment or emotional salience. Factor scores will be used as dependent variables

1. Anger (self-reported): Participant rating of how angry the scenario makes them feel, measured on a 7-point ordinal scale. Measured at the vignette level.
2. Willingness to Intervene (self-reported): Self-reported likelihood of reporting the incident to someone in authority, rated on a 7-point ordinal scale. Measured at the vignette level.

3. Punishability (self-reported): Perception of the most appropriate sanction for the behaviour described in the vignette, choosing one option between:
a. No punishment
b. Verbal warning from someone in a position of authority
c. Formal police warning
d. Restorative justice meetings between the victim and the offender, facilitated by trained professionals
e. Community service
f. Prison sentence

4. Responsibility Attribution (self-reported): Participant ratings of the degree of responsibility attributed to:
a. The perpetrator
b. The victim
Both measured on 7-point ordinal scales. Measured at the vignette level.

5. Donation Decision (incentivised behaviour): Participants are entered into a lottery to win £100. They can choose how much of the prize to donate to a UK charity addressing online harassment. We will analyse:
a. A binary indicator of whether the participant donates any amount (0/1).
b. A continuous measure of the amount donated (ranging from £0 to £100).
Measured at the participant level.

6. Petition Support (behavioural measure): Participants are asked whether they support a petition advocating for stronger legal protections against online sexual harassment. This will be analysed as a binary variable (1 = supports petition, 0 = does not).
Measured at the participant level.

We conduct an online survey experiment to examine public attitudes towards online sexual harassment, using a UK sample of approximately 600 participants recruited via Prolific. The study uses a vignette-based design, where participants read short, fictional scenarios describing incidents of harassment. Key features of each scenario, such as the setting (online vs in person), the gender of the victim, and the form of harassment (verbal or image-based), are systematically varied across vignette. After reading each vignette, participants rate the incident on several outcomes, including perceived severity, responsibility, emotional response, and likelihood of intervention. The study also includes questions about social norms and measures of real-world behaviour.

Not available

Randomization will be done through Qualtrics

Individual level

No

We aim to recruit around 600 participants.

Around 600 participants

We have a total of twelve unique vignette combinations, created by varying three dimensions with 2x3x2 levels. Each participant will be randomly shown four vignettes, drawn without replacement from the full set, such that no participant sees the same combination twice, for a total of 600 × 4 = 2400 total vignette responses. Dividing this evenly across the twelve vignettes, 2400 / 12 ≈ 200 responses per vignette. Thus, we expect each vignette to be rated by approximately 200 participants.