A new book by Ernest Sosa is always an event. In a philosophical age where much of the focus is on piecemeal issues, Sosa has forged ahead with a novel virtue-theoretic treatment of a range of core questions in epistemology that is self-consciously systematic. Note that ‘epistemology’ is here broadly conceived. Indeed, a key part of the Sosa project has been to enlarge the reach of mainstream epistemology and thereby draw out connections with other areas of philosophy that have hitherto been underexplored, especially ethics, philosophy of mind and action, and metaphysics. Moreover, Sosa is also unusual among contemporary philosophers in having an acute grasp of the history of the subject, which he brings to bear in support of his program. The result is an incredibly sophisticated vision of how a range of topics in epistemology fit together.1Epistemic Explanations: A Theory of Telic Normativity, and What it Explains is the new installment in Sosa’s distinctive brand of virtue epistemology. The book is organized into four parts. Part 1 is devoted to articulating the telic virtue epistemology framework that Sosa defends. In light of this framework, he explores how we should account for the undoubted importance of first-hand knowledge and understanding and how we should conceive of the relationship between the theory of knowledge and intellectual ethics. Part 2 offers a comprehensive treatment of the epistemology of suspension. Part 3 is primarily concerned with default assumptions and understanding how they lead to refinements of telic virtue epistemology. Among other things, this part defends a metaphysical hierarchy of epistemic categories that includes a discussion of what Sosa terms secure knowledge, which is a particularly important epistemic category within his framework. Part 4 builds on the account of default assumptions in part 3 by offering an extended discussion of how this bears on the Wittgenstein–Moore debate (roughly, the clash of hinge epistemology with a form of epistemic foundationalism). Like all Sosa’s work, the writing is refreshingly crisp. It is also deceptively readable, in that one can find oneself surprised at just how much philosophical ground is being covered.There is much that I agree with in this book, but I would like to take this opportunity to critically focus on Sosa’s intriguing appeal to default assumptions and how it plays out both in terms of his theory of knowledge and his approach to radical skepticism. As will be familiar to readers of Sosa’s work, he understands knowledge in terms of what he calls aptness. Roughly, a performance is apt (‘accurate because adroit’) when one’s success in the target endeavour is properly attributable to one’s manifestation of relevant skill. As applied to the epistemic realm, we thus get the idea that knowledge is apt belief—that is, one knows when one’s cognitive success (true belief) is properly attributable to one’s manifestation of relevant cognitive agency.2Sosa’s novel claim is that apt performances can legitimately presuppose certain background conditions that the agent might have no way of knowing obtain, and which might in fact obtain simply by luck. Sosa gives the example of the performance of a baseball fielder in a nighttime game who presupposes that the lighting is working (160). Sosa argues that these background conditions can be non-negligently assumed to be in place (at least if one is given no explicit reason to consider them), even if their obtaining is just a matter of luck, and hence is unsafe. Indeed, to concern oneself with the obtaining of these conditions, and thereby weaken one’s focus on the skillful task in hand, would be negligence. (The baseball fielder would likely be less effective at his game if he is concerned about whether the lighting is working safely).3When such an account of apt performance is applied to the epistemic realm in the form of apt belief, we end up with a version of relevant alternatives theory. Recall that the basic idea behind relevant alternatives theory is that in order to know we do not need to exclude all possibilities of error but just the relevant ones. The irrelevant error possibilities can be legitimately assumed to be false (i.e., without one having any specific reason to think they are false). That certainly sounds right. Knowledge can be fallible, after all (i.e., acquired via fallible processes), and hence why would it be required for knowledge that all possibility of error be excluded? If that is correct, however, then the view seems to have immediate antiskeptical import, since if any error possibility looks like it would be irrelevant in the target sense of the term it is surely radical sceptical scenarios. Accordingly, our everyday knowledge of shoes, ships, sealing wax, cabbages, and kings can be compatible with our failure to know the denials of radical skeptical hypotheses.All that we need to make this story stick is a principled account of how to delineate the relevant alternatives from the irrelevant ones. Various proposals have been offered in this regard, such as accounts in terms of the modal closeness of the error possibility, its conversational salience, and so on. Sosa is offering his own slant on this topic by arguing that apt performance, and hence apt belief, can legitimately involve assuming default assumptions and that this can be so even when those assumptions concern modally close error possibilities. So long as the default assumptions are false and the subject is given no specific reason for entertaining them, then she can nonetheless manifest apt belief and thereby acquire knowledge. Moreover, Sosa explicitly draws an antiskeptical moral from this point by claiming that everyday knowledge can legitimately coexist with our default assumption that radical skeptical hypotheses are false. As he puts it, “Conscientiously enough, without negligence or recklessness, we normally assume ourselves free of skeptical scenarios. And this assumption is proper even on the rare occasions when it is true but not known to be true and even quite unsafe” (159).Notice that in taking this line Sosa is embracing what is regarded by many as an unfortunate consequence of (a straightforward version of) relevant alternatives theory, which is the denial of the closure principle for knowledge.4 After all, one can now know everyday claims and be fully aware that they entail the denials of radical skeptical hypotheses, but be merely assuming that these denials are false. Sosa also makes a further claim that to my knowledge no other proponent of relevant alternatives theory makes (at least not explicitly anyway), which is that one’s knowledge can be unsafe and no less bona fide as a result. In particular, since it is compatible with one’s knowledge of the target proposition that the background assumptions are only luckily true, and hence could have easily been false, and since the falsity of these background assumptions can make one’s belief in the target proposition false, then one’s belief in the known target proposition can be only luckily true (i.e., it could have easily been false).I think Sosa is quite right about apt performance in general being compatible with unsafety; indeed, I take this to be a deep insight of Sosa’s work. As we can put the point, achievements (i.e., successes that are because of ability, or apt performances in Sosa’s terminology) can be modally fragile (i.e., the target success could have easily been a failure). And Sosa has put his finger on the source of the issue here: all that matters for apt performance is that the legitimate background assumptions are true—in particular, it does not matter whether they are luckily true. That is why the fielder’s apt performance is compatible with the fact that the stadium lights could have easily failed (in which case his attempted catch would have gone awry).Where I would diverge with Sosa is that I do not think the same is true of knowledge. That is, if one’s cognitive success (true belief) could have easily been failure, then it is not knowledge. Knowledge excludes luck (fragility) of this kind because it excludes high levels of epistemic risk (i.e., the epistemic risk that one’s belief is false). That is why, when one knows, one’s basis for belief is such that one could not easily be wrong, which is what safety demands. In contrast, Sosa’s proposed alternative picture would commit us to allowing that knowledge can coexist with high levels of epistemic risk. The upshot then is that knowledge is not apt belief (relatedly, knowledge is not a cognitive achievement).5Even if we grant to Sosa that apt belief (or cognitive achievement), even when unsafe, can amount to knowledge, it is not clear that this affords us a purchase on the problem of radical skepticism as he supposes. Consider what his view amounts to. We can recognize that our everyday knowledge presupposes the falsity of global error possibilities that we can never (even in principle) exclude, and yet be content to continue regardless as if we have knowledge. So long as skeptical scenarios do not in fact obtain (and even if they could very easily obtain), then our beliefs can be aptly formed even though we groundlessly presuppose that we are not radically in error, and hence can amount to knowledge. Relatedly, and in line with the rejection of closure, one can self-consciously be aware that one has knowledge of everyday propositions that one knows entail the denials of skeptical hypotheses (e.g., that one has hands) while being unable to know the entailed proposition (e.g., that one is not a BIV).While I find Sosa’s general line in epistemology very persuasive, I find the line he takes on radical skepticism deeply unpersuasive. To begin with, if radical skeptical scenarios could easily occur, then surely we are all epistemically doomed! Why would we be tempted to suppose otherwise (except unless we were already in the grip of an—undoubtedly compelling—picture that suggested this position)? But our inability to exclude them is troubling even if we do not grant that they are modally close. The worry is that if we cannot exclude them, then how can we be sure that our beliefs have satisfied any bona fide epistemic standard, even of the lowest kind? This is precisely why they are very different from normal, ‘local,’ background conditions, which do not call into question one’s general epistemic relationship to the world.In defense of his antiskeptical line, Sosa notes that “no-one is likely to entirely avoid illusion and every other perceptual error” (138). But what is the relevance of this in the context of radical skepticism? Remember that the radical sceptic is not appealing to a high epistemic standard, as if we need to be infallible in order to know. Their claim is rather that on the face of it, we have satisfied no epistemic standard at all (i.e., not even a fallible one). Consider also this remark: “Because we are essentially rational animals, we have no real option on how to proceed cognitively over the enormous span of the animal knowledge we rely on in any ordinary day” (138). I do not think anyone would dispute this, but what exactly is meant to follow? For note that even the radical sceptic can agree with such a claim. It may well be that we need to act as if we have knowledge. But that does not provide any reason for thinking that we do have knowledge.For what it is worth, I entirely agree with Sosa that we cannot know that we are not radically in error, but I do not think he correctly captures why this is so. For that, we need to embrace Wittgenstein’s conception of hinge commitments, albeit in a very different form to how Sosa conceives of this notion (see Wittgenstein 1969). Our conviction that we are not radically in error is not an assumption, nor is it an incidental lack on our parts. What is needed is an understanding of why such claims are not properly even in the market for knowledge (such that our not knowing them does not amount to ignorance). But that entails, contra Sosa, differentiating our hinge commitments from local background conditions. Wittgenstein’s point is precisely that our failure to know our hinge commitments is not a cognitive limitation on our parts, but reveals an important truth about the structure of rational evaluation, a structure that both the radical skeptic and the traditional antiskeptic (including, I would argue, Sosa) misunderstands.6 But a defense of that claim is a topic for another day.7

中文翻译：

---

认知解释：目的规范性理论及其解释

欧内斯特·索萨 (Ernest Sosa) 的新书总是一件大事。在一个大部分注意力都集中在零碎问题上的哲学时代，索萨对认识论中的一系列核心问题进行了新颖的美德理论处理，这是一种自觉的系统化。请注意，这里的“认识论”是广义的。事实上，索萨计划的一个关键部分是扩大主流认识论的范围，从而与迄今为止未被充分探索的其他哲学领域建立联系，尤其是伦理学、心灵和行动哲学以及形而上学。此外，索萨在当代哲学家中也不同寻常，因为他对主题的历史有着敏锐的把握，他用这些来支持他的计划。结果是关于认识论中的一系列主题如何组合在一起的令人难以置信的复杂愿景。1认知解释：目的规范性理论及其解释是 Sosa 独特的美德认识论品牌的新部分。本书分为四个部分。第 1 部分致力于阐明 Sosa 捍卫的目的美德认识论框架。根据这个框架，他探讨了我们应该如何解释第一手知识和理解的毋庸置疑的重要性，以及我们应该如何理解知识理论与知识伦理之间的关系。第 2 部分全面介绍了暂停的认识论。第 3 部分主要关注默认假设并理解它们如何导致目的美德认识论的改进。除其他事项外，这部分捍卫了认知范畴的形而上学层次结构，其中包括对 Sosa 所说的安全知识的讨论，这是他的框架中一个特别重要的认知范畴。第 4 部分以第 3 部分中默认假设的说明为基础，详细讨论了这如何影响维特根斯坦-摩尔辩论（粗略地说，铰链认识论与某种形式的认知基础主义的冲突）。与 Sosa 的所有作品一样，文笔清新爽快。它也具有欺骗性的可读性，因为你会发现自己对所涵盖的哲学基础如此之多感到惊讶。我同意这本书中的很多内容，但我想借此机会批判性地关注 Sosa 对默认假设的有趣吸引力，以及它如何在他的知识理论和激进怀疑主义方法方面发挥作用。正如索萨作品的读者所熟悉的那样，他根据他所谓的恰当性来理解知识。粗略地说，当一个人在目标努力中的成功适当地归因于一个人相关技能的表现时，表现是恰当的（“准确因为熟练”）。当应用于认知领域时，我们因此得到知识是恰当的信念的想法——也就是说，一个人知道什么时候一个人的认知成功（真正的信念）可以适当地归因于一个人相关认知机构的表现。2Sosa 的新颖主张是，apt 表演可以合法地预设某些背景条件，代理人可能无法知道这些条件是否获得，而实际上可能只是靠运气获得。Sosa 举了一个棒球守场员在夜间比赛中的表现的例子，他假设灯光正常工作 (160)。Sosa 认为，这些背景条件可以被非疏忽地假设到位（至少在没有明确理由考虑它们的情况下），即使它们的获得只是运气问题，因此是不安全的。的确，只关注获得这些条件，从而削弱了对手头善巧任务的专注，将是疏忽。（如果棒球守场员担心照明是否安全工作，他的比赛效率可能会降低）。3当这种对恰当表现的解释以恰当信念的形式应用于认知领域时，我们最终得到了相关替代理论的一个版本。回想一下相关替代理论背后的基本思想是，为了知道我们不需要排除所有错误的可能性，而只需排除相关的错误。不相关的错误可能性可以合理地假设为假（即，没有任何特定理由认为它们是假的）。这听起来当然是对的。毕竟，知识可能会出错（即，通过容易出错的过程获得），因此为什么知识需要排除所有出错的可能性？然而，如果这是正确的，那么该观点似乎具有直接的反怀疑意义，因为如果任何错误可能性看起来与术语的目标意义无关，那肯定是激进的怀疑情景。因此，我们对鞋子、船只、封蜡、卷心菜和国王的日常知识可以与我们对激进怀疑论假说的否认的失败相容。从不相关的选择中选择相关的选择。在这方面已经提出了各种建议，例如根据错误可能性的模态接近度、其会话显着性等进行说明。索萨通过论证恰当的表现和因此恰当的信念，在这个话题上提出了自己的观点，可以合理地涉及假设默认假设，即使这些假设涉及模态关闭错误的可能性，也可以如此。只要默认假设是错误的，并且没有给主体提供接受这些假设的具体理由，那么她仍然可以表现出恰当的信念，从而获得知识。此外，Sosa 通过声称日常知识可以与我们默认的激进怀疑假设是错误的假设合理共存，从这一点明确得出了反怀疑主义的道德观。正如他所说，“足够认真，没有疏忽或鲁莽，我们通常会假设自己没有怀疑的情景。即使在极少数情况下，这种假设也是正确的，即它是真实的但不知道是真实的，甚至是非常不安全的”（159）。请注意，在采取这条路线时，Sosa 接受了许多人认为相关替代理论（的直接版本）的不幸后果，即否认知识的封闭原则。 4 毕竟，人们现在可以知道日常主张并充分意识到它们需要否定激进的怀疑假设，但只是假设这些否定是错误的。Sosa 还进一步声称，据我所知，没有其他相关替代理论的支持者提出（至少没有明确表示），这就是一个人的知识可能是不安全的，结果也同样是善意的。特别是，由于它与目标命题的知识相容，即背景假设只是幸运地为真，因此很容易为假，并且由于这些背景假设的错误会使一个人对目标命题的信念为假，那么一个人对已知目标命题的信念只能幸运地为真（即，它很容易为假）。我认为 Sosa 对 apt 的看法是正确的总体性能与不安全性相容；事实上，我认为这是对索萨作品的深刻理解。正如我们可以指出的那样，成就（即由于能力而获得的成功，或者用 Sosa 的术语来说是适当的表现）可能是模态脆弱的（即，目标成功很容易成为失败）。Sosa 在这里指出了问题的根源：对 apt 性能而言，重要的是合理的背景假设是真实的——特别是，它们是否幸运地真实并不重要。这就是为什么守场员的恰当表现与体育场灯很容易失效的事实相符（在这种情况下，他的尝试接球会出错）。我与索萨不同的地方是，我认为情况并非如此知识。也就是说，如果一个人的认知成功（真正的信念）很容易失败，那么它就不是知识。知识排除了这种运气（脆弱性），因为它排除了高水平的认知风险（即，一个人的信念是错误的认知风险）。这就是为什么当一个人知道时，一个人的信念基础就是这样一个人就不会轻易犯错，这就是安全所要求的。相比之下，Sosa 提出的替代图景将使我们承诺允许知识可以与高水平的认知风险共存。那么结果就是知识不是恰当的信念（相关地，知识不是认知成就）。5即使我们承认 Sosa 恰当的信念（或认知成就），即使在不安全的情况下，也可以构成知识，但尚不清楚这是否为我们提供了解决激进怀疑主义问题的依据，因为他假设。考虑一下他的观点。我们可以认识到，我们的日常知识预设了我们永远（甚至在原则上）无法排除的全局错误可能性的虚假性，但我们仍然满足于继续下去，就好像我们拥有知识一样。只要怀疑的情景实际上没有获得（即使它们很容易获得），那么我们的信念就可以恰当地形成，即使我们毫无根据地预设我们并没有完全错误，因此可以构成知识。相关地，根据拒绝关闭，一个人可以自觉地意识到自己拥有日常命题的知识，而这些日常命题会导致对怀疑假设的否定（例如，有人有手），但无法知道所蕴含的命题（例如，一个人不是 BIV）。虽然我发现索萨在认识论中的一般路线非常有说服力，但我发现他对激进怀疑主义的路线非常没有说服力。首先，如果激进的怀疑情景很容易发生，那么我们肯定都在认知上注定要失败！为什么我们会试图做出不同的假设（除非我们已经掌握了表明这一立场的——无疑是引人注目的——图片）？但是，即使我们不承认它们在模态上很接近，我们也无法排除它们，这令人不安。担心的是，如果我们不能排除它们，那么我们如何确定我们的信念满足任何真正的认识标准，即使是最低的标准？这正是它们与正常的、“本地的”背景条件截然不同的原因，这些条件不会质疑一个人与世界的一般认知关系。为了捍卫他的反怀疑主义路线，索萨指出“没有人可能完全避免错觉和其他所有知觉错误”（138）。但是，在激进怀疑主义的背景下，这有什么相关性呢？请记住，激进的怀疑论者并没有吸引高认知标准，就好像我们需要万无一失才能知道一样。他们的主张更确切地说，从表面上看，我们根本没有满足任何认知标准（即，甚至不是一个容易犯错的标准）。还请考虑一下这句话：“因为我们本质上是理性动物，对于我们在日常生活中所依赖的大量动物知识，我们没有真正的选择来进行认知”（138）。我认为没有人会对此提出异议，但究竟是什么意思呢？请注意，即使是激进的怀疑论者也会同意这样的说法。很可能我们需要表现得好像我们有知识一样。但这并没有提供任何理由认为我们确实拥有知识。就其价值而言，我完全同意 Sosa 的观点，即我们无法知道我们并没有完全错误，但我认为他没有正确地抓住为什么会这样. 为此，我们需要接受维特根斯坦的铰链承诺概念，尽管其形式与 Sosa 对这一概念的构想截然不同（参见 Wittgenstein 1969）。我们坚信我们并没有完全错误不是一种假设，这也不是我们方面的偶然缺乏。需要的是理解为什么即使在知识市场上这样的主张也不恰当（这样我们不知道它们并不等于无知）。但是，与 Sosa 相反，这需要将我们的关键承诺与当地背景条件区分开来。维特根斯坦的观点恰恰是，我们不知道我们的关键承诺并不是我们的认知局限，而是揭示了关于理性评估结构的一个重要事实，这种结构无论是激进的怀疑论者还是传统的反怀疑论者（包括，我认为, Sosa) 误解了。6 但为该主张辩护是另一天的话题。 7 需要的是理解为什么即使在知识市场上这样的主张也不恰当（这样我们不知道它们并不等于无知）。但是，与 Sosa 相反，这需要将我们的关键承诺与当地背景条件区分开来。维特根斯坦的观点恰恰是，我们不知道我们的关键承诺并不是我们的认知局限，而是揭示了关于理性评估结构的一个重要事实，这种结构无论是激进的怀疑论者还是传统的反怀疑论者（包括，我认为, Sosa) 误解了。6 但为该主张辩护是另一天的话题。 7 需要的是理解为什么即使在知识市场上这样的主张也不恰当（这样我们不知道它们并不等于无知）。但是，与 Sosa 相反，这需要将我们的关键承诺与当地背景条件区分开来。维特根斯坦的观点恰恰是，我们不知道我们的关键承诺并不是我们的认知局限，而是揭示了关于理性评估结构的一个重要事实，这种结构无论是激进的怀疑论者还是传统的反怀疑论者（包括，我认为, Sosa) 误解了。6 但为该主张辩护是另一天的话题。 7