Fig 1.
a—c: Representative ‘social media’ content (representative images taken from the Socio-Moral Image Database (SMID), [30]): (a) Morally positive image with poster approval; (b) Morally neutral image with poster disapproval; and (c) Morally negative image with poster approval. d & e: Judgment and Punishment tasks: (d) Judgment task with choice options (‘Like’, ‘Dislike’ or ‘Report’); (e) Punishment task with slider ranging from zero-day punishment (No Ban) to 30+ day punishment (Permanent Ban).
Fig 2.
(a) Median reporting rates between Groups, visualised using boxplot and violin plot with jittered data points. The y-axis represents reporting rate (in %), and the x-axis shows the two Groups. The white box indicates mean values and violin plot displays the distribution of data. (b) Model estimates expressed as odds ratio for each of the model parameters. Significance ***p < .001; *p < .05. (c) Proportion of decisions to ‘Report’ across Images and Poster Intention, between Groups. Error bars represent 95% Confidence Intervals.
Table 1.
Logistic regression model output for the likelihood of reporting social media content, showing fixed effects estimates, standard errors, z-values, and p-values.
Fig 3.
(a) Distribution of punishment (No ban (0 days)–Permanent ban (30+ days)) between Groups, and across Images and Poster Intention. (b) Boxplot depicting punishment score per Image Category (separated into panels according to Poster Intention and Group). The lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles). Jittered data reflects outliers. (c) Scatter plot depicting the relationship between reporting rates (in percentage) in the Judgment Task and mean punishment (in days) in the Punishment Task.
Table 2.
Linear mixed regression model output for the likelihood of reporting social media content, showing fixed effects estimates, standard errors, z-values, and p-values.
Fig 4.
Trust and use of social media platforms.
(a) Trust in platform (in percentage), captured in response to the question: "I trust this social media platform and believe that the people/company that run and manage it are honest."; (b) Distribution of frequency of use (in percentage) across 12 social media platforms; and (c) Total reporting rate (percentage reported in the Judgment task) as a function of participants trust in YouTube, Instagram, Twitter/X and Facebook.
Fig 5.
Scatter plots depicting relationship between empathy subscales and reporting rates in the Judgment Task (top row) and mean punishment in the Punishment Task (bottom row). Spearman’s rho correlations were conducted to explore the association between empathy subscales (Empathetic Concern (EC), Perspective Taking (PT), and Personal Distress (PD)) and reporting rates in the Judgment Task and mean punishment in the Punishment Task respectively.