STUDIO TORIUMI
Article 1 of Japan's Seventeen-Article Constitution (604 CE) is widely quoted as "cherish harmony; avoid discord." Read to the end, it says the opposite. The point sits in the final sentence — and the method that genuinely changes people's minds has now been tested in the field. A 1,400-year-old sentence and a 10-year-old experiment agree.
一に曰く、和を以て貴しと為し、忤ふること無きを宗とせよ。人皆党あり、亦達れる者は少し。是を以て、或いは君父に順わず、上に逆らい下を悩ます。然れども、上和ぎ下睦びて事を論ずれば、事理自ずから通ずべし。
Seventeen-Article Constitution, Article 1 (original text)
What gets quoted is the opening — "cherish harmony, let there be no resistance." Stop there, and the article reads as "don't rock the boat."
But the sentence continues: "everyone has their own faction, and few are truly perceptive" — it begins by granting that disagreement is the default. And then it ends like this:
When those above and below are reconciled and then argue the matter out,
the right course will appear of itself.
Not "don't argue." Build harmony first, so that you can argue. Harmony was never the goal. It was the precondition.
Note what the text assumes: opinions will split. It is not aiming for consensus. The stated aim is jiri onozukara tsūzu — a state in which the sound line of reasoning becomes visible on its own.
This sounds like moralizing. There is an experiment.
In 2016, political scientists Broockman and Kalla reported in Science on door-to-door canvassing: a conversation of roughly ten minutes. Effects were measured with a randomized controlled trial.
And it worked across party lines — Democrats and Republicans, liberals and conservatives, across gender and race.
What didn't work, by contrast: television ads and direct mail. Measured in the same research tradition, their effect is near zero, and what little there is fades within days. "Deliver the correct message, loudly, over and over" does not change minds. It is also the approach that receives the most money and time.
The content of the conversation. The canvassers did not argue. What they did was ask people to recall and speak about their own experience of being treated differently, then draw an analogy between that experience and the group under discussion.
The literature calls this analogic perspective-taking. Nothing is installed from outside; what is already there is drawn out.
What changes a person is not
our correctness. It is their own memory.
This is nearly the same instruction as the 1,400-year-old sentence: reconcile first, then discuss. Reverse the order and it stops working.
The experimental procedure, translated into daily life.
A caveat, to keep expectations honest. Even in the experiment, only one person in ten changed. Nine did not. It still outperforms ads and flyers by orders of magnitude. The gap between "barely works" and "doesn't work at all" is enormous.
Studio Toriumi, which hosts this site, has a comment section — with a filter that lets criticism through and blocks insults. It is the constitution's Article 1 written not as a rule but as running code.
Opposition and strong complaints pass. What gets blocked is only language aimed at a person rather than at the content. "I don't understand what you're saying" is a harsh sentence — and it passes, because it is about the substance.
The line is harder to draw than it looks. When I tested my own filter while writing this page, "whoever made this is an idiot" was slipping through — the pattern matched too narrow a set of endings.
Fixing it required care, too. Sweep in "he's no idiot — he researches carefully" (a defense) or "that's no small feat" (praise), and the filter starts blocking compliments. The most common failure of moderation is reducing language in the name of reducing insults.
Ban the words themselves, and legitimate argument gets caught too. What should be banned is direction, not vocabulary — aimed at the substance, or aimed at the person. Article 1's "argue the matter out" lives exactly inside that line.
These numbers have a history. In 2014, Science published a study with the same aim. Its data turned out to be fabricated — discovered when a graduate student called the survey firm that supposedly polled 9,500 people, and the firm said it had conducted no such survey. The paper was retracted in May 2015.
The people who made that call were Broockman and Kalla — graduate students at the time.
And they did not stop at exposing it. The next year they redesigned the study from scratch, ran it themselves, and produced a real result — the 2016 study cited above. They didn't just point at the error; they did the work. That seemed like the very subject of this page, so the story is included.