Key claims
COVID-19 misinformation has severe real-world consequences. People have died from drinking disinfectant, believing sunlight cures COVID, and rejecting vaccines based on misleading headlines. The Wikipedia list of "unproven methods against COVID-19" documents hundreds of false treatments. Misleading yet technically factual headlines (e.g., "Healthy doctor died two weeks after receiving vaccine") went viral and eroded vaccine trust despite subsequent corrections.
Misinformation can be factually correct yet deeply misleading. Truth and manipulativeness are distinct. A headline containing only correct facts can imply falsehoods through omission, emotional language, or juxtaposition. This reframes the problem from "fact vs. fiction" to "how manipulative is this message," bypassing political contestation over truth.
Technique-based inoculation scales better than issue-specific inoculation. Teaching people to recognize specific false claims (e.g., "vaccines contain microchips") does not work because there are infinitely many such claims. Instead, teaching people to spot recurrent manipulation techniques (emotional manipulation, logical fallacies, false expertise, conspiracy framing) creates broad-spectrum resistance applicable to novel misinformation.
Inoculation theory parallels medical immunology. A weakened pathogen triggers antibodies; weakened exposure to misinformation techniques triggers cognitive resistance. Two mechanisms: forewarning ("watch out, someone might manipulate you") and preemptive refutation ("this claim uses X technique, which is unreliable because...").
The Go Viral game teaches three core manipulation techniques related to COVID-19. Participants play as operators of an online echo chamber spreading COVID-19 misinformation using three scenarios: (1) emotional language (exploiting fear and anger), (2) fake expertise (false authority), and (3) conspiracy reasoning (attributing events to coordinated plots).
Study 1 results validated the game's learning effect. A simple pre-post survey (N unspecified, 6 social media items) showed that playing Go Viral significantly increased participants' perception of misinformation posts as manipulative, while leaving perception of real news unchanged — evidence that the game taught discrimination rather than general skepticism.
Study 2: RCT across three countries with 1,777 participants. Participants were randomly assigned to: (a) Go Viral game (b) UNESCO infographics on misinformation tactics (c) Tetris control. Measurements at three timepoints: before intervention (T1), immediately after (T2), and one week later (T3). The study tested three research questions: Does the intervention improve ability to spot manipulative content? Does it boost confidence in that ability? Does it reduce willingness to share misinformation?
Go Viral substantially outperformed UNESCO infographics. Immediately after intervention (T2), the Go Viral condition showed significantly larger gains in perceived manipulativeness of COVID-19 misinformation than the infographics condition. Effect sizes differed substantially — the game's effect was much larger.
The Go Viral effect persisted one week later; infographics' did not. At T3 (one week post-intervention), participants in the Go Viral condition still showed significantly better discrimination between manipulative and non-manipulative content compared to control. Infographics' effects had essentially dissipated and were no longer significantly different from control.
Paradoxical finding: heightened skepticism toward real news immediately post-game. Immediately after playing Go Viral, participants also rated genuine COVID-19 news as significantly more manipulative than baseline, suggesting an initial over-generalization of skepticism. However, this effect disappeared by the one-week follow-up, suggesting a normalization process.
Study limitations honestly acknowledged. Study 1 lacked a control group, had only 6 items, and was English-only, raising questions about generalizability and whether reading alone might work as well as playing. Study 2 addressed these but is still a lab-based online study, not a field deployment.
Psychological resilience is necessary but insufficient. Teaching people to recognize misinformation techniques is valuable but does not solve the problem if algorithms preferentially amplify manipulative content or if malicious actors escalate sophistication. Regulatory, algorithmic, and institutional solutions are complementary.
Longitudinal effects require reinforcement. Work by Roozenbeek's collaborators on the original Bad News game showed effects persist weeks post-play, but fade over time without booster reminders. Real-world deployment would require periodic reinforcement, not one-time exposure.
Methods and evidence
Inoculation framework:
- Theory originated with William McGuire (1960s), initially applied to resist "brainwashing" of U.S. POWs in the Korean War
- Two components: forewarning (activates critical attention) and preemptive refutation (provides cognitive tools)
- Analogy to medical vaccination: weakened exposure to pathogen/technique triggers defenses
Study 1 (simple pre-post, English-language Go Viral):
- Participants: N not explicitly stated
- Design: within-subjects; participants rated social media posts before and after playing
- Measurement: 6 social media items (3 misinformation, 3 real news) rated for manipulativeness on a scale
- Results: Misinformation perceived as significantly more manipulative post-game; real news unchanged
Study 2 (RCT, cross-cultural):
- Participants: N=1,777 across UK, Germany, France
- Design: randomized between-subjects with three conditions (Go Viral game, UNESCO infographics, Tetris control)
- Timepoints: T1 (baseline), T2 (immediately post-intervention), T3 (one week post-intervention)
- Interventions:
- Go Viral: ~5-minute game; participants play as disinformation operator using three manipulation techniques
- UNESCO "Think Before Sharing" infographics: static visual materials teaching how to spot misinformation
- Tetris: 5-minute control game, unrelated to misinformation
- Measures:
- 18 social media items (misinformation and real news)
- Three questions per item: (1) "This post is manipulative" (7-point Likert), (2) "I'm confident in my ability to judge manipulativeness," (3) "I would share this with my network"
- Results:
- T2: Go Viral condition > infographics condition > control in perceived manipulativeness of misinformation (large effect size for game, smaller for infographics)
- T3: Go Viral still significantly better than control; infographics no longer significantly different from control
- Real news: Go Viral condition showed heightened skepticism at T2 (unexpected), which normalized by T3
- Confidence and sharing intentions followed similar patterns (improvements for Go Viral > infographics > control)
Notes
This talk is significant for presenting detailed empirical results from one of the largest cross-cultural RCTs of inoculation-based interventions against COVID-19 misinformation. By comparing an interactive game (Go Viral) directly against a lower-friction alternative (infographics), the study provides evidence for what makes inoculation interventions effective beyond mere exposure to warnings.
The key finding—that the game substantially outperformed infographics in both immediate effect size and one-week persistence—suggests that active, scenario-based learning (where participants role-play and make decisions) produces stronger and more durable cognitive resilience than passive consumption of educational materials. This has implications for scaling psychological interventions: while infographics distribute easily, the behavioral engagement required by gamified approaches may be necessary for robust effects.
The cross-cultural dimension (UK, Germany, France) strengthens the claim that technique-based inoculation generalizes across linguistic and political contexts. The resilience of the effect one week later is particularly important for real-world deployment, where forgetting and fade-out are major threats to intervention durability.
The paradoxical increase in skepticism toward legitimate news immediately post-game is noteworthy and under-explained in the talk. One interpretation: the game creates an acute heightened state of critical cognition that subsides as normal information processing resumes. Alternatively, participants may initially over-apply newly-learned patterns before calibrating. This deserves follow-up investigation, as it hints at a potential trade-off between skepticism and gullibility.
The acknowledged limitations are honest: this is a controlled online study, not a field deployment. Real-world effectiveness depends on whether participants actually behave differently (e.g., share less misinformation), not just perceive it differently. The speaker notes ongoing experiments to test behavioral outcomes, a crucial next step.
Finally, the emphasis that inoculation is necessary but not sufficient is intellectually candid. Psychological interventions operate at the cognitive level; if platforms systematically amplify manipulation or if disinformation operators escalate sophistication, individual psychological resilience alone cannot solve the problem. Complementary regulatory and algorithmic interventions are essential.