Skip to content

Bad News / Harmony Square / Bad News Game

FreeFirst-line

Publisher: goviralgame.com | Type: Non-detector | Cost: Free

TL;DR

The foundational inoculation-game architecture in the prebunking field, developed by Cambridge's Social Decision- Making Lab and DROG. Available in fifteen languages, none of them currently SEA. Included in the toolkit as adaptable architecture rather than as a deployable Bahasa, Thai, Filipino, Sinhala, Tamil, Malay, or Lao tool. Where a fact-checker wants to commission a localised inoculation game, this is the model.

What it does

Bad News is a browser-based inoculation game in which the player adopts the role of a misinformation producer, building a fake news outlet, accumulating followers, and progressing through six manipulation techniques (impersonation, emotional language, polarisation, conspiracy, discrediting, trolling). The pedagogical premise is inoculation theory: by walking the player through how misinformation is produced, the game builds recognition of those same techniques when the player encounters them in the wild. The Cambridge SDML / DROG team has published extensively on the game's effectiveness, with cross-cultural validation across four-plus languages and a 5,061-participant sample, and Roozenbeek & van der Linden's foundational 2019 study reached approximately 15,000 participants.

The Bad News game is one of three games in the family. Harmony Square covers political disinformation and intergroup polarisation; Go Viral covers COVID-19 misinformation. Cambridge SDML and DROG describe more games in development. The Bad News intervention has been adapted in collaboration with the UK Foreign and Commonwealth Office into fifteen language versions — Czech, Dutch, English, Esperanto, German, Greek, Polish, Portuguese, Romanian, Russian, Serbian, Slovenian, Swedish, Ukrainian, plus dedicated Bosnian and Moldovan-Romanian deployments, but none of the seven SEA languages the toolkit covers.

For the toolkit, Bad News is the architectural reference. A fact-checker thinking about commissioning a Thai, Filipino, Sinhala, Tamil, Malay, or Lao inoculation game has the Bad News codebase, methodology, and effectiveness research as the starting point. Direct deployment in any of the six focus countries faces the language-coverage gap; adaptation is the operational route.

When to use it

  • A media-literacy programme operator is designing a localised SEA-language inoculation game and wants the foundational Cambridge SDML / DROG methodology and game architecture as the starting point.
  • A trainer working in English with mixed-language audiences (Bahasa-with-English speakers, Filipino-with-English speakers, English-medium classrooms in Sri Lanka or Malaysia) wants a research-anchored game whose mechanics transfer across languages even where the localised version does not exist.
  • A research partner needs the published peer-reviewed effectiveness research on inoculation games for grant reporting, advocacy work, or sustained programme design.
  • A workshop is teaching the prebunking field as a discipline — not deploying a single game, and Bad News is the canonical example of the inoculation-by-active-play approach.

Limitations

Limitations

  • Effects decay within days to weeks without booster sessions. Single-session deployment is one input into a sustained programme, not a stand-alone intervention.
  • Reach is limited. Active-play games require deliberate engagement; passive audiences who would never click into a 15-minute browser game are not reached.
  • Not yet localised in any SEA language. The fifteen available language versions are Eastern European, Western European, and Esperanto. Bahasa, Thai, Filipino, Sinhala, Tamil, Malay, and Lao adaptations would need to be commissioned from Cambridge SDML / DROG / Gusmanson under the standard adaptation route. The toolkit's cell-level honest-gap on localised prebunking effectiveness outside Indonesia applies here directly.
  • Game scenarios were authored against Western and European information environments; adaptation requires local creative work to ensure scenarios resonate with SEA digital habits (the way Tempo Detektif Deepfake and IREX's Gali Fakta were both deliberately tuned to Indonesian context).

Privacy and threat model

The game is publicly accessible at getbadnews.com (English) and the language-specific URLs. Players load the game in the browser; no account creation is required. Standard web analytics apply for the Cambridge SDML and DROG infrastructure; players who want to participate in research evaluations sometimes opt into surveys, and that participation is documented inside the game flow.

The deployment-side threat model is the same as the rest of the 2A.3 cell: the game is editorial content, not data-handling infrastructure. The decision a fact-checker makes is whether to deploy it to an English-comfortable SEA audience as a stop-gap, or to commission localised adaptation as the longer-term investment.

Country and platform applicability

  • Indonesia: not localised in Bahasa Indonesia. The Cambridge SDML / DROG game architecture is the reference pattern; the Indonesian deployment that exists is Tempo Detektif Deepfake (Tempo Detektif Deepfake) and IREX's Gali Fakta (IREX Learn to Discern (L2D) / Gali Fakta), both of which sit operationally in the same inoculation-game family. The Facciani et al. 2026 cross-cultural study explicitly compared Gali Fakta and Harmony Square in Indonesia and found Gali Fakta the more effective deployment for Indonesian audiences, sustaining the case that locally authored adaptations outperform direct cross-cultural deployment.
  • Laos: not localised. The cell-level honest-gap on localised prebunking outside Indonesia applies. Lao adaptation would require partner-led commissioning.
  • Malaysia: not localised in Malay. Sebenarnya.my, a Malaysian university partnership, or MAFINDO-Malaysia could plausibly lead a Malay adaptation.
  • Philippines: not localised in Filipino. The #FactsFirstPH coalition or a VERA Files-led partnership could plausibly lead.
  • Sri Lanka: not localised in Sinhala or Tamil. Hashtag Generation, Watchdog, or DRI could plausibly lead a Sri Lanka-specific adaptation; the asymmetric Sinhala-vs- Tamil tooling pattern from regional research means an adaptation effort would need to plan deliberately for both languages rather than starting in Sinhala and trusting Tamil to follow.
  • Thailand: not localised in Thai. Cofact Thailand or AFP Fact Check Thailand could plausibly lead.

Platform applicability: the game is a web page; reach across any platform that distributes URLs to players (Facebook, Facebook Groups, TikTok organic posts, LINE, WhatsApp shareable links).

How to access

Free public access at getbadnews.com (English) and at the fifteen language-specific URLs documented on the Cambridge SDML research page (badnewsgame.se for Swedish, getbadnews.de for German, getbadnews.gr for Greek, and so on). The inoculation.science research-hub site documents the broader game family. Cambridge SDML invites organisations interested in translation or adaptation to contact Dr Jon Roozenbeek or Prof van der Linden directly.

Cost (current as of 2026-05)

Free for play in any of the existing fifteen language versions. Adaptation cost (translation, contextual rewriting, language-specific deployment hosting, evaluation methodology) is absorbed by the deploying institution, the partnership with Cambridge SDML / DROG / Gusmanson, and any funder behind the adaptation effort. The toolkit does not pin a figure for a hypothetical SEA-language adaptation.

Quickstart (deployment-side)

  1. Identify the deployment frame: direct play in English with an English-comfortable audience as a stop-gap; commissioning a localised adaptation as the longer-term investment; methodology reference for designing a separate game.
  2. For direct stop-gap deployment, route the audience to getbadnews.com (English) and frame the activity as a 15-minute prebunking exercise.
  3. For localised adaptation, contact Dr Jon Roozenbeek or Prof van der Linden through the Cambridge SDML page; plan for the standard adaptation timeline and the funder relationship that adaptation requires.
  4. For methodology reference (designing a different game on the inoculation-by-active-play pattern), cite the Roozenbeek & van der Linden 2019 study, the cross-cultural validation literature, and the Facciani et al. 2026 cross-cultural comparison.
  5. Pair the deployment with reinforcement; inoculation effects decay without booster sessions.
  6. Read the Facciani et al. 2026 finding closely if the audience is Indonesian; Gali Fakta outperformed Harmony Square in that study, and the cross-cultural transfer evidence supports prioritising locally authored deployments over direct adaptation.

In the toolkit's workflow

Source-history and behaviour pillar non-detector audience- engagement reference per Architectural Anchor 1. Sits in 2A.3 as the alternative entry that anchors the inoculation-game landscape; the operational deployment recommendations for the focus countries route through the locally authored Tempo Detektif Deepfake and the IREX-led Gali Fakta instead of through direct Bad News deployment.

Standard combinations (deployment combinations):

  • With Tempo Detektif Deepfake at 2A.3 – Bad News is the methodology reference, Tempo Detektif Deepfake is the Indonesian deployment that operationalises the same pattern in Bahasa with deepfake-specific scenarios.
  • With IREX L2D / Gali Fakta at 2A.3 – Bad News family covers the underlying game pattern, IREX's Gali Fakta is the Bahasa-tuned WhatsApp-group-chat-style deployment; Facciani et al. 2026 shows Gali Fakta outperforms Harmony Square for Indonesian audiences specifically.
  • With Jigsaw Prebunking at 2A.3 – as the research-base complement; Jigsaw and Bad News together represent the two largest published prebunking research bases.
  • With Cambridge SDML / DROG as adaptation partners when an SEA-language game commissioning is on the table.

Decision-tree references: this is not a verification routine; the decision trees in T1–T7 do not invoke Bad News at runtime. The reference sits in the editorial-strategic layer that shapes audience preparation and the methodology layer that shapes future SEA-language adaptation work.

This card carries no detector signal class: Bad News is an inoculation game and adaptable architecture; it does not produce a verdict on any specific piece of suspect content.

Override notes

Override notes

  • No SEA language localisation yet (b2-disclaimer-pass-no- SEA-language; sea-localisation-not-yet-pointer): the fifteen available language versions are Eastern European, Western European, and Esperanto. Bahasa, Thai, Filipino, Sinhala, Tamil, Malay, and Lao adaptations require partner-led commissioning. The toolkit treats this as the foundational architecture for adaptation rather than as a directly deployable SEA tool, and routes operational Indonesian deployment through Tempo Detektif Deepfake and IREX Gali Fakta.

Sources