This superb book should help set the agenda for philosophical work on causation for years to come. Indeed, its impact deserves to be felt more widely: philosophers who don’t give a fig about causation will still profit a great deal by studying this book closely. It is not faultless: in particular, Woodward’s treatment in chapter 2 of his favored “interventionist” account of causation and its alternatives may leave the uninitiated reader lost. That said, these defects do nothing to undermine the main philosophical achievement of the book, which is to articulate, defend, and provide detailed illustrations of the power of a certain methodological orientation to the philosophical study of causation that Woodward calls the “functional approach.”The word functional signals that the questions that should be of primary concern to the philosophical study of causation are questions, roughly, about the function—point, rationale, goal-serving capacity—of our concept of causation, along with corresponding questions about the kinds of worldly structures this concept answers to. (We’ll see shortly that, for reasons Woodward expertly lays out, talk of “our concept” of causation is too limiting. But if you’re a philosopher new to Woodward’s approach, it’s probably the best way to start to orient you.) Chapter 1 lays out the functional approach in the abstract, paying special attention to the ways in which it involves weaving together descriptive and normative inquiry into human causal cognition. Chapter 3 draws out a number of specific and important lessons for how inquiry into causation ought to proceed, given the functional approach; along the way, it patiently constructs devastating challenges to standard ways that philosophers rely on “intuitions” about cases, while at the same defending importantly different ways in which such intuitions can help guide inquiry. Together, these chapters are a methodological tour de force. Summarizing them adequately isn’t possible here. What I’ll do instead is to approach the main ideas obliquely, by considering a way that philosophical dialectic about causation could, but typically doesn’t, proceed.Suppose I advance a simple Mackie-style analysis of causation: event C is a cause of event E iff (1) C and E both occur; (2) C precedes E; and (3) there is some true proposition P about the circumstances in which C and E occur such that the conjunction of P, the proposition that C occurs, and the proposition L that captures the fundamental laws entails the proposition that E occurs, but the conjunction of P and L alone does not. A nonstarter! After all, it’s child’s play to produce cases where C and E meet these conditions, but where intuition judges C and E to be joint effects of some common cause. In short, this analysis conflates causation with common-cause-induced correlation, and so must be rejected. Maybe we add bells and whistles to fix the analysis; maybe we try something completely different. But either way, the analysis as it stands has been refuted.What I’ve just described is, I trust, an utterly familiar example of how philosophical back-and-forth proceeds (and not just when the subject is causation). But now consider a kind of thought experiment rarely entertained in these dialectical contexts: Imagine a community of people whose concept of causation was exactly captured by this analysis. So they don’t see any distinction between what we call “causation” and what we call “common-cause-induced correlation.” What, if anything, would be wrong with them? Put another way, why not adopt their way of thinking about causation? Those are the kinds of question Woodward thinks we should be asking. I think he is dead right.Note that to answer such questions well, we have to set aside the temptation to pound the table and say things like “They wouldn’t really be thinking about causation; that’s what’s wrong!” Woodward gives a thorough and conscientious airing (in chapters 1 and 3) to similar kinds of philosophical moves and, to my mind, decisively shows how unproductive they are. What he offers instead is a program for addressing these questions by integrating different kinds of inquiry, some more empirical and some more philosophical. Adapting his own description (151) somewhat: (1) We need a clear understanding of the structure of human causal cognition, one that highlights its most important and central features and explains how these features subserve such things as decision-making and strategies for learning.(2) We need a clear formulation, at the right level of abstraction and generality, of goals whose achievement is plausibly enabled by the ability to engage in causal cognition with the features described in (1) (for example, the goals of manipulating, predicting, and controlling aspects of the world around us).(3) We need a kind of empirical-cum-normative analysis of how and why different features of causal cognition contribute to achieving the goals described in (2).(4) In pursuit of (3), we need an account—again, at the right level of abstraction and generality—of the kinds of worldly structures (for example, structures of interventionist counterfactual dependence) that human causal cognition is responsive to.(5) We need a clear understanding of why (and to what extent) achieving the goals described in (2) is worthwhile—why, in Woodward’s words, achieving them counts as a kind of “success.”Returning briefly to our thought experiment, I think we can see, dimly but well enough, that if the goals in (2) at least include those Woodward consistently highlights—prediction, manipulation, and control—then our imaginary humans will do quite badly at using what they recognize as information about the causal structure of the world around them as a basis for manipulating and controlling their world. But the point of the thought experiment is really to get us to think harder and more productively about non-imaginary humans, and the complex interrelationships between the structure of human causal cognition, the structure of the world, and core sources of human value.Woodward’s articulation, defense, and illustration of this program is so rich and sophisticated that I can only hope to highlight a couple of key points. First, the kind of inquiry Woodward is urging philosophers of causation to engage in is necessarily interdisciplinary. All of it? Can’t the understanding described in (1) be achieved just by empirical psychology? In practice, no: as chapter 4’s well-curated review of results from the empirical psychological literature demonstrates, philosophical inquiry has played a crucial role both in generating hypotheses worth testing and in developing conceptual frameworks that permit the formulation of such hypotheses. (We might add that what counts as “important” or “central” can’t exactly be settled by experiment.) On the flip side, even inquiries like (5) will require close attention to empirically discoverable limitations on human causal cognition, lest we propose “from the armchair” goals that, however desirable, only angelic causal cognition could achieve.Second, you may have noticed the shift from talk of “our concept” of causation to talk of “human causal cognition.” Here I am following Woodward’s lead: he argues persuasively that focusing exclusively on what is part of, or involved in, “our concept” of causation will lead us to overlook features of causal cognition of vital importance. Woodward’s working example is invariance: roughly, the extent to which a given causal relationship holds across a wide variety of actual and possible circumstances. Chapter 5 provides a detailed philosophical exposition of invariance; chapter 6 insightfully applies this discussion to a range of well-known examples from the philosophical literature, including (my favorite) puzzles about double-prevention; chapter 7 walks the reader through a series of fascinating empirical studies that help bring out the role of invariance in causal cognition. These chapters make a compelling case that, from the perspective of Woodward’s functionalist approach, cognitive strategies for tracking and attending to different kinds of invariance are a very important part of human causal cognition. As just one example, chapter 5 ends with a brilliant discussion comparing invariance as a goal of inquiry with “explanatoriness” as such a goal; this passage should be required reading for any philosopher interested in “inference to the best explanation.” But as Woodward notes, invariance does not seem to matter to the traditional philosophical task of analyzing our concept of causation, as witness the ease with which we can construct examples in which it is “intuitively clear” that C causes E, but where this causal relationship is highly non-invariant. So much the worse for letting “our concept” delimit the boundaries of inquiry into causation. Chapter 8 closes out the book with some more tentative—in part because largely empirically untested—hypotheses about the importance to causal cognition of the concept of “proportionality” between cause and effect.Chapter 2 surveys philosophical theories of causation, devoting the most space to the “interventionist” approach Woodward favors. This chapter is the one weak spot in the book; readers not already at least somewhat familiar with contemporary philosophical debates about causation may find it more confusing than helpful. The treatment of rival views is oddly selective: for example, regularity theories of causation need not presuppose (as Woodward seems to suggest) a regularity theory of law; and one can pursue a Lewis-style counterfactual account without adopting Lewis’s peculiar semantics for counterfactuals. The treatment of interventionism doesn’t alert the reader to a range of important critiques in the literature. Finally, the distinction between “type-causal claims” and “token-causal claims” that the chapter begins with—and that is crucial to understanding the target of inquiry for the functional approach—never comes into focus sharply enough. The upshot is that “novice” readers would be well advised to augment chapter 2 with some well-chosen Stanford Encyclopedia of Philosophy articles and perhaps a foray into the literature critical of interventionism.Still, some perspective: while the problems in chapter 2 create a kind of barrier to entry, they in no way undermine the impressive case Woodward makes for his functionalist approach. My own view, to put it bluntly, is that philosophers working on causation should just stop doing the kind of thing Lewis taught us to do, and devote their energies and talents to pursuing the functionalist program instead. More: one thing that makes the functional approach so genuinely thrilling is the prospect of applying it to other topics. Woodward mentions a few in passing: “knowledge,” “free choice,” “morally responsible” (34–35). (Parroting our earlier thought experiment, we might productively ask what would go wrong for a community of people whose only knowledge-like concept just was the concept of justified true belief.) Like Woodward, I suspect the list could be extended both considerably and profitably, and the reason I hope this book gets read very widely within philosophy is precisely so that the invitation Woodward extends gets taken up: “I see what follows as a sort of testing ground, in connection with one particular set of concepts, of the merits of [the functionalist] approach” (35). In my view, these merits are not just considerable but game-changing.

中文翻译：

---

人性的因果关系：规范理论和描述心理学


这本精彩的书应该有助于为未来几年关于因果关系的哲学工作制定议程。事实上，它的影响值得更广泛地感受到：那些不关心因果关系的哲学家仍然可以通过仔细研究这本书而受益匪浅。它并非完美无缺：特别是，伍德沃德在第二章中对他所青睐的“干预主义”因果关系及其替代方案的处理可能会让外行读者迷失方向。也就是说，这些缺陷并没有破坏本书的主要哲学成就，即阐明、捍卫并详细说明某种方法论方向对因果关系哲学研究的力量，伍德沃德称之为“功能方法”。 “功能性这个词表明，因果关系的哲学研究应该首先关注的问题，粗略地说，是关于我们因果关系概念的功能——点、基本原理、服务目标的能力——的问题，以及关于因果关系的相应问题。这个概念所对应的各种世俗结构。 （我们很快就会看到，由于伍德沃德专业地阐述的原因，谈论因果关系的“我们的概念”太过局限性。但如果你是一位刚接触伍德沃德方法的哲学家，这可能是开始引导你的最佳方式。 ）第一章抽象地阐述了功能方法，特别关注它将描述性和规范性探究交织在一起对人类因果认知的方式。 第 3 章针对如何根据功能方法进行因果关系探究提出了一些具体且重要的经验教训；在此过程中，它耐心地对哲学家依赖案例“直觉”的标准方式提出了毁灭性的挑战，同时捍卫了这种直觉可以帮助指导探究的重要不同方式。这些章节共同构成了方法论的杰作。在这里不可能充分总结它们。相反，我要做的是通过考虑一种关于因果关系的哲学辩证法可以（但通常不会）继续进行的方式来间接地探讨主要思想。假设我提出一个简单的麦基式因果关系分析：事件 C 是一个原因事件 E 当且仅当 (1) C 和 E 都发生； (2) C 在 E 之前； (3) 关于 C 和 E 发生的情况，存在某个真命题 P，使得 P、C 发生的命题和捕捉基本定律的命题 L 的合取必然包含 E 发生的命题，但是单独使用 P 和 L 则不会。毫无进展！毕竟，产生 C 和 E 满足这些条件的案例是轻而易举的事，但直觉判断 C 和 E 是某些共同原因的联合影响。简而言之，这种分析将因果关系与共同原因引起的相关性混为一谈，因此必须予以拒绝。也许我们添加一些花哨的东西来修复分析；也许我们会尝试一些完全不同的东西。但无论哪种方式，现有的分析都被驳斥了。我相信，我刚才描述的是一个完全熟悉的例子，说明哲学的来回如何进行（而不仅仅是当主题是因果关系时）。 但现在考虑一下在这些辩证背景下很少进行的一种思想实验：想象一个人们的社区，其因果关系的概念恰好被这种分析所捕获。因此，他们认为我们所说的“因果关系”和我们所说的“共同原因引起的相关性”之间没有任何区别。如果有的话，他们会有什么问题吗？换句话说，为什么不采用他们思考因果关系的方式呢？伍德沃德认为我们应该问这些问题。我认为他是完全正确的。请注意，要很好地回答此类问题，我们必须抛开敲桌子的诱惑，说“他们不会真正考虑因果关系；他们不会真正考虑因果关系；他们不会真正考虑因果关系。”就是这样！”伍德沃德（在第一章和第三章）对类似的哲学举措进行了彻底而认真的阐述，在我看来，他果断地表明了它们是多么无效。相反，他提供的是一个通过整合不同类型的探究来解决这些问题的计划，其中一些更实证，一些更哲学。稍微调整一下他自己的描述（151）：（1）我们需要清楚地理解人类因果认知的结构，强调其最重要和最核心的特征，并解释这些特征如何促进决策和学习策略等事物(2) 我们需要在正确的抽象和一般性水平上对目标进行清晰的表述，这些目标的实现似乎是通过具有 (1) 中描述的特征进行因果认知的能力来实现的（例如，操纵的目标） ，预测和控制我们周围世界的各个方面）。（3）我们需要一种经验兼规范的分析来分析因果认知的不同特征如何以及为何有助于实现（2）中描述的目标。（4）为了追求（3），我们需要再次在正确的抽象和普遍性水平上解释人类因果认知所响应的世俗结构（例如，干预主义反事实依赖的结构）。 （5）我们需要清楚地理解为什么（以及在多大程度上）实现（2）中描述的目标是值得的——用伍德沃德的话来说，为什么实现这些目标算作一种“成功”。简单地回到我们的思想实验，我认为我们可以隐约但足够清楚地看到，如果（2）中的目标至少包括伍德沃德一贯强调的那些预测、操纵和控制，那么我们想象中的人类在使用他们所识别的信息时会做得很糟糕关于他们周围世界的因果结构作为操纵和控制他们的世界的基础。但思想实验的目的实际上是让我们更努力、更富有成效地思考非想象中的人类，以及人类因果认知结构、世界结构和人类价值核心来源之间复杂的相互关系。这个程序的阐述、辩护和说明是如此丰富和复杂，我只能希望强调几个关键点。首先，伍德沃德敦促因果哲学家进行的探究必然是跨学科的。所有的？ （1）中所描述的理解不能仅通过经验心理学来实现吗？实际上，答案是否定的：正如第四章对实证心理学文献结果的精心回顾所表明的那样，哲学探究在产生值得检验的假设和开发允许制定此类假设的概念框架方面都发挥着至关重要的作用。 （我们可能会补充说，什么算作“重要”或“核心”并不能完全通过实验来解决。）另一方面，即使是像（5）这样的询问也需要密切关注人类因果认知的经验可发现的局限性，以免我们提出“坐在扶手椅上”的目标，无论多么理想，只有天使般的因果认知才能实现。其次，你可能已经注意到从谈论“我们的因果概念”到谈论“人类因果认知”的转变。在这里，我追随伍德沃德的领导：他令人信服地指出，仅仅关注“我们的因果概念”的一部分或所涉及的内容将导致我们忽视因果认知的至关重要的特征。伍德沃德的例子是不变性：粗略地说，是给定的因果关系在各种实际和可能的情况下保持的程度。第五章对不变性进行了详细的哲学阐述；第六章富有洞察力地将这一讨论应用于哲学文献中的一系列著名例子，包括（我最喜欢的）关于双重预防的谜题；第七章引导读者完成一系列引人入胜的实证研究，这些研究有助于揭示因果认知中不变性的作用。这些章节提供了一个令人信服的案例，从伍德沃德功能主义方法的角度来看，跟踪和关注不同类型不变性的认知策略是人类因果认知的非常重要的一部分。仅举一个例子，第 5 章以一场精彩的讨论结束，将作为探究目标的不变性与作为这样的目标的“解释性”进行比较；对于任何对“推断最佳解释”感兴趣的哲学家来说，这段话应该是必读的。但正如伍德沃德指出的那样，不变性似乎对于分析我们因果关系概念的传统哲学任务并不重要，正如我们可以轻松地构造例子一样，在这些例子中“直观地清楚”C 导致了 E，但在哪里因果关系是高度非不变的。更糟糕的是让“我们的概念”划定了因果关系探究的界限。第 8 章以一些更具尝试性的假设（部分原因是在很大程度上未经实证检验）结束了本书，这些假设涉及因果关系之间“比例性”概念对因果认知的重要性。第 2 章概述了因果关系的哲学理论，其中用了最多的篇幅伍德沃德青睐的“干预主义”方法。本章是本书的一个薄弱环节；尚未至少有点熟悉当代关于因果关系的哲学辩论的读者可能会发现它更令人困惑而不是有帮助。奇怪的是，对敌对观点的处理是有选择性的：例如，因果关系的规律性理论不需要预设（正如伍德沃德似乎暗示的那样）法律的规律性理论；人们可以追求一种刘易斯式的反事实解释，而不需要采用刘易斯特有的反事实语义。对干预主义的处理并没有提醒读者注意文献中的一系列重要批评。最后，本章开头的“类型因果断言”和“符号因果断言”之间的区别——这对于理解功能方法的探究目标至关重要——从未引起足够的关注。结果是，建议“新手”读者在第二章中添加一些精心挑选的斯坦福哲学百科全书文章，或许还可以涉足批评干预主义的文献。尽管如此，还是有一些观点：虽然第二章中的问题造成了某种进入壁垒，但它们绝不会削弱伍德沃德为其功能主义方法所做的令人印象深刻的案例。坦白地说，我自己的观点是，研究因果关系的哲学家应该停止做刘易斯教我们做的事情，而将他们的精力和才能投入到追求功能主义纲领上。更多：使函数式方法真正令人兴奋的一件事是将其应用于其他主题的前景。伍德沃德顺便提到了一些：“知识”、“自由选择”、“道德责任”（34-35）。 （重复我们之前的思想实验，我们可能会富有成效地问，对于一个唯一的类似知识的概念就是合理的真实信念的概念的人来说，会出现什么问题。）像伍德沃德一样，我怀疑这个列表可以大大扩展并且有利可图。 ，我希望这本书在哲学领域得到广泛阅读的原因正是为了让伍德沃德发出的邀请得到接受：“我认为接下来的内容是一种试验场，与一组特定的概念有关，它的优点[功能主义]方法”（35）。在我看来，这些优点不仅是巨大的，而且是改变游戏规则的。