Concise Summary简洁概述
Yudkowsky distinguishes two kinds of rationality: epistemic (building accurate maps of the world) and instrumental (choosing actions that achieve your goals). Both have formal foundations — probability theory and decision theory — that provide an objective standard for evaluating beliefs and decisions. Popular images of rationality, like Mr. Spock, miss the point entirely. But even Bayesian formalism has limits: it is computationally intractable in real life and faces open questions in edge cases. So rationality also requires an art — understanding your own cognitive flaws, overcoming biases, and working from inside a finite human mind. The word itself is secondary; what matters is the underlying question of what is a good way to think.
Yudkowsky 区分了两种理性：知识论理性（建构关于世界的准确地图）与工具理性（选择能实现目标的行动）。两者都有形式化基础——概率论与决策论——为评估信念和决策提供客观标准。斯波克先生式的流行理性形象完全错失了要点。但即便是贝叶斯形式体系也有局限：在现实生活中计算上不可行，在边缘案例中也面临悬而未决的问题。因此，理性还需要一种艺术——理解自身的认知缺陷、克服偏差，从一个有限的人类心智内部出发。这个词本身是次要的；真正重要的是其背后的问题：什么是好的思维方式。
Infographic信息图
Epistemic rationality: accurate maps
知识论理性：准确的地图
It's about building a mental model that corresponds to reality — the correspondence between belief and reality is simply called truth.
它关乎建构与现实相对应的心智模型——信念与现实之间的对应，就简单地叫做真理。
Instrumental rationality: winning
工具理性：赢
It's the art of steering reality — choosing actions that lead to outcomes ranked higher in your preferences, including what you value about other people.
这是驾驭现实的艺术——选择能带来更符合你偏好之结果的行动，包括你所关心的一切，也包括他人。
Math as the gold standard
数学作为黄金标准
Psychologists use probability theory and decision theory to objectively label reasoning errors — like the conjunction fallacy — as incorrect, not just unusual.
心理学家使用概率论与决策论来客观地将推理错误——如合取谬误——标注为不正确，而非只是不寻常。
Spock is not rational
斯波克并非理性
Popular culture's image of rationality (cold verbal logic) is orthogonal to Bayesian rationality, which fully encompasses urges, perceptions, and wordless intuitions.
流行文化的理性形象（冷酷的言语逻辑）与贝叶斯理性正交，后者完全涵盖冲动、感知和无言的直觉。
Formalism + art
形式体系 + 艺术
Because full Bayesian computation is intractable for humans, rationality also requires an art: learning your biases, avoiding self-deception, and thinking clearly under real conditions.
由于完整的贝叶斯计算对人类而言不可行，理性还需要一种艺术：了解自己的偏差、避免自我欺骗，在真实条件下清晰思考。
Detailed Summary详细概述
Two Definitions
Yudkowsky opens by stating he means two things by "rationality." Epistemic rationality is the systematic improvement of belief accuracy — building maps that correspond to the territory. A false belief is like a map that says there's a bookcase where none exists; when you go to get a book, reality disappoints. This correspondence between belief and world is simply called truth. Instrumental rationality is steering reality — choosing actions that lead to outcomes ranked higher in your preferences. Yudkowsky calls this "winning," with the caveat that winning doesn't mean winning at others' expense, since our values include everything we care about.
Why a Special Word?
A critic might ask: why say "rational" when you could say "true" or "good"? Yudkowsky's answer mirrors the defense of the concept of truth itself. "Snow is white" doesn't need the word true — but "true models generally produce better predictions than false models" does, because it generalizes across all map-territory relationships. Similarly, "Rational agents maximize the probabilistic expectation of a coherent utility function" captures something that "it's useful to eat vegetables" cannot. We need rational to describe systematic features of thinking that reliably produce truth or value.
The Mathematical Gold Standards
Experimental psychologists need an objective vantage point to call a judgment wrong — not just unusual. The tools are probability theory (the laws of rational belief) and decision theory (the laws of rational action). These apply universally, whether you're locating a bookcase or estimating how many hairs Julius Caesar had. The conjunction fallacy — rating P(Bill plays jazz) as lower than P(Bill is an accountant who plays jazz) — is demonstrably incorrect by P(A) ≥ P(A,B). Beliefs and decisions conforming to these formalisms are called "Bayesian."
"Bayesian" here doesn't mean cold and robotic. Footnote 2 stresses that Bayesian rationality fully applies to urges, hunches, and wordless intuitions — not only to verbal assertions.
Limits of Formalism
Yudkowsky is honest about two major limits. First, full Bayesian computation is computationally intractable on real-world problems — no one can actually calculate the math, any more than you can predict the stock market from quark movements. This is why a whole site called Less Wrong exists rather than a one-page statement of axioms. There is an entire further art to finding truth from inside a human mind: learning your biases, overcoming motivated reasoning, keeping yourself emotionally honest.
Second, the formal rules themselves are sometimes called into question — anthropic problems unsettle probability theory; Newcomb-like problems unsettle standard decision theory.
In such edge cases, Yudkowsky recommends returning to simpler, pre-formal ideas like "truth" and "winning." Like a scientist pointing to a campfire rather than invoking phlogiston, you shouldn't ignore something just because you can't define it precisely. Cognitive biases hurt regardless of whether we can compactly define "irrationality."
The Real Question
The essay ends by deflating the word itself. Trying to "win" a definitional debate about "rationality" is futile; definitions communicate, they don't settle substantive questions. If replacing "rational" with the nonsense word "foozal" throughout the essay changes its force, then the argument was resting on connotations, not substance. What matters is the underlying question: what is a good way to think? Use "rational" when it communicates clearly, but never use it as a mere incantation.
两个定义
Yudkowsky 开篇就说明，他所谓的「理性」有两层含义。知识论理性是对信念准确性的系统性提升——建构与疆域相对应的地图。一个错误的信念就像一张标注了书架却根本不存在书架的地图；当你走过去拿书时，现实会让你失望。信念与世界之间的这种对应，就简单地称为真理。工具理性是驾驭现实——选择能带来更符合你偏好之结果的行动。Yudkowsky 把这称为「赢」，并附带说明：赢并不意味着以牺牲他人为代价，因为我们所有关心的事物——包括他人——都属于我们的价值范畴。
为什么需要一个特殊的词？
批评者也许会问：既然可以说「真」或「好」，为何还要说「理性」？Yudkowsky 的回答与捍卫「真」这一概念的逻辑如出一辙。「雪是白的」不需要「真」这个词——但「真实的模型通常比虚假的模型产生更好的预测」就需要，因为它将所有地图-疆域关系都纳入了一个普遍化的陈述。类似地，「理性主体最大化相干效用函数的概率期望」捕捉到了「吃蔬菜是有益的」所无法表达的东西。我们需要「理性」来描述那些系统性地产生真理或价值的思维特征。
数学黄金标准
实验心理学家需要一个客观的制高点来将某个判断标注为错误——而非只是不寻常。这些工具就是概率论（理性信念的法则）与决策论（理性行动的法则）。它们的适用范围是普遍的，无论你是在定位书架还是估算恺撒大帝的发量。合取谬误——认为 P（比尔演奏爵士乐）低于 P（比尔是会演奏爵士乐的会计师）——根据 P(A) ≥ P(A,B) 这一定律，被证明是明确错误的。符合这些形式体系的信念与决策，被称为「贝叶斯式的」。
「贝叶斯式」并不意味着冷酷机械。脚注 2 强调，贝叶斯理性完全适用于冲动、直觉和无言的感受——而不仅仅适用于言语断言。
形式体系的局限
Yudkowsky 坦诚地指出两个主要局限。首先，完整的贝叶斯计算在现实问题中计算上不可行——没有人能真正算出那些数学，就像你无法通过计算夸克的运动来预测股市。这正是存在整个《少有人走的路》网站、而非仅仅一页公理声明的原因。从人类心智内部找到真理，还需要一整套额外的艺术：了解自己的偏差、克服动机性推理、保持情感上的诚实。
其次，形式规则本身有时也受到质疑——人择问题撼动了概率论；纽科姆式问题撼动了标准决策论。
在这类边缘情况下，Yudkowsky 建议回归更简单的前形式化概念，如「真理」与「赢」。就像一个科学家指向营火而非援引燃素说，你不应该仅仅因为无法精确定义某样东西就忽视它。认知偏差造成的伤害，不会因为我们无法简洁地定义「非理性」而减少分毫。
真正的问题
文章最后贬低了这个词本身的地位。试图在「理性」的定义之争中「取胜」是徒劳的；定义用来沟通，不能解决实质性问题。如果把文章中每一处「理性」替换成无意义的词「福扎尔」会改变文章的力度，那说明论证依赖的是词语的内涵，而非实质。真正重要的是背后的问题：什么是好的思维方式？ 当「理性」能清晰沟通时就用它，但绝不要把它当作一种空洞的咒语。
FAQ常见问答
What is the difference between epistemic and instrumental rationality?知识论理性与工具理性的区别是什么？
Epistemic rationality is about believing accurately — making your mental map match the territory. Instrumental rationality is about acting effectively — choosing actions that lead to outcomes you prefer. Both are aspects of a single underlying drive toward truth and winning, but the distinction matters because you can be instrumentally savvy while holding false beliefs, or vice versa.
知识论理性关乎准确地相信——让你的心智地图与疆域相符。工具理性关乎有效地行动——选择能带来你所偏好之结果的行动。两者都是同一底层驱动——追求真理与赢——的不同面向，但这一区分是重要的：你可以在持有错误信念的同时在工具上很精明，反之亦然。
Why is the conjunction fallacy useful as an example here?为什么合取谬误在这里是一个有用的例子？
It illustrates why we need a formal gold standard. Without probability theory, we might just say the judgment is "weird" or "counterintuitive." With it, we can say it is provably wrong: P(A) ≥ P(A,B) is a universal law, and the judgment violates it. This shows how mathematical formalisms give rationality objective teeth.
它说明了为何我们需要一个形式化的黄金标准。没有概率论，我们也许只能说这个判断「奇怪」或「反直觉」。有了概率论，我们就能说它经过证明是错误的：P(A) ≥ P(A,B) 是一条普遍定律，而该判断违反了它。这展示了数学形式体系是如何赋予理性客观牙齿的。
Why isn't Spock-like behavior rational?为什么斯波克式的行为不是理性的？
Spock's popular-culture image of rationality — cold verbal reasoning, suppressing emotions — conflates rationality with a specific style of human personality. Bayesian rationality is indifferent to style; it applies equally to gut feelings, emotional responses, and wordless intuitions. What matters is whether your beliefs and actions systematically track truth and value, not whether you look analytical.
斯波克的流行文化理性形象——冷酷的言语推理、压制情感——把理性与人类个性的某种特定风格混为一谈。贝叶斯理性对风格漠不关心；它同样适用于直觉、情感反应和无言的感受。重要的是你的信念和行动是否系统性地追踪真理与价值，而非你是否看起来很有分析范儿。
If Bayesian formalism is computationally intractable, why bother with it at all?如果贝叶斯形式体系在计算上不可行，为什么还要去理会它？
Because it provides the correct normative standard — it tells us what perfect reasoning would look like, which lets us measure and diagnose our deviations from it. We don't need to compute Bayes exactly to benefit from knowing in which direction our biases tend to push us. The art of rationality is navigating the gap between the ideal and what a human mind can actually do.
因为它提供了正确的规范性标准——它告诉我们完美的推理是什么样子，从而让我们能够衡量并诊断自己的偏差。我们不需要精确地计算贝叶斯，就能从了解我们的偏差倾向于把我们推向哪个方向中获益。理性的艺术，正是在理想与人类心智实际能做到的事之间穿行。
What does Yudkowsky mean by saying the word 'rational' is secondary?Yudkowsky 说「理性」这个词是次要的，是什么意思？
He means that definitional debates — arguing over what 'rational' means — cannot resolve the underlying substantive question of how to think well. Words communicate; they don't settle matters. If you replace 'rational' with a nonsense word throughout the essay and the argument still works, then it was never about the word. The real question is: what is a good way to think?
他的意思是，关于「理性」意味着什么的定义之争，无法解决「如何好好思考」这一底层实质性问题。词语用来沟通，不能解决问题。如果把文章里的「理性」全部替换成一个无意义的词，论证仍然成立，那么争论就从来不是关于这个词的。真正的问题是：什么是好的思维方式？
What is the relevance of Newcomb's Problem and anthropic problems?纽科姆问题和人择问题有什么相关性？
They are edge cases where our best formal theories — probability theory and decision theory — either give counterintuitive answers or genuinely disagree among competing formalizations. Yudkowsky mentions them to show that even the mathematical gold standards have open questions, which is a further reason the full "art" of rationality cannot be reduced to simply following the math.
它们是边缘案例，在这些情况下，我们最好的形式化理论——概率论与决策论——要么给出反直觉的答案，要么在相互竞争的形式化方案之间产生真实的分歧。Yudkowsky 提到它们，是为了说明即便是数学黄金标准也有悬而未决的问题，这进一步说明了为什么完整的理性「艺术」不能简化为单纯地遵循数学。
In-depth Analysis · Pros & Cons深入解读 · 优缺点
This essay serves as the definitional anchor for the entire Rationality: A-Z sequence. By distinguishing epistemic from instrumental rationality, grounding both in formal theory, and then honestly acknowledging those theories' limits, it stakes out a position that is neither naive scientism nor defeatist relativism.
这篇文章是整个《理性：从A到Z》系列的定义锚点。通过区分知识论理性与工具理性、将两者奠基于形式化理论之上，然后诚实地承认这些理论的局限，它确立了一个既非天真的科学主义、也非失败主义相对主义的立场。
- Clean, memorable two-part definition清晰易记的两部分定义Splitting rationality into epistemic and instrumental gives readers a durable conceptual scaffold that genuinely carves the problem space at its joints.将理性分为知识论理性与工具理性，为读者提供了一个经久耐用的概念脚手架，真正地在问题空间的关节处进行了切割。
- Mathematical grounding without mathematical gatekeeping有数学基础却无数学门槛Citing probability theory and decision theory as the gold standard elevates the discussion above mere opinion, yet Yudkowsky doesn't require readers to know the math — just to accept that it exists and applies.援引概率论与决策论作为黄金标准，将讨论提升到单纯意见之上，但 Yudkowsky 并不要求读者了解这些数学——只需接受它们存在且适用即可。
- Honest about the limits of formalism对形式体系的局限诚实The acknowledgment that Bayesian computation is intractable and that edge cases remain open prevents the essay from over-promising. It sets up the rest of the sequence rather than claiming to finish the job.承认贝叶斯计算不可行且边缘案例尚未解决，防止了文章过度承诺。它为系列后续篇章做好铺垫，而非声称已经大功告成。
- The 'foozal' deflation is rhetorically elegant「福扎尔」的消气手法修辞上优雅Replacing 'rational' with a nonsense word to test whether the argument still works is a compact, memorable device for teaching the difference between substance and connotation.用无意义词替换「理性」来测试论证是否仍然成立，是一个简洁易记的装置，用来传授实质与内涵之间的区别。
- The epistemic/instrumental distinction is underexplored知识论/工具理性的区分探讨不足The two types are introduced and named but their interactions — e.g., how epistemic errors compound into bad decisions, or how goal-setting itself may require epistemic input — are left for later essays. Readers may leave with the impression the two are fully independent.两种理性被引入并命名，但它们之间的交互——例如知识论错误如何复合为糟糕的决策，或目标设定本身如何可能需要知识论投入——留给了后续文章。读者可能会带着两者完全独立的印象离开。
- The formal gold standards are invoked rather than justified形式黄金标准被援引而非被证明Probability theory and decision theory are presented as authoritative without explaining why they are the right norms — their axiomatic derivations (Dutch Book, Cox's theorem, etc.) are left entirely implicit. Readers inclined to question the standards have little to hold onto.概率论与决策论被当作权威呈现，却没有解释为什么它们是正确的规范——它们的公理推导（荷兰赌论证、考克斯定理等）完全是隐含的。倾向于质疑这些标准的读者几乎找不到可以抓住的东西。
- The treatment of instrumental rationality is narrower than claimed工具理性的处理比宣称的更狭窄Saying that winning 'includes everything we care about, including other people' gestures at altruism but doesn't address the hard cases where self-interest and other-regarding values genuinely conflict — or where the structure of 'preferences' is itself contested.说赢「包括我们关心的一切，包括他人」暗示了利他主义，但并未处理自身利益与关照他人的价值真正冲突的困难情形——也没有处理「偏好」结构本身受到质疑的情况。
- Open problems in decision theory are mentioned but not engaged决策论中的开放问题被提及却未被正视Newcomb's Problem is cited in a footnote as unsettling decision theory, yet the body of the essay treats decision theory as settled enough to serve as a gold standard. This creates a tension the essay does not resolve.纽科姆问题在脚注中被引用为撼动决策论的案例，然而文章正文却将决策论视为足够稳固的黄金标准。这造成了一个文章本身并未化解的张力。
A necessary and well-executed definitional essay that grounds the sequence in formal theory while staying honest about the theory's limits. Its main risk is making rationality sound more settled than it is — the formal apparatus is invoked as an authority before its own foundations have been examined. Read it as a useful first map, not a finished picture.
一篇必要且执行良好的定义性文章，将系列奠基于形式化理论之上，同时对理论的局限保持诚实。其主要风险是让理性听起来比实际上更为确定——形式化装置在其自身基础尚未被审视之前就被援引为权威。把它当作一张有用的初始地图来读，而非一幅完成的图景。
Original Text原文
I mean two things:
1\. Epistemic rationality: systematically improving the accuracy of your beliefs.
2\. Instrumental rationality: systematically achieving your values.
The first concept is simple enough. When you open your eyes and look at the room around you, you’ll locate your laptop in relation to the table, and you’ll locate a bookcase in relation to the wall. If something goes wrong with your eyes, or your brain, then your mental model might say there’s a bookcase where no bookcase exists, and when you go over to get a book, you’ll be disappointed.
This is what it’s like to have a false belief, a map of the world that doesn’t correspond to the territory. Epistemic rationality is about building accurate maps instead. This correspondence between belief and reality is commonly called “truth,” and I’m happy to call it that.^1^
Instrumental rationality, on the other hand, is about steering reality—sending the future where you want it to go. It’s the art of choosing actions that lead to outcomes ranked higher in your preferences. I sometimes call this “winning.”
So rationality is about forming true beliefs and making decisions that help you win.
(Where truth doesn't mean “certainty,” since we can do plenty to increase the probability that our beliefs are accurate even though we're uncertain; and winning doesn't mean “winning at others' expense,” since our values include everything we care about, including other people.)
When people say “X is rational!” it’s usually just a more strident way of saying “I think X is true” or “I think X is good.” So why have an additional word for “rational” as well as “true” and “good”?
An analogous argument can be given against using “true.” There is no need to say “it is true that snow is white” when you could just say “snow is white.” What makes the idea of truth useful is that it allows us to talk about the general features of map-territory correspondence. “True models usually produce better experimental predictions than false models” is a useful generalization, and it’s not one you can make without using a concept like “true” or “accurate.”
Similarly, “Rational agents make decisions that maximize the probabilistic expectation of a coherent utility function” is the kind of thought that depends on a concept of (instrumental) rationality, whereas “It’s rational to eat vegetables” can probably be replaced with “It’s useful to eat vegetables” or “It’s in your interest to eat vegetables.” We need a concept like “rational” in order to note general facts about those ways of thinking that systematically produce truth or value—and the systematic ways in which we fall short of those standards.
As we’ve observed in the previous essays, experimental psychologists sometimes uncover human reasoning that seems very strange. For example, someone rates the probability “Bill plays jazz” as less than the probability “Bill is an accountant who plays jazz.” This seems like an odd judgment, since any particular jazz-playing accountant is obviously a jazz player. But to what higher vantage point do we appeal in saying that the judgment is wrong ?
Experimental psychologists use two gold standards: probability theory, and decision theory.
Probability theory is the set of laws underlying rational belief. The mathematics of probability applies equally to “figuring out where your bookcase is” and “estimating how many hairs were on Julius Caesars head,” even though our evidence for the claim “Julius Caesar was bald” is likely to be more complicated and indirect than our evidence for the claim “theres a bookcase in my room.” It’s all the same problem of how to process the evidence and observations to update one’s beliefs. Similarly, decision theory is the set of laws underlying rational action, and is equally applicable regardless of what one’s goals and available options are.
Let “P(such-and-such)” stand for “the probability that such-and-such happens,” and “P(A,B)” for “the probability that both A and B happen.” Since it is a universal law of probability theory that P(A) ≥ P(A,B), the judgment that P(Bill plays jazz) is less than P(Bill plays jazz, Bill is an accountant) is labeled incorrect.
To keep it technical, you would say that this probability judgment is non-Bayesian. Beliefs that conform to a coherent probability distribution, and decisions that maximize the probabilistic expectation of a coherent utility function, are called “Bayesian.”
I should emphasize that this isn't the notion of rationality thats common in popular culture. People may use the same string of sounds, “ra-tio-nal,” to refer to “acting like Mr. Spock of Star Trek” and “acting like a Bayesian”; but this doesn't mean that acting Spock-like helps one hair with epistemic or instrumental rationality.^2^
All of this does not quite exhaust the problem of what is meant in practice by “rationality,” for two major reasons:
First, the Bayesian formalisms in their full form are computationally intractable on most real-world problems. No one can actually calculate and obey the math, any more than you can predict the stock market by calculating the movements of quarks.
This is why there is a whole site called “Less Wrong,” rather than a single page that simply states the formal axioms and calls it a day. There’s a whole further art to finding the truth and accomplishing value from inside a human mind: we have to learn our own flaws, overcome our biases, prevent ourselves from self-deceiving, get ourselves into good emotional shape to confront the truth and do what needs doing, et cetera, et cetera.
Second, sometimes the meaning of the math itself is called into question. The exact rules of probability theory are called into question by, e.g., anthropic problems in which the number of observers is uncertain. The exact rules of decision theory are called into question by, e.g., Newcomblike problems in which other agents may predict your decision before it happens.^3^
In cases where our best formalizations still come up short, we can return to simpler ideas like “truth” and “winning.” If you are a scientist just beginning to investigate fire, it might be a lot wiser to point to a campfire and say “Fire is that orangey-bright hot stuff over there,” rather than saying “I define fire as an alchemical transmutation of substances which releases phlogiston.” You certainly shouldn’t ignore something just because you can’t define it. I can't quote the equations of General Relativity from memory, but nonetheless if I walk off a cliff, I'll fall. And we can say the same of cognitive biases and other obstacles to truth—they won't hit any less hard if it turns out we can't define compactly what “irrationality” is.
In cases like these, it is futile to try to settle the problem by coming up with some new definition of the word “rational” and saying, “Therefore my preferred answer, by definition, is what is meant by the word ‘rational.’ ” This simply raises the question of why anyone should pay attention to your definition. I’m not interested in probability theory because it is the holy word handed down from Laplace. I’m interested in Bayesian-style belief-updating (with Occam priors) because I expect that this style of thinking gets us systematically closer to, you know, accuracy, the map that reflects the territory.
And then there are questions of how to think that seem not quite answered by either probability theory or decision theory—like the question of how to feel about the truth once you have it. Here, again, trying to define “rationality” a particular way doesn’t support an answer, but merely presumes one.
I am not here to argue the meaning of a word, not even if that word is “rationality.” The point of attaching sequences of letters to particular concepts is to let two people communicate—to help transport thoughts from one mind to another. You cannot change reality, or prove the thought, by manipulating which meanings go with which words.
So if you understand what concept I am generally getting at with this word “rationality,” and with the sub-terms “epistemic rationality” and “instrumental rationality,” we have communicated: we have accomplished everything there is to accomplish by talking about how to define “rationality.” What’s left to discuss is not what meaning to attach to the syllables “ra-tio-na-li-ty”; what’s left to discuss is what is a good way to think.
If you say, “It’s (epistemically) rational for me to believe X, but the truth is Y,” then you are probably using the word “rational” to mean something other than what I have in mind. (E.g., “rationality” should be consistent under reflection—“rationally” looking at the evidence, and “rationally” considering how your mind processes the evidence, shouldn’t lead to two different conclusions.)
Similarly, if you find yourself saying, “The (instrumentally) rational thing for me to do is X, but the right thing for me to do is Y,” then you are almost certainly using some other meaning for the word “rational” or the word “right.” I use the term “rationality” normatively, to pick out desirable patterns of thought.
In this case—or in any other case where people disagree about word meanings—you should substitute more specific language in place of “rational”: “The self-benefiting thing to do is to run away, but I hope I would at least try to drag the child off the railroad tracks,” or “Causal decision theory as usually formulated says you should two-box on Newcomb’s Problem, but I’d rather have a million dollars.”
In fact, I recommend reading back through this essay, replacing every instance of “rational” with “foozal,” and seeing if that changes the connotations of what I’m saying any. If so, I say: strive not for rationality, but for foozality.
The word “rational” has potential pitfalls, but there are plenty of non-borderline cases where “rational” works fine to communicate what I’m getting at. Likewise “irrational.” In these cases I’m not afraid to use it.
Yet one should be careful not to overuse that word. One receives no points merely for pronouncing it loudly. If you speak overmuch of the Way, you will not attain it.
^1^ For a longer discussion of truth, see “The Simple Truth” at the very end of this volume.
^2^ The idea that rationality is about strictly privileging verbal reasoning over feelings is a case in point. Bayesian rationality applies to urges, hunches, perceptions, and wordless intuitions, not just to assertions.
I gave the example of opening your eyes, looking around you, and building a mental model of a room containing a bookcase against the wall. The modern idea of rationality is general enough to include your eyes and your brains visual areas as things-that-map, and to include instincts and emotions in the belief-and-goal calculus.
^3^ For an informal statement of Newcomb’s Problem, see Jim Holt, “Thinking Inside the Boxes,” Slate, 2002, http://www.slate.com/articles/arts/egghead/2002/02/thinkinginside\_the\_boxes.single.html.
我指的是两件事：
1\. 知识论理性：系统性地提升你信念的准确性。
2\. 工具理性：系统性地实现你的价值观。
第一个概念简单明了。当你睁开眼睛打量周围的房间，你会确定笔记本电脑相对于桌子的位置，确定书架相对于墙壁的位置。如果你的眼睛或大脑出了问题，你的心智模型也许会说某处有书架，而那里其实空无一物；当你走过去取书，就会失望而归。
这就是持有错误信念的感受——一张与疆域不符的世界地图。知识论理性就是去建构准确的地图。信念与现实之间的这种对应，通常被称为「真理」，我也乐于这样称呼它。^1^
另一方面，工具理性关乎的是驾驭现实——把未来送往你希望它去的地方。它是选择能带来更符合你偏好之结果的行动的艺术。我有时把这称为「赢」。
所以，理性就是形成真实的信念，并做出帮助你赢的决策。
（这里的「真实」并不意味着「确定」，因为即使在不确定的情况下，我们也能做很多事来提高信念准确的概率；而「赢」也不意味着「以牺牲他人为代价取胜」，因为我们的价值观包含了我们关心的一切，包括他人。）
当人们说「X 是理性的！」时，通常只是在更加强调地说「我认为 X 是真的」或「我认为 X 是好的」。那么，除了「真」和「好」之外，为什么还需要一个额外的词「理性」呢？
类似的论证也可以用来反对「真」这个词。「雪是白的」已经说明了一切，「雪是白的这件事是真的」并不需要。使「真理」这一概念有用的地方，在于它让我们能谈论地图-疆域对应关系的一般特征。「真实的模型通常比虚假的模型产生更好的实验预测」是一个有用的概括，而不使用「真」或「准确」这样的概念，你是无法表达这个概括的。
类似地，「理性主体做出能最大化相干效用函数的概率期望的决策」是那种依赖于（工具）理性概念的思想，而「吃蔬菜是理性的」则大概可以替换为「吃蔬菜是有益的」或「吃蔬菜符合你的利益」。我们需要一个类似「理性」的概念，才能指出那些系统性地产生真理或价值的思维方式的一般性事实——以及我们在系统性上落后于这些标准的方式。
正如我们在前几篇文章中所观察到的，实验心理学家有时会发现人类推理中一些看起来非常奇怪的现象。例如，有人认为「比尔演奏爵士乐」的概率低于「比尔是一位演奏爵士乐的会计师」的概率。这看起来是个奇怪的判断，因为任何一个演奏爵士乐的会计师显然都是爵士乐演奏者。但我们依据什么更高的立足点来说这个判断是错误的呢？
实验心理学家使用两个黄金标准：概率论和决策论。
概率论是理性信念背后的法则体系。概率论的数学既适用于「搞清楚你的书架在哪里」，也适用于「估算恺撒大帝头上有多少根头发」，尽管我们关于「恺撒大帝是秃头」的证据，很可能比「我房间里有一个书架」的证据更复杂、更间接。这都是同一个问题：如何处理证据和观察，以更新自己的信念。类似地，决策论是理性行动背后的法则体系，不论目标和可选方案是什么，它都同样适用。
让 P（如此这般）代表「如此这般发生的概率」，P（A，B）代表「A 和 B 同时发生的概率」。由于 P(A) ≥ P(A,B) 是概率论的普遍定律，因此认为 P（比尔演奏爵士乐）低于 P（比尔演奏爵士乐，比尔是会计师）的判断被标注为错误。
用专业术语来说，这个概率判断是非贝叶斯的。符合相干概率分布的信念，以及能最大化相干效用函数的概率期望的决策，被称为「贝叶斯式的」。
我必须强调，这不是流行文化中常见的那种理性概念。人们也许会用同一串发音「ra-tio-nal」来指称「像《星际迷航》中的斯波克先生那样行事」和「像贝叶斯主义者那样行事」；但这并不意味着斯波克式的行为对知识论理性或工具理性有丝毫帮助。^2^
以上这些并没有完全穷尽实践中「理性」究竟意味着什么的问题，原因有两个：
首先，贝叶斯形式体系在其完整形式上，对于大多数现实问题而言在计算上是不可行的。没有人能真正计算出那些数学，就像你无法通过计算夸克的运动来预测股市一样。
这就是为什么存在一个叫做「少有人走的路」的完整网站，而不是一个仅仅陈述形式公理就收工的单页页面。从人类心智内部去找到真理、实现价值，还有一整套进一步的艺术：我们必须学习自己的缺陷，克服我们的偏见，阻止自我欺骗，让自己保持良好的情绪状态以直面真相并做该做的事，等等，等等。
其次，有时候数学本身的含义会受到质疑。概率论的精确规则受到了例如人择问题的质疑，在这类问题中观察者的数量是不确定的。决策论的精确规则则受到了例如纽科姆式问题的质疑，在这类问题中其他主体可能在你做出决策之前就预测到你的决策。^3^
在我们最好的形式化理论仍然不够用的情况下，我们可以回归更简单的想法，比如「真理」和「赢」。如果你是一位刚开始研究火的科学家，指着一堆篝火说「火就是那边那个橘红色的、滚烫的东西」，也许比说「我将火定义为一种释放燃素的物质炼金转化」要明智得多。你当然不应该仅仅因为无法定义某样东西就忽视它。我背不出广义相对论的方程式，但如果我从悬崖上走下去，我仍然会坠落。认知偏差和其他通往真理的障碍也是如此——即便事实证明我们无法简洁地定义「非理性」是什么，它们造成的冲击也不会因此减轻一分。
在这类情况下，试图通过给出「理性」这个词的某个新定义来解决问题是徒劳的，因为这样做无非是说「因此，按照定义，我偏好的答案就是'理性'一词的含义。」这只会引出另一个问题：为什么任何人都应该关注你的定义。我对概率论感兴趣，并非因为它是拉普拉斯传下来的神圣话语。我对贝叶斯式信念更新（与奥卡姆先验结合）感兴趣，是因为我期望这种思维风格能系统性地让我们更接近——你知道的，准确性——那张映照疆域的地图。
还有一些关于如何思考的问题，似乎既不能由概率论也不能由决策论来完整回答——比如，一旦你掌握了真理，该如何感受它。在这里，同样，试图以某种特定方式定义「理性」并不能支持一个答案，而只是预设了一个答案。
我来这里不是为了争论一个词的含义，哪怕这个词是「理性」。给特定概念附上字母序列的意义在于让两个人能够沟通——帮助把思想从一个心智传递到另一个心智。你无法通过操纵哪些含义对应哪些词语来改变现实，或证明某个思想。
所以，如果你理解了我用「理性」这个词大致想表达的概念，以及「知识论理性」和「工具理性」这两个子术语，我们就已经实现了沟通：我们已经完成了谈论如何定义「理性」所能完成的一切。剩下要讨论的，不是把什么含义附在「理性」这几个音节上；剩下要讨论的，是什么是好的思维方式。
如果你说「对我来说，（知识论上）相信 X 是理性的，但真相是 Y」，那么你使用「理性」这个词，很可能意味着某种与我心目中不同的东西。（例如，「理性」应该在反思下保持一致——「理性地」审视证据，与「理性地」考量你的心智如何处理证据，不应该得出两个不同的结论。）
类似地，如果你发现自己在说「对我来说，（工具上）理性的做法是 X，但正确的做法是 Y」，那么你几乎可以肯定是在用某种其他含义使用「理性」或「正确」这两个词中的一个。我规范性地使用「理性」这个术语，来指出那些值得向往的思维模式。
在这种情况下——或任何其他人们在词义上存在分歧的情况下——你应该用更具体的语言来替代「理性」：「自利的做法是逃跑，但我希望我至少会试图把那个孩子从铁轨上拖走」，或者「通常所阐述的因果决策论说你应该在纽科姆问题上选两个盒子，但我宁愿要一百万美元。」
事实上，我建议重读这篇文章，把每一处「理性」都替换成「福扎尔」，看看这是否改变了我所说内容的内涵。如果改变了，我会说：不要追求理性，而要追求福扎性。
「理性」这个词有潜在的陷阱，但也有大量非边缘的情况，在那些情况下「理性」能很好地传达我的意思。「非理性」也是如此。在这些情况下，我不害怕使用它。
然而，人们应当小心，不要过度使用那个词。仅仅大声说出它，并不能得分。如果你谈论道路的话太多，你就无法到达那里。
^1^ 关于真理的更长讨论，请参阅本卷末尾的「简单的真理」。
^2^ 理性就是严格地将言语推理凌驾于感受之上——这一观念恰好是个典型例子。贝叶斯理性适用于冲动、直觉、感知和无言的本能，而不仅仅适用于言语断言。
我举了睁开眼睛、打量四周、在心中建立一个包含靠墙书架的房间模型的例子。现代理性观念足够普遍，足以把你的眼睛和大脑的视觉区域纳入「进行映射的东西」之列，并把本能和情感纳入信念-目标的计算之中。
^3^ 关于纽科姆问题的非正式陈述，见 Jim Holt，《在盒子里思考》，Slate，2002 年，http://www.slate.com/articles/arts/egghead/2002/02/thinkinginside\_the\_boxes.single.html。