Help Center › Why AlphaVote exists
Why we hide the results until you vote 🔬
It feels like friction. It's actually the entire point — and there are 20 years of science behind it.
Crowds are only wise under one condition
In 2004, James Surowiecki's The Wisdom of Crowds made a famous observation: large groups can out-judge experts — but only when three things hold: diversity of opinion, independence of judgment, and a way to aggregate. Break independence, and the magic dies.
Social media broke independence. Every poll shows you the score before you answer; every post shows likes before you form a view. You're not asked what you think — you're asked whether you agree with the crowd.
The experiment that proves it
In 2011, researchers at ETH Zurich ran a now-classic experiment (Lorenz, Rauhut, Schweitzer & Helbing, published in PNAS): groups answered estimation questions, with or without seeing others' answers.
The result: merely showing people the group's estimates made opinions converge, killed the diversity that makes crowds smart, and — the cruel part — made everyone more confident while the group got less accurate.
That's herding: the loudest early answers drag everyone else. Psychology has documented the ingredients for decades — Asch's conformity experiments (people deny what their own eyes see to match a group), and "information cascades" (once a few people choose, it becomes rational for everyone after to just copy).
What the decision scientists prescribe
Daniel Kahneman's last book, Noise (2021), gives organisations one core piece of "decision hygiene": collect everyone's independent judgment first, discuss second. The Delphi method — used since the 1960s for serious forecasting — is built on the same rule: structured, independent rounds beat open discussion.
AlphaVote is that rule, turned into a consumer product:
- Commit before reveal — your judgment enters the pool uncontaminated;
- Reasoning attached — confidence, a reason, a time horizon (because opinions with arguments are worth more than likes);
- The counter-view — after committing, you face the other side's best argument, not its worst strawman;
- Calibration rewarded — predictions settle against reality, and being right and calibrated earns more than being loud.
And the Learn cards?
Same philosophy, different science: answering before seeing the solution (the testing effect, Roediger & Karpicke 2006) is the single most evidence-backed study technique there is. A Learn card is that technique in card form — commit an answer, then the explanation unlocks.
The one-sentence version
Social networks broke the independence condition of the wisdom of crowds. AlphaVote rebuilds it — one committed, reasoned vote at a time.