Concise Summary简洁概述
Yudkowsky opens with Wason's 2-4-6 task: most people try examples that fit their hypothesis and never test cases that would falsify it. Only 21% find the real rule. This is positive bias — the instinct to look into the light rather than the dark. It differs from confirmation bias in that it operates sub-verbally, on the level of which example automatically pops into your head. Knowing about it intellectually is not enough; you must retrain the reflex to flinch toward disconfirming evidence. The same flaw haunts broader reasoning: if a theory can explain both an outcome and its opposite equally well, it has zero predictive power — and you only discover this by looking for the negative cases you'd rather ignore.
Yudkowsky 以沃森的 2-4-6 任务开场：大多数人只测试符合自己假设的例子，从不测试能证伪它的情形。只有 21% 的人找到了真正的规则。这就是正向偏差——本能地只看向光亮而非黑暗。它与确认偏差的区别在于：它在语言之下运作，在哪个例子自动蹦入脑海的层面运作。仅凭理智层面的了解是不够的；你必须重新训练那个本能，让它转向反驳性证据。同样的缺陷也困扰着更宏观的推理：若一个理论对某结果及其对立面均能解释，它的预测力为零——而你只有主动寻找那些你宁愿忽视的负面案例，才能发现这一点。
Infographic信息图
We only shine the flashlight on 'Yes'
我们只把手电筒照向「是」
When testing a hypothesis, people naturally generate examples that should fit, and stop when they get a 'Yes' — never probing what a 'No' would look like.
测试假设时，人们自然地生成应该符合的例子，得到「是」便停手——从不追究「否」会是什么样子。
Cold logic, hot mistake
冷逻辑，热错误
Positive bias operates even in emotionless, purely logical tasks — it's a sub-verbal reflex, not a consequence of emotional investment.
正向偏差即便在无情绪的纯逻辑任务中也会出现——它是一种语言之下的本能反应，而非情感投入的产物。
Phlogiston's unfalsifiable glow
燃素说无法证伪的光芒
Phlogiston theory could explain a flame going out and not going out — which means it explained neither. Recognizing this requires actively searching for negative evidence.
燃素说既能解释火焰熄灭，也能解释火焰不熄灭——这意味着它什么都没解释。意识到这一点，需要主动搜寻负面证据。
Strength lies in what a theory cannot explain
理论的强度在于它不能解释什么
If you can explain every outcome equally well, you have zero knowledge. A theory's true power comes from the results it rules out.
如果你对所有结果都能同样好地解释，你就没有任何知识。理论真正的力量来自它所排除的结果。
Retraining the reflex
重新训练那个本能
The fix cannot be purely verbal — you must learn, wordlessly, to zag instead of zig: to flinch toward the zero, toward the dark, instead of away.
修正不能只停留在语言层面——你必须在无言中学会：该向左时向右，本能地朝向零点、朝向黑暗，而不是逃离它。
Detailed Summary详细概述
The Classroom Setup
Yudkowsky opens in the voice of a teacher writing three numbers on a blackboard: 2-4-6. Students are told a rule governs such sequences and may test as many triplets as they like. One student's record shows 4-6-2 (No), 4-6-8 (Yes), 10-12-14 (Yes) — then they announce their rule. What rule would you have guessed? And would you have tested anything else first?
This is Peter Wason's famous 2-4-6 task. Despite high reported confidence, only 21% of subjects correctly identified the real rule. The study was provocatively titled "On the failure to eliminate hypotheses in a conceptual task."
The Core Flaw: Positive Rather Than Negative Examples
Subjects typically generate triplets that confirm their working hypothesis. Someone who suspects "numbers increasing by two" tests 8-10-12, hears "Yes," and stops. Someone who suspects X-2X-3X tests 3-6-9 and announces that rule. In every case the actual rule is the same: the three numbers must be in ascending order. Discovering this requires testing triplets that shouldn't fit — something like 20-23-26, or a decreasing sequence — and seeing if they are labeled "No."
Yudkowsky distinguishes this from ordinary confirmation bias. Ordinary confirmation bias involves protecting a cherished belief. Positive bias is more primitive: it's the reflex of generating positive examples rather than negative ones when testing any hypothesis, even a cold, purely logical one with no emotional stakes at all.
The Phlogiston Example
Phlogiston theory — the pre-Lavoisier idea that burning things release a substance called phlogiston — could explain a flame going out in an enclosed box: the air became saturated with phlogiston, so no more could be released. But the theory could equally well have explained the flame not going out. When a theory accounts for every possible outcome, it has no genuine explanatory force. To notice this, you have to search for negative examples instead of positive ones, look into zero instead of one — which cuts against human instinct.
For by instinct, we human beings only live in half the world.
Why Knowing Isn't Fixing
One may be lectured on positive bias for days and still overlook it in the moment. The 2-4-6 task is emotionally "cold" — there is no attachment to a treasured hypothesis — and yet the mistake persists. That is because positive bias does not arise from following a deliberate rule; it operates at the level of imagery and instinctive reaction. It is sub-verbal.
This matters for the cure. You cannot fix a sub-verbal habit with a verbal reminder. You have to wordlessly retrain yourself to "zag instead of zig" — to flinch toward zero, toward the dark case, instead of away from it.
Strength = What a Theory Cannot Explain
Yudkowsky links positive bias to his broader epistemological claim: the strength of a hypothesis is what it cannot explain, not what it can. If a hypothesis is equally good at explaining any outcome, it carries zero information. He illustrates this with an example from his own writing: when he challenged the usefulness of "emergence" as a concept, a commenter cited superconductivity and ferromagnetism. Yudkowsky replied that non-superconductivity and non-ferromagnetism are also examples of "emergence" — which is precisely the problem.
The Closing Challenge
The essay ends with a direct challenge to the reader: Are you right now searching for positive examples of positive bias — or are you also looking for where positive bias should lead you not to see? Did you look toward the light, or the darkness? This recursive close forces the reader to notice whether they are falling into the very trap being described.
课堂场景
Yudkowsky 以一个教师的口吻开场，在黑板上写下三个数字：2-4-6。学生被告知有一条规则支配这样的数字序列，他们可以测试任意多个三元组。某个学生的记录显示：4-6-2（否）、4-6-8（是）、10-12-14（是）——然后他宣布了自己猜到的规则。你会猜出什么规则？你会在宣布之前再测试其他什么东西吗？
这是彼得·沃森著名的 2-4-6 任务。尽管受试者普遍表达了极高的自信，只有 21% 的人正确找出了真正的规则。这项研究有一个挑衅性的标题：「论概念任务中排除假设的失败」。
核心缺陷：正例而非反例
受试者通常只生成确认其当前假设的三元组。猜测「每次增加二」的人测试 8-10-12，听到「是」，就停手了。猜测 X-2X-3X 规则的人测试 3-6-9，然后宣布那就是规则。在所有情况下，真正的规则都是同一个：三个数字必须是升序的。 发现这一点需要测试那些不应该符合的三元组——比如 20-23-26，或一个降序序列——然后看它们是否被标记为「否」。
Yudkowsky 将此与普通的确认偏差区分开来。普通的确认偏差涉及保护一个珍视的信念。正向偏差更为原始：它是在测试任何假设时，本能地生成正面例子而非负面例子的反射——即使是一个完全冷静、纯逻辑、没有任何情感牵绊的任务。
燃素说的例子
燃素说——拉瓦锡之前认为燃烧物释放一种叫燃素的物质的理论——能够解释火焰在封闭容器中熄灭：空气被燃素饱和，不能再释放更多了。但该理论同样能解释火焰不熄灭。当一个理论能解释所有可能的结果，它就没有真正的解释力。要注意到这一点，你必须搜寻负例而非正例，看向零而非一——这与人类的本能背道而驰。
因为从本能上讲，我们人类只活在世界的一半。
为什么知道不等于修复
一个人可能被讲授正向偏差好几天，却仍然在当下忽视它。2-4-6 任务在情感上是「冷」的——没有对珍视假设的依附——然而错误依然存在。这是因为正向偏差并非源于遵循某条刻意的规则；它在意象和本能反应的层面运作。它是语言之下的。
这对修复方式至关重要。你无法用语言提醒来修复一个语言之下的习惯。你必须在无言中重新训练自己「该向左时向右」——本能地朝向零点、朝向黑暗，而不是逃离。
强度 = 理论不能解释什么
Yudkowsky 将正向偏差与他更宏观的认识论主张联系起来：一个假设的强度在于它不能解释什么，而非它能解释什么。 如果一个假设对任何结果都同样擅长解释，它携带的信息为零。他以自己的写作为例加以说明：当他质疑「涌现」作为一个概念的有用性时，一位评论者援引了超导性和铁磁性作为涌现的例子。Yudkowsky 回应说，非超导性和非铁磁性同样也是「涌现」的例子——而这恰恰是问题所在。
结尾的挑战
文章以一个对读者的直接挑战作结：你现在是在搜寻正向偏差的正面例子——还是也在寻找正向偏差本该让你看不见的地方？ 你看向了光明，还是黑暗？这个递归式的收尾，迫使读者注意自己是否正在落入文章所描述的那个陷阱。
FAQ常见问答
What exactly is the 2-4-6 task, and what does it demonstrate?2-4-6 任务究竟是什么，它证明了什么？
The 2-4-6 task is a classic experiment where subjects are given one triplet that fits a secret rule and must discover the rule by proposing their own triplets. Most people only test triplets that confirm their current guess, so they never discover that the real rule — simply "ascending order" — is far broader than their hypothesis. It demonstrates that humans instinctively generate positive examples when testing ideas, rather than the disconfirming examples that would actually reveal the truth.
2-4-6 任务是一个经典实验，受试者得到一个符合某条隐藏规则的三元组，必须通过提出自己的三元组来找出该规则。大多数人只测试确认其当前猜测的三元组，因此他们永远发现不了真正的规则——仅仅是「升序」——远比他们的假设宽泛。它证明了：在测试想法时，人类本能地生成正面例子，而非那些真正能揭示真相的反驳性例子。
How is positive bias different from confirmation bias?正向偏差与确认偏差有何不同？
Confirmation bias usually refers to the tendency to favor evidence that supports a pre-existing, emotionally held belief. Positive bias is more primitive and operates even in cold, purely logical tasks with no emotional attachment. It's the instinct to generate examples that fit a hypothesis rather than examples that might break it — happening at the level of automatic imagery, not deliberate reasoning. Yudkowsky argues positive bias deserves its own name because the mechanism is distinct.
确认偏差通常指倾向于偏爱支持已有的、有情感投入的信念的证据。正向偏差更为原始，即使在没有情感牵绊的冷静纯逻辑任务中也会出现。它是本能地生成符合假设的例子、而非可能破坏假设的例子——发生在自动意象的层面，而非刻意推理的层面。Yudkowsky 认为正向偏差值得有自己专属的名称，因为其机制是不同的。
Why can't you fix positive bias just by knowing about it?为什么仅仅知道正向偏差的存在不足以修复它？
Because positive bias is sub-verbal — it operates below the level of deliberate rule-following. The 2-4-6 task is emotionally neutral, so the bias can't be blamed on defensive feelings; it fires at the level of which example automatically pops into your head. Knowing verbally that "we should test negative examples" doesn't rewire that reflex. The fix requires actual practice: training yourself to notice when you're about to reach for a positive example and to flinch toward the disconfirming one instead.
因为正向偏差是语言之下的——它在刻意遵循规则的层面之下运作。2-4-6 任务在情感上是中性的，所以偏差不能归咎于防御性情绪；它在哪个例子自动蹦入脑海的层面上触发。在语言层面知道「我们应该测试负面例子」，并不能重新接通那个反射回路。修复需要真正的练习：训练自己察觉到自己即将伸手去拿正面例子的那一刻，然后转而朝向反驳性的那一个。
What is the connection between positive bias and the strength of a theory?正向偏差与理论的强度之间有什么联系？
Yudkowsky's epistemological claim is that a theory's strength lies in what it cannot explain, not in what it can. If a theory can accommodate every possible outcome, it has zero predictive power — it's as useless as phlogiston theory, which could explain a flame going out and not going out. Positive bias is the cognitive habit that makes us miss this: we focus on the cases our theory handles well and don't go looking for the cases it can't handle.
Yudkowsky 的认识论主张是：一个理论的强度在于它不能解释什么，而非它能解释什么。如果一个理论能容纳所有可能的结果，它的预测力为零——就像燃素说一样毫无用处，它既能解释火焰熄灭，也能解释火焰不熄灭。正向偏差是让我们错过这一点的认知习惯：我们只关注理论处理得好的情形，而不去寻找它处理不了的情形。
What does the essay's closing challenge mean — 'did you look toward light or darkness?'文章结尾的挑战——「你看向了光明还是黑暗」——是什么意思？
It's a recursive trap. The essay has just explained positive bias, and the final question asks whether you searched only for positive examples of positive bias in your own thinking — cases where you do exhibit it — rather than also looking for cases where you don't. If you only nodded along thinking "yes, I see how positive bias works," you probably just did it again. The darkness is the disconfirming case: moments when you avoid positive evidence and actively seek the negative.
这是一个递归式的陷阱。文章刚刚解释了正向偏差，最后的问题问的是：你是否只在自己的思维中搜寻了正向偏差的正面例子——你确实表现出它的情形——而没有也去寻找你没有表现出它的情形。如果你只是一路点头，心想「是的，我明白正向偏差是怎么运作的」，你大概刚刚又犯了一次这个错误。黑暗是反驳性的情形：那些你主动避开正面证据、积极寻求负面证据的时刻。
Is the essay's prescription — 'look into the dark' — actually sufficient to fix positive bias?文章的处方——「看向黑暗」——真的足以修复正向偏差吗？
Yudkowsky himself admits he's still working on it. The prescription is correct in direction but doesn't specify how to retrain the sub-verbal reflex in practice. He acknowledges that "people will agree with you, but then, in the next sentence, do something subdeliberative that goes in the opposite direction." The essay gives the diagnosis and the goal; it doesn't provide a training protocol — that work is left to later essays and to the reader's own deliberate practice.
Yudkowsky 自己也承认他还在努力中。这个处方方向是正确的，但没有说明在实践中如何重新训练那个语言之下的反射回路。他承认「人们会同意你的观点，但在下一句话里，就会做出某个潜意识的动作，朝相反的方向走去」。文章给出了诊断和目标，却没有提供训练方案——那项工作留给了后续文章和读者自己的刻意练习。
In-depth Analysis · Pros & Cons深入解读 · 优缺点
This short essay is one of Yudkowsky's most practically useful pieces in the Sequences: it isolates a specific, demonstrable cognitive flaw, gives a vivid experimental illustration, and correctly identifies why verbal knowledge of the flaw is insufficient to fix it. At under 900 words it packs more actionable epistemology than many book-length treatments.
这篇短文是 Yudkowsky 在「序列」中最具实践价值的文章之一：它分离出一个具体的、可演示的认知缺陷，给出生动的实验性说明，并正确指出为何对该缺陷的语言层面了解不足以修复它。不到 900 词，却比许多整本书的论述包含更多可操作的认识论内容。
- Concretely falsifiable demonstration具体可验证的演示The 2-4-6 task is not a thought experiment — it's a real, replicable study with a hard number (21% success). This grounds the essay in evidence rather than introspection.2-4-6 任务不是思想实验——它是一项真实的、可重复的研究，有一个确切的数字（21% 成功率）。这让文章立足于证据，而非内省。
- Correctly identifies the sub-verbal nature of the bias正确识别出偏差的语言之下性质By noting the task is emotionally 'cold' yet still produces the error, Yudkowsky pinpoints why the flaw is hard to fix: it's not motivated reasoning, it's automatic imagery. This is a genuine theoretical contribution.通过指出该任务在情感上是「冷」的却仍然产生错误，Yudkowsky 准确定位了为何这个缺陷难以修复：它不是动机性推理，而是自动意象。这是一个真正的理论贡献。
- Recursive self-application in the closing结尾处的递归自我应用The final question — did you look toward light or darkness? — turns the essay's lesson back on the reader in real time, making the point experientially rather than just propositionally.最后那个问题——你看向了光明还是黑暗？——实时将文章的教训反向作用于读者，以经验而非命题的方式传达要点。
- Strong connection to hypothesis testing与假设检验的有力联系Linking positive bias to the broader principle that theory strength lies in what it cannot explain elevates the essay from a psychological curiosity to a core epistemological point.将正向偏差与「理论强度在于它不能解释什么」这一更宏观原则联系起来，将文章从一个心理学趣闻提升为核心认识论要点。
- The cure is underspecified修复方案过于模糊"Train yourself to flinch toward the zero" is directionally correct but tells the reader almost nothing about how to perform this retraining. The essay diagnoses the sub-verbal problem, then prescribes a sub-verbal solution without a practical method.「训练自己朝向零点退缩」方向上是正确的，但几乎没有告诉读者如何进行这种重新训练。文章诊断了语言之下的问题，然后开出了一个语言之下的解方，却没有给出实际方法。
- Overstates how representative the 2-4-6 task is高估了 2-4-6 任务的代表性The 2-4-6 task is a highly stylized lab setting. Real-world hypothesis testing involves costs, time, and prior knowledge that change the optimal search strategy. Treating the lab result as a universal instinct may overstate the scope of the problem.2-4-6 任务是一个高度程式化的实验室场景。现实世界中的假设检验涉及成本、时间和先验知识，这些都会改变最优搜索策略。将实验室结果视为普遍本能，可能高估了问题的范围。
- Positive bias vs. rational prior updating正向偏差 vs. 理性先验更新Seeking positive examples is not always irrational: if a hypothesis is very unlikely, confirming instances are more informative than disconfirming ones. The essay doesn't acknowledge that the balance between positive and negative search depends on prior probability and search costs.寻求正面例子并不总是非理性的：如果一个假设非常不可能，确认性实例比否定性实例更具信息量。文章没有承认正负向搜索之间的平衡取决于先验概率和搜索成本。
- The phlogiston example is underdeveloped燃素说例子发展不足Phlogiston is used to illustrate a theory that can explain any outcome — but the example is briefly stated and relies on the reader already knowing enough chemistry history to see that the theory really was that flexible. Readers unfamiliar with phlogiston may not grasp the point.燃素说被用来说明一个能解释任何结果的理论——但这个例子陈述得很简短，依赖于读者已经足够了解化学史才能明白该理论真的有那么灵活。不熟悉燃素说的读者可能无法抓住要点。
A compact, well-evidenced essay that correctly identifies positive bias as a distinct and sub-verbal phenomenon, and persuasively links it to the broader epistemological principle that a theory's strength is what it cannot explain. Its weakness is practical: it diagnoses the problem clearly but provides little guidance on how to actually fix the reflex. Essential reading, but pair it with deliberate practice — not just comprehension.
这是一篇简洁、有据可查的文章，正确地将正向偏差识别为一种独特的、语言之下的现象，并有说服力地将其与更宏观的认识论原则联系起来：理论的强度在于它不能解释什么。其弱点在于实践层面：它对问题的诊断清晰，但几乎没有提供如何真正修复这一反射的指导。必读，但需配合刻意练习——而非仅仅理解。
Original Text原文
I am teaching a class, and I write upon the blackboard three numbers: 2-4-6. “I am thinking of a rule,” I say, “which governs sequences of three numbers. The sequence 2-4-6, as it so happens, obeys this rule. Each of you will find, on your desk, a pile of index cards. Write down a sequence of three numbers on a card, and I’ll mark it ‘Yes’ for fits the rule, or ‘No’ for not fitting the rule. Then you can write down another set of three numbers and ask whether it fits again, and so on. When you’re confident that you know the rule, write down the rule on a card. You can test as many triplets as you like.”
Here’s the record of one student’s guesses:
| 4-6-2 | No | | --- | --- | | 4-6-8 | Yes | | 10-12-14 | Yes |
At this point the student wrote down their guess at the rule. What do you think the rule is? Would you have wanted to test another triplet, and if so, what would it be? Take a moment to think before continuing.
The challenge above is based on a classic experiment due to Peter Wason, the 2-4-6 task. Although subjects given this task typically expressed high confidence in their guesses, only 21% of the subjects successfully guessed the experimenter’s real rule, and replications since then have continued to show success rates of around 20%.
The study was called “On the failure to eliminate hypotheses in a conceptual task.” Subjects who attempt the 2-4-6 task usually try to generate positive examples, rather than negative examples—they apply the hypothetical rule to generate a representative instance, and see if it is labeled “Yes.”
Thus, someone who forms the hypothesis “numbers increasing by two” will test the triplet 8-10-12, hear that it fits, and confidently announce the rule. Someone who forms the hypothesis X-2X-3X will test the triplet 3-6-9, discover that it fits, and then announce that rule.
In every case the actual rule is the same: the three numbers must be in ascending order.
But to discover this, you would have to generate triplets that shouldn’t fit, such as 20-23-26, and see if they are labeled “No.” Which people tend not to do, in this experiment. In some cases, subjects devise, “test,” and announce rules far more complicated than the actual answer.
This cognitive phenomenon is usually lumped in with “confirmation bias.” However, it seems to me that the phenomenon of trying to test positive rather than negative examples, ought to be distinguished from the phenomenon of trying to preserve the belief you started with. “Positive bias” is sometimes used as a synonym for “confirmation bias,” and fits this particular flaw much better.
It once seemed that phlogiston theory could explain a flame going out in an enclosed box (the air became saturated with phlogiston and no more could be released). But phlogiston theory could just as well have explained the flame not going out. To notice this, you have to search for negative examples instead of positive examples, look into zero instead of one; which goes against the grain of what experiment has shown to be human instinct.
For by instinct, we human beings only live in half the world.
One may be lectured on positive bias for days, and yet overlook it in-the-moment. Positive bias is not something we do as a matter of logic, or even as a matter of emotional attachment. The 2-4-6 task is “cold,” logical, not affectively “hot.” And yet the mistake is sub-verbal, on the level of imagery, of instinctive reactions. Because the problem doesn’t arise from following a deliberate rule that says “Only think about positive examples,” it can’t be solved just by knowing verbally that “We ought to think about both positive and negative examples.” Which example automatically pops into your head? You have to learn, wordlessly, to zag instead of zig. You have to learn to flinch toward the zero, instead of away from it.
I have been writing for quite some time now on the notion that the strength of a hypothesis is what it can’t explain, not what it can—if you are equally good at explaining any outcome, you have zero knowledge. So to spot an explanation that isn’t helpful, it’s not enough to think of what it does explain very well—you also have to search for results it couldn’t explain, and this is the true strength of the theory.
So I said all this, and then I challenged the usefulness of “emergence” as a concept. One commenter cited superconductivity and ferromagnetism as examples of emergence. I replied that non-superconductivity and non-ferromagnetism were also examples of emergence, which was the problem. But be it far from me to criticize the commenter! Despite having read extensively on “confirmation bias,” I didn’t spot the “gotcha” in the 2-4-6 task the first time I read about it. It’s a subverbal blink-reaction that has to be retrained. I’m still working on it myself.
So much of a rationalist’s skill is below the level of words. It makes for challenging work in trying to convey the Art through words. People will agree with you, but then, in the next sentence, do something subdeliberative that goes in the opposite direction. Not that I’m complaining! A major reason I’m writing this is to observe what my words haven’t conveyed.
Are you searching for positive examples of positive bias right now, or sparing a fraction of your search on what positive bias should lead you to not see? Did you look toward light or darkness?
我正在上一堂课，在黑板上写下三个数字：2-4-6。“我心里有一条规则，”我说，“支配着三个数字的序列。2-4-6 这个序列，恰好满足这条规则。你们每个人的桌上都有一叠索引卡。在卡上写下一个三个数字的序列，我会标注「是」（符合规则）或「否」（不符合规则）。然后你可以再写下另一组三个数字来问是否符合，如此类推。当你有把握知道规则时，把规则写在一张卡上。你可以测试任意多个三元组。”
以下是一个学生的猜测记录：
| 4-6-2 | 否 | | --- | --- | | 4-6-8 | 是 | | 10-12-14 | 是 |
到这里，学生写下了对规则的猜测。你认为规则是什么？你会想再测试另一个三元组吗？如果会，会测试什么？在继续阅读之前，花一点时间思考一下。
上面这个挑战基于彼得·沃森的一个经典实验，即 2-4-6 任务。尽管参与该任务的受试者通常对自己的猜测表达了高度自信，却只有 21% 的受试者成功猜出了实验者的真正规则，此后的复现研究持续显示约 20% 的成功率。
该研究题为《论概念任务中排除假设的失败》。尝试 2-4-6 任务的受试者通常试图生成正面例子，而不是负面例子——他们将假设中的规则应用于生成一个代表性实例，然后看它是否被标注为「是」。
因此，形成「每次增加二的数字」这一假设的人，会测试三元组 8-10-12，听到它符合，然后自信地宣布规则。形成 X-2X-3X 假设的人，会测试三元组 3-6-9，发现它符合，然后宣布那条规则。
在每种情况下，真正的规则都是同一个：这三个数字必须是升序的。
但要发现这一点，你必须生成那些不应该符合的三元组，比如 20-23-26，看它们是否被标注为「否」。而在这个实验中，人们往往不这样做。在某些情况下，受试者设计、「测试」并宣布的规则，远比真正的答案复杂得多。
这种认知现象通常被归入「确认偏差」。然而，在我看来，试图测试正面而非负面例子这一现象，应当与试图维护自己最初持有的信念这一现象区分开来。「正向偏差」有时被用作「确认偏差」的同义词，但它更贴切地描述了这个特定的缺陷。
曾经，燃素理论似乎可以解释火焰在封闭容器中熄灭（空气被燃素饱和，无法再释放更多燃素）。但燃素理论同样可以解释火焰不熄灭。要注意到这一点，你必须搜寻负面例子而不是正面例子，看向零而不是一；而这与实验所揭示的人类本能背道而驰。
因为从本能上讲，我们人类只活在世界的一半。
一个人可能被讲授正向偏差好几天，却仍然在当下忽视它。正向偏差不是我们在逻辑层面或情感依附层面做的事情。2-4-6 任务在逻辑上是「冷」的，在情感上也并不「热」。然而这个错误是语言之下的，在意象的层面、在本能反应的层面发生。因为问题并非源于某条说「只想正面例子」的刻意规则，所以它无法仅凭语言上知道「我们应该同时想到正面和负面例子」就得以解决。哪个例子会自动蹦入你的脑海？你必须在无言中学会该向右时向左。你必须学会朝向零点退缩，而不是逃离它。
我已经写了相当长一段时间，论述一个假设的强度在于它不能解释什么，而非它能解释什么——如果你对任何结果都同样擅长解释，你就拥有零知识。所以，要发现一个无益的解释，光想它能很好地解释什么是不够的——你还必须搜寻它不能解释的结果，这才是理论真正的强度所在。
我说了所有这些，然后我质疑了「涌现」作为一个概念的有用性。一位评论者援引超导性和铁磁性作为涌现的例子。我回复说，非超导性和非铁磁性也是涌现的例子，而这正是问题所在。但请勿由此批评那位评论者！尽管我已广泛阅读了关于「确认偏差」的文章，在我第一次读到 2-4-6 任务时，我也没有发现其中的「玄机」。这是一种需要重新训练的语言之下的瞬间反应。我自己仍在努力中。
理性主义者如此多的技能都在语言层面之下。这让试图通过文字来传授这门艺术的工作充满挑战。人们会同意你的观点，但随后，在下一句话里，就会做出某个潜意识的动作，朝相反的方向走去。我并不是在抱怨！我写这些文章的一个重要原因，就是观察我的文字没有传递出什么。
你现在是在搜寻正向偏差的正面例子，还是也在花一部分搜寻力去找正向偏差本该让你看不见的东西？你看向了光明还是黑暗？