Skip to content
Citizen Infrastructure

Pre-bunking

By the time a false or inflammatory claim has gone viral, debunking it is usually a lost cause. The correction never travels as far as the lie, and the more you argue, the more attention the lie gets. Worse, when a government tries to arbitrate a fight between two angry camps, it tends to kindle the fire — fuelling the sense of a cover-up.

Pre-bunking flips the timing. Instead of correcting after the fact, you inoculate beforehand: before a divisive frame takes hold, you put out a small, shared idea that almost everyone already accepts — what Audrey Tang calls an uncommon ground — so the polarising version has less room to grow.

The cute-dog example

Section titled “The cute-dog example”

In early 2020, before the science on masks was settled, Taiwan saw the same fight brewing as everywhere else: one camp insisting only N95 masks counted, another insisting masks were useless or harmful. Rather than referee that fight, Taiwan’s government pushed out a meme of the country’s then–Premier’s dog with a paw to its mouth: “wear a mask to remind each other to keep your dirty hand away from your face.”

Hand-washing is something no one argues against — there is no anti-hand-washing camp — so the meme defused the mask war into something everybody could agree on. It was funny, so it spread on its own; measured tap-water use actually went up. Humour, Tang stresses, is a large part of why pre-bunking works: a joke you laughed at is one you remember.

From literacy to competence

Section titled “From literacy to competence”

Pre-bunking changes what “being informed” means. The old idea is media literacy — equipping each individual with critical-thinking skills to evaluate what they read. But against AI-powered persuasion, asking every person to fact-check everything alone is, in Tang’s words, “just impossible.”

So in 2019 Taiwan rewrote its school curriculum around competence instead: students work with their families and communities to investigate a claim like journalists — weighing sources, building a balanced account, publishing it for others to rely on. The three competencies are autonomy (curiosity), interaction (listening to people unlike you), and the common good (turning a zero-sum fight into a positive-sum one). By a 2022 international study, Taiwan’s 15-year-olds ranked top of the world in the sense that they can affect society — without their science scores slipping.

This sits alongside the epistemic case for democracy and the AI-assisted sensemaking in the Taiwan Alignment Assemblies, where the shared uncommon ground is itself shared back as a pre-bunk.

The one debunk that seems to work

Section titled “The one debunk that seems to work”

If pre-bunking flips the timing because debunking barely works, there’s a recent exception worth noting: a tailored AI dialogue. In a 2024 Science study, an LLM chatbot (“DebunkBot,” by Thomas Costello, Gordon Pennycook, and David Rand) held short, evidence-based, personalised conversations with people about a conspiracy they believed — and durably reduced their confidence by about 20%, an effect still measurable months later. What one-size-fits-all corrections can’t do, a patient back-and-forth that meets each person’s specific reasons apparently can.

Sources

Section titled “Sources”