Self-reported fake news exposure greatly exceeds model-based estimates

Evidence

However, all of these studies still run the risk of severely underestimating the problem. According to a national survey from Pew, in December 2016, 32 per cent of Americans said they often encountered political fake news online; 51 per cent said they often see inaccurate news; and 23 per cent self-report having shared fake news. In Europe, 68 per cent of those surveyed report seeing fake news on a weekly basis. Something doesn’t add up. Either people’s memories are suspect, or the models are way off. — Sander van der Linden, Foolproof, 525

Categorization

• QUESTION: Why does self-reported fake news exposure (32-68%) dwarf machine-measured shares (0.15-6%)? Worth chasing what explains such a wide mismatch — recall bias, model flaws, or category differences.
• SEED: A measurable tension between survey self-reports and trace-data models — usable for an article on why the ‘fake news was trivial’ rebuttal may understate the problem.
• XREF: connects to: Average US adult saw only about three fake news pages during the 2016 election — Sander van der Linden, Foolproof: Why Misinformation Infects…, 517; The economists’ estimate that fake news was rare depends on sampling and memory assumptions — Sander van der Linden, Foolproof: Why Misinformation Infects…, 519; Self-reported fake news exposure is mostly false memory rather than real contact — Sander van der Linden, Foolproof: Why Misinformation Infects…, 518