Might People On The Internet Sometimes Lie?

Starting from Reddit 'past lives' and 'glitch in the matrix' stories that seem to demand paranormal or wild-coincidence explanations, Scott offers the obvious-but-emotionally-hard alternative: people lie. On a site with a million daily readers, even 1/10,000 trolls yields ~100 fabrications, of which the ten most convincing rise to the top. Generalized: 'interesting things should usually be lies,' because liars can craft stories that perfectly fit a narrative while true stories carry complicating subtleties. Three applications: Toxoplasma (the crusade-anecdotes that bubble up are disproportionately false), his own links posts, and — the standout — bad science (sloppy/biased researchers are 'the Reddit liars of science,' producing more interesting false positives, out-competing and driving out the careful truth-tellers into a stable bad equilibrium). Ends puzzled that his System 1 still refuses to believe the Reddit storytellers are lying.

Quality 76 · Excellent. A strong, original epistemics essay with a memorable, transferable principle (selection-for-fabrication in large communities) and a genuinely sharp application to the replication crisis (interesting-but-false results dominate because their producers can generate more of them). Excellent-low.

Claude’s paradigm shift 62 · Notable shift. The 'interesting things should usually be lies / selection-for-fabrication' framing and the bad-science-equilibrium application are original and memorable. Notable.

Real-world impact 2 · Minor. A within-discourse epistemics essay; no material-world effect. Within-blog influence.