How to "inoculate" against multimodal misinformation: A conceptual replication of Roozenbeek and van der Linden (2020)

Abstract

Building misinformation resilience at scale continues to pose a challenge. Gamified "inoculation" interventions have shown promise in improving people's ability to recognize manipulation techniques commonly used in misinformation, but so far few interventions exist that tackle multimodal misinformation (e.g., videos, images). We developed a game called Cat Park, in which players learn about five manipulation techniques (trolling, emotional manipulation, amplification, polarization, and conspiracism), and how misinformation can spread through images. To test the game's efficacy, we conducted a conceptual replication (N = 380) of Roozenbeek and van der Linden's 2020 study about Harmony Square, with the same study design, item set, and hypotheses. Like the original study, we find that people who play Cat Park find misinformation significantly less reliable post-gameplay (d = 0.95, p < 0.001) compared to a control group, and are significantly less willing to share misinformation with people in their network (d = 0.54, p < 0.001). These effects are robust across different covariates. However, unlike the original study, Cat Park players do not become significantly more confident in their ability to identify misinformation (p = 0.204, d = - 0.13). We did not find that the game increases people's self-reported motivation and confidence to counter misinformation online.

Figure 1

Screenshots from the Cat Park game.

Figure 2

Experimental design flowchart.

Figure 3

Box plots and point range plots with violin plots and data jitter for the pre-post difference scores for the reliability (top left), confidence (top middle), and sharing (top right) measures, for both the inoculation (Cat Park) and control (Tetris) groups.

Figure 4

Box plots and point range plots with violin plots and data jitter for the pre-post difference scores for the perceived reliability of real misinformation (top left), perceived reliability of fake misinformation (top right), confidence in detecting real misinformation (middle left), confidence in detecting fake misinformation (middle right), sharing intentions of real misinformation (bottom left), and sharing intentions of fake misinformation (bottom right) measures, for both the inoculation (Cat Park) and control (Tetris) groups.

Figure 5

Example of a social media post used in the survey, along with the outcome measures (reliability, confidence, and sharing intentions).