宗教信仰和推荐系统解决同一问题

人工智能 ,哲学, 舆论 (Artificial Intelligence, Philosophy, Opinion)

On the surface, religions and recommender systems appear to have nothing in common. One is a historical and social phenomenon, providing the moral foundations for our deepest convictions and beliefs. The other is a recent, fairly mundane technology mostly aimed at giving us better search results, ads, or movie recommendations. Religion answers the big questions, while recommender systems answer the trivial ones.

从表面上看,宗教和推荐系统似乎没有什么共同之处。 一种是历史和社会现象,为我们最深刻的信念和信念提供了道德基础。 另一种是最新的相当平凡的技术,主要旨在为我们提供更好的搜索结果,广告或电影推荐。 宗教回答了大问题,而推荐系统回答了小问题。

Yet I will argue here that both are methods for dealing with uncertainty and information overload, albeit in very different ways. Religious thought and recommender systems are two ways of solving the same basic problem: coordinating and directing our actions towards specific goals while navigating chaotic, uncertain environments. But might there be any benefit to actively seeking out situations of uncertainty? Is information overload always bad? When we look back on today from 100 years in the future, we might wonder:

然而,我将在这里争辩说,这两种方法都是用于处理不确定性和信息过载的方法,尽管方式非常不同。 宗教思想和推荐系统是解决同一基本问题的两种方法:协调和引导我们的行动朝着特定目标发展,同时在混乱,不确定的环境中导航。 但是,积极寻找不确定的情况可能有什么好处? 信息过载总是不好吗? 当我们回顾未来100年的今天时,我们可能会怀疑:

Will recommender systems eventually do for us in complex informational environments what religion did for us in complex natural environments?

推荐系统最终会在复杂的信息环境中为我们做些什么,宗教在复杂的自然环境中为我们做些什么?

推荐系统和信息超载 (Recommender Systems and Information Overload)

Recommender systems can be seen as a solution to the problem of information overload on the Internet. In other words, recommender systems can be viewed as a kind of cognitive technology that filters out irrelevant information, thereby providing meaningful structure to a morass of complex digital information surrounding us in digital reality. This kind of cognitive technology has real business value as well, as evidenced by the ubiquity of recommender systems on major Behavioral Big Data (BBD) platforms such as Netflix, Spotify, Youtube, and Facebook, just to name a few. A recommender system might also be characterized as a metaproduct — it’s a product whose purpose is to recommend other products.

推荐系统可以看作是Internet上信息过载问题的一种解决方案。 换句话说,推荐系统可以看作是一种过滤掉不相关信息的认知技术,从而为数字现实中包围我们的复杂数字信息提供了有意义的结构。 诸如Netflix,Spotify,Youtube和Facebook等主要行为大数据(BBD)平台上的推荐系统无处不在,这种认知技术也具有真正的商业价值。 推荐系统也可能被描述为元产品 -它是一种旨在推荐其他产品的产品。

Interestingly, the process described above mimics what happens in the human brain, whereby non-conscious perceptual information is “selected” for via conscious access. In the study of consciousness this is termed selective attention and largely operates at the non-conscious level. According to the neuroscientist Stanislas Dehaene, “attention serves as the gateway for consciousness.” Said differently, some act of selective filtering must occur before any kind of conscious access to mental content is possible.

有趣的是,上述过程模仿了人脑中发生的事情,从而通过有意识的访问“选择”了无意识的感知信息。 在意识研究中,这被称为选择性注意 ,主要在无意识水平上起作用。 根据神经科学家Stanislas Dehaene的说法,“注意力是意识的门户。” 换句话说,在任何形式的有意识的对精神内容的访问成为可能之前,必须发生一些选择性过滤的动作。

So the phenomenon of information overload is not unique to living in the digital age. Brains have been doing this for millions of years already. Today the problem is not getting information, it’s getting the right information. We’ve seen this is a matter of filtering or selecting that which is most relevant to us at the present moment. The tricky question is figuring out what exactly is most relevant! If we were pre-historic hunter gatherers, relevant information might be related to the sounds of nearby predators or memories of poisonous snakes. But times have changed.

因此,信息超载现象并非生活在数字时代。 大脑已经这样做了数百万年 今天的问题不是获取信息,而是获取正确的信息。 我们已经看到,这是过滤或选择当前与我们最相关的问题。 棘手的问题是找出最相关的是什么! 如果我们是史前猎人的采集者,则相关信息可能与附近捕食者的声音或毒蛇的记忆有关。 但是时代变了。

人类对信息处理封闭的需求 (The Human Need for Information Processing Closure)

There appears to be a fundamental human need for reducing uncertainty, a need with us since the first roaming bands of humans formed social collectives and told one another mythical stories of creation. These stories made simple what was otherwise a complex natural environment, full of chaos and unpredictability. Given these simple beliefs, we were able to infer relatively well (enough to reproduce!) what to expect from a wide variety of natural events and organisms. That made decision-making much easier.

人类似乎存在着减少不确定性的根本需求,这是我们的需求,因为人类的最初漫游乐队形成了社会集体,并彼此讲述了关于创造的神话故事。 这些故事使原本复杂的自然环境变得简单,充满了混乱和不可预测性。 有了这些简单的信念,我们就能够相对较好地推断(足以繁殖!)对各种自然事件和有机体的期望。 这使决策更加容易。

Today, psychologists refer to this problem as one of cognitive overload. Social psychologist Arie Kruglanski claims we have a basic need for information processing closure and are inherently motivated to find answers to “ambiguous situations.” Anyone who’s ever excitedly Googled some obscure trivia question would probably concur with Kruglanski. In states of information non-closure, Kruglanski claims we tend to either “freeze or seize.” Reaching closure means being able to make a decision and act, but in doing so we risk missing out on potentially better, more accurate information. We can either exploit the imperfect information we have now, or explore potentially better options and risk missing a golden window of opportunity.

今天,心理学家将此问题称为认知超负荷之一 。 社会心理学家Arie Kruglanski声称我们对信息处理的关闭有基本的需求,并且天生就有动机去寻找“模棱两可的情况”的答案。 任何曾经激动地用一些晦涩的琐事问题搜索过Google的人都可能会同意Kruglanski的观点。 克鲁格斯基声称,在信息不公开的状态下,我们倾向于“冻结或抓住”。 达成封闭意味着能够做出决定并采取行动,但是这样做有可能会丢失可能更好,更准确的信息。 我们可以利用现有的不完善信息,也可以潜在地探索 更好 选择权和风险,错失良机。

Here is just small sampling of various ways humans have traditionally dealt with uncertainty:

这只是人类传统上处理不确定性的各种方式的小样本:

  • Heuristics (Pick the first one that satisfies constraint x)

    启发式 ( 选择第一个满足x的约束 )

  • Prejudice (I don’t like Americans)

    偏见 ( 我不喜欢美国人 )

  • Stereotypes (Asians are good at math)

    刻板印象 ( 亚洲人擅长数学)

Also related to information-processing closure are what we would typically call the traits of “closed” and “open-mindedness.” Those who are open-minded might pause, collect more information, and consult others before making a decision to act. In contrast, those scoring high in “close mindedness” will tend to look for quick, easy sources of information to relieve the cognitive discomfort of uncertainty.

与信息处理闭包相关的还有我们通常所说的“封闭”和“开放思想”的特征。 胸襟开阔的人可能会停下来,收集更多信息,并在决定采取行动之前咨询他人。 相反,得分高的“近距离思维”人士将倾向于寻找快速,简便的信息来源,以减轻不确定性带来的认知不适。

I’m sure we can all think of some popular news media outlets which present information in this way. Simple political stories explaining observed “data” (e.g., news events) are given to help people cope with the complexities of democratic government. Ben Novak’s book Hitler and Abductive Logic: The Strategy of a Tyrant is a great case study in the potential for mass manipulation through clever abductive storytelling.

我敢肯定,我们都会想到一些以这种方式提供信息的流行新闻媒体。 简单的政治故事解释了观察到的“数据”(例如新闻事件),以帮助人们应对民主政府的复杂性。 本诺瓦克(Ben Novak)的书《 希特勒与归纳逻辑 :暴君的策略》是一个很好的案例研究,说明了通过巧妙的诱拐故事讲述进行大规模操纵的可能性。

In the language of recommender systems we might refer to open mindedness as “serendipity” and close-mindedness as being stuck in a “filter bubble.”

用推荐系统的语言,我们可以将开放思想称为“偶然性”,将封闭思想称为“过滤器泡沫”。

If a major goal of recommender systems is to remove irrelevant information for a given user, what counts as irrelevant? The process of personalization links this filtering process with inferences about a person’s underlying interests and preferences. For example, if I know you’re searching for flights to Barcelona in December, maybe you prefer beaches to snowy mountains. For designers of recommender systems, the fundamental problem to be solved is this: how can I predict the interests or preferences of my customers or users?

如果推荐系统的主要目标是删除给定用户的不相关信息,那么什么算不相关个性化过程将此过滤过程与关于一个人的基本兴趣和偏好的推断联系在一起。 例如,如果我知道您要搜索12月飞往巴塞罗那的航班,也许您更喜欢海滩而不是雪山。 对于推荐系统的设计人员来说,要解决的基本问题是:如何预测客户或用户的兴趣或偏好?

自我不确定性:在数字时代敢于认识自己 (Self-Uncertainty: Daring to Know Ourselves in the Digital Age)

OK, so perhaps it’s in our nature to want to reach information-processing closure. The normative question still remains: should we? Might there be some advantage to reveling in uncertainty and doubt?

好的,所以达到信息处理封闭也许是我们的本性。 规范性问题仍然存在:我们应该吗? 陶醉于不确定和怀疑中可能会有一些优势吗?

Immanuel Kant, in his 1784 essay “What Is Enlightenment?” asked us to dare to know and thereby reject all claims to truth based on the appeal to dogma, faith, or authority. “Enlightenment is man’s emergence from his self-incurred immaturity,” he wrote. For too long, we unquestioningly accepted what others told us. For Kant, daring to know means staring uncertainty in the eyes and attempting to use your intellectual capacities for critical thinking, even when doing so might be dangerous or socially unpopular.

依曼纽尔·康德(Immanuel Kant)在1784年的文章“ 什么是启蒙运动? 要求我们知道并基于对教条,信仰或权威的呼吁拒绝所有对真理的主张。 他写道:“启蒙是人类从自身的不成熟中脱颖而出。” 长期以来,我们毫无疑问地接受了别人告诉我们的。 对于康德来说, 敢于知道意味着在眼睛中凝视不确定性,并试图将自己的智力能力用于批判性思维,即使这样做可能很危险或在社会上不受欢迎。

In Kant’s view, the key to freedom and autonomy is the courage to think and reason for oneself, from foundational principles without influence from oppressive historical, political, social, or religious forces. Kant’s Groundwork of the Metaphysics of Morals explores our capacity for reason as a fundamental, human universal, uniting, and providing us with a theoretical foundation for morality — as an action guide whose principles would be acceptable to any reasonable/rational thinker.

康德认为,自由和自治的关键是勇于思考和理性思考,他们从基本原则出发,不受历史,政治,社会或宗教势力的压迫。 康德的《 道德形而上学基础》探讨了我们作为基本的,人类的普遍性,团结性以及为我们提供道德的理论基础的理性能力,以此作为行动指导,其原则可以为任何理性/理性思想家所接受。

Even if we are predisposed by nature to appeals to faith, authority, and habit, Kant would say that realizing our human potential demands more. He would implore us to use our unique faculty of reason to think about and understand the nature of things around us.

即使我们天生就倾向于信仰,权威和习惯,康德也会说,实现我们的人类潜力还需要更多。 他会恳求我们利用我们独特的理性能力去思考和理解我们周围事物的本质。

道德不确定性:您是谁? 你该怎么办? (Ethical Uncertainty: Who Are You? What Should You Do?)

Some of the most pressing decisions in life must be made under near complete uncertainty and this fact seems unlikely to change. These kinds of situations are very different from the ones where recommender systems are currently deployed, however. Those situations usually involve trivial decisions about what videos to watch on Youtube, or which accounts to follow on Instagram.

生活中一些最紧迫的决定必须在几乎完全不确定的情况下做出,这一事实似乎不太可能改变。 但是,这些情况与当前部署推荐系统的情况有很大不同。 这些情况通常涉及对要在Youtube上观看哪些视频或在Instagram上关注哪个帐户的重大决定。

Despite this, I suggest we must not blindly accept the use of recommendation systems where they are not needed. Learning to make difficult decisions in the face of uncertainty is, in my opinion, a skill necessary for a good life. As Kant argued, we should dare to know and claim our own capacity for practical reasoning instead of relying on the dogma and authority of those responsible (corporations and data scientists) for building recommender systems.

尽管如此,我建议我们不要盲目接受不需要它们的推荐系统。 我认为,面对不确定性,学会做出艰难的决定是一种美好生活所必需的技能。 正如康德所言,我们应该敢于知道并要求我们拥有进行实际推理的能力,而不是依靠负责人(机构和数据科学家)的教条和权威来建立推荐系统。

Why might blind reliance on automated recommendation be a bad thing for human development? Recommender systems, taken to their logical extreme, represent a form of moral outsourcing, the outsourcing of practical reasoning in two key, interrelated areas:

为什么盲目依赖自动推荐对人类发展是一件坏事? 推荐系统在逻辑上达到了极致,代表了道德外包的一种形式,这是在两个相互关联的关键领域中的实践推理的外包:

1) Practical reasoning about which type of persons are better and worse to become

1)关于哪种类型的人变得更好和更坏的实践推理

2) Practical reasoning about which actions are better and worse to perform

2)关于哪些动作执行得更好和更坏的实用推理

These aspects of practical reasoning are related because only once we have decided on what a good person is can we begin to act in the appropriate way to become such a person. It’s no surprise then, that in modern life many people have never taken the time to consider what type of person they would like to be. Without religious beliefs to guide us, these kinds of questions are apt to foster an existential crisis. Sartre famously claimed that existence precedes essence: we are born into this world, and without the guidance of God, are faced with the arduous and scary task of figuring out what exactly we are in this world. Heidegger too grappled with this problem through his concept of dasein.

实践推理的这些方面是相关的,因为只有当我们确定一个好人是什么之后,我们才能开始以适当的方式采取行动成为这样的人。 那么就不足为奇了,在现代生活中,许多人从来没有花时间去考虑自己想成为什么样的人。 没有宗教信仰的指导,这些问题很容易引发生存危机。 萨特(Sartre)著名地宣称, 存在先于本质 :我们出生于这个世界,没有上帝的指导,我们面临着艰巨而可怕的任务,要弄清我们到底是什么。 海德格尔也通过他dasein概念来解决这个问题。

Yet in the age of Aristotle, these kinds of existential questions were at the front and center of human life and experience. Aristotle believed that the truly happy person could only rightly be judged on his deathbed after he had accumulated and exhibited a life-long tendency towards virtuous behavior. A virtuous person is defined as one who simply acts virtuously, without having to weigh the pros and cons of each situation.

然而,在亚里士多德时代,这些存在性问题处于人类生活和经验的前沿和中心。 亚里士多德相信,只有在他积累并表现出终生的良性行为倾向之后,才能真正对真正快乐的人做出判断。 善良的人被定义为仅行善良的行为而不必权衡每种情况的利弊的人。

If we allocate these fundamental questions of being and doing to the “dogma” and “authority” of the designers of recommender systems, then we lose an essential aspect of what it means to be a person capable of reflective judgment. In the limit, the outsourcing of such questions to the designers of such algorithms threatens to turn us into mere fleshly automata whose purposes are no greater than their collective role in the advancement in GDP growth and stock-market valuations.

如果我们将这些关于存在和执行的基本问题分配给推荐系统设计者的“教条”和“权威”,那么我们将失去一个具有反思能力的人意味着什么的基本方面。 在极限情况下,将此类问题外包给这类算法的设计者有可能使我们变成纯粹的肉体自动机,其目的不超过其在GDP增长和股票市场估值中的集体作用。

上帝:对复杂问题的简单答案 (God(s): Simple Answers to Complex Questions)

The natural world is a complex and uncertain place. The history and evolution of science is a testament to this fact. We’re constantly updating our best picture of what reality really is like. It’s not surprising that a major function of religion is to give us a rudimentary and simple cognitive tool for understanding the complex natural processes around us, such as hurricanes, meteor showers, the movement of celestial objects, and the like. Similarly, the tradition of Christian theodicy aims to explain how evil could co-exist with a God who was obviously omniscient, omnibenevolent, and omnipotent.

自然世界是一个复杂而不确定的地方。 科学的历史和发展就是这一事实的证明。 我们不断更新的现实是什么真的就像是我们最好的图片。 宗教的主要功能是给我们一个基本而又简单的认知工具来理解我们周围复杂的自然过程 (如飓风,流星雨,天体运动等)也就不足为奇了。 同样,基督教的传统自然神学目的来解释罪恶如何能共存与上帝谁明显无所不知,omnibenevolent的,万能的。

Let’s look at one salient example of how religion reduces uncertainty. The Book of Genesis tells an extremely simple story about arguably the most complex event imaginable: the creation of the world. One measure of its simplicity is that a small child can understand it. Meanwhile, cosmologists and geologists tell us the Earth is roughly 4.5 billion years old. In fact, without our modern, scientific understanding of the physical process of radioactive decay, we might not even know this basic fact about our planet.

让我们看一个宗教如何减少不确定性的重要例子。 《创世记》讲述了一个非常简单的故事,它可以说是可以想象的最复杂的事件:世界的创造。 简单性的一种衡量标准是,一个小孩可以理解它。 同时,宇宙学家和地质学家告诉我们地球大约有45亿年的历史 。 实际上,如果没有对放射性衰变物理过程的现代科学认识,我们甚至可能不知道有关我们星球的这一基本事实。

Without scientific training, it’s hard to wrap your head around terms like “isotopic analysis” and numbers like 4.5 billion, but it’s easy to imagine six days of work and a day off on the seventh. That’s what Daddy does every week.

没有科学的培训,很难像“同位素分析”这样的术语和诸如45亿这样的数字缠住你的头,但是很容易想象六天的工作和第七天的休息。 那就是爸爸每周做的事。

Religions also provide their followers with simple answers to complex moral arguments, such as in the abortion debate, where thorny questions of personhood must be faced. Can something be 50% of a person? What would that even mean?

宗教还为追随者提供了对复杂道德论点的简单答案,例如在堕胎辩论中,必须面对棘手的人格问题。 可以占一个人的50%吗? 那甚至意味着什么?

By following the received doctrine of our religion, we can get quick and clear answers to these kinds of tricky moral questions. Pope Benedict, for instance, has told his Catholic followers that “life begins at conception.” After being told this, the uncertainty surrounding our belief that abortion is morally wrong completely evaporates. We are now able to focus our mental energies in other areas of life as the result of this evaporation of moral uncertainty.

通过遵循我们所接受的宗教教义,我们可以对这些棘手的道德问题得到快速而明确的答案。 例如,教皇本尼迪克特(Benedict)告诉他的天主教信徒,“ 生命始于受孕。 被告知后,围绕我们认为堕胎在道德上是错误的信念的不确定性完全消失了。 由于道德不确定性的蒸发,我们现在能够将精力集中在生活的其他领域。

确定性的认知魅力:以个人身份为例 (The Epistemic Allure of Certainty: The Case of Personal Identity)

Sometimes simple answers to complex questions are practically necessary. Persons, for example, must be considered as discrete wholes. You can’t buy half an egg at the store, after all.

有时实际上需要对复杂问题的简单答案。 例如,必须将人视为离散的整体。 毕竟,你不能在商店买到半个鸡蛋

The philosopher Derek Parfit famously argued against conceiving of personal identity as an all-or-nothing affair whose determinations hinged on questions of “survival.” For example, if one were to undergo brain fission or fusion experiment, would you survive? Would there be two of you? None of you? Parfit argues this way of thinking about identity is not productive.

哲学家德里克·帕菲特 ( Derek Parfit)著名地反对将个人身份视为全有或全无的事情,其决定取决于“生存”问题。 例如,如果要进行脑裂变或融合实验, 您会生存吗? 你们两个人 ? 你们都不是 帕菲特(Parfit)辩称,这种关于身份的思考方式无济于事。

Image for post
Parfit imagined personal identity as something that could “branch” into many possible future selves. Personal identity would be defined by the “relation-R” between a current self and past or future self. Photo by Tim Johnson on Unsplash
帕菲特(Parfit)将个人身份想象为可以“融入”许多未来自我的事物。 个人身份将由当前自我与过去或未来自我之间的“关系-R”来定义。 蒂姆·约翰逊 ( Tim Johnson)Unsplash

Parfit instead proposed a “Relation-R,” which permits of degrees of personal identity, operationalized via psychological continuity. Importantly, this relation must not be one-to-one. In other words, there may be two persons psychologically continuous with another person if they meet the right conditions. For instance, our current self at time t will be in different degrees of relation R with various past selves (time t-1, t-2,…, t-n), depending on the psychological continuity of our present with past selves. After traumatic events or religious conversions, for instance, what might survive or die is enough to constitute a new and different person in Parfit’s view.

相反,Parfit提出了一种“ Relation-R”,它可以通过心理连续性实现一定程度的个人身份认同。 重要的是,这种关系不能一对一。 换句话说,如果两个人符合正确的条件,他们可能会在心理上与另一个人连续。 例如,根据当前与过去自我的心理连续性,我们在时间t处的当前自我与过去的各种自我(时间t-1t-2 ,…, tn )之间的关系程度R会不同。 例如,在发生创伤性事件或宗教religious依之后,可能幸存下来或死亡的东西足以构成一个新的和不同的人,在帕菲特看来。

Of course, such philosophical views are logically appealing but practically (and legally) unfeasible. Only one person can go to jail, even if we might agree that the same person who committed a crime 20 years and underwent religious conversion is, in essence, a “new person.” Parfit might argue there is a clear psychological connectedness between these selves at different times, but changes in beliefs, values, and behavior may be radically different, more so than between two different persons. For moral and legal reasons we, unfortunately, cannot split the person (after all, we call him an individual) to reflect such a reality. We are thus held socially and morally responsible for acts we commit, even if those acts were committed long ago, by a former self-bearing little psychological resemblance to the person who exists today.

当然,这种哲学观点在逻辑上很有吸引力,但实际上(和法律上)是不可行的。 即使我们可以同意,一个犯罪20年并进行宗教信仰改造的人从本质上讲是一个“新人”,但只有一个可以入狱。 帕菲特可能会辩称,这些自我在不同时间之间存在明显的心理联系,但信念,价值观和行为的变化可能根本不同,比两个不同的人之间更是如此。 对于道德和法律的原因,我们遗憾的是,不能分割的人(毕竟,我们叫他灵敏度独立),以反映这种现实。 因此,即使我们的行为是很久以前实施的,我们也要在社会和道德上承担责任,因为他们过去的自生心理与今天的人几乎没有相似之处。

In the real world decisions must be made. And often times the real world presents us with only two options: Go or stay? Buy or sell? Love or hate? Cry or laugh? Fetus or person?

在现实世界中,必须做出决定。 通常情况下,现实世界只为我们提供两种选择:去还是留下? 买还是卖? 爱或恨? 哭还是笑? 胎儿还是人?

贝叶斯大脑:宗教信仰更新的案例 (Bayesian Brains: The Case of Religious Belief Updating)

Have you ever wondered why you can’t tickle yourself?

您是否曾经想过为什么不能给自己挠痒痒?

The philosopher Andy Clark has famously argued that human brains’ traffic chiefly in generative probability density functions, which are updated — via a form of neuronally-based gradient descent — when prior expectations meet new sensory stimuli. Our brains engage in a form of predictive coding of low-level sensory input, and when these predictions don’t pan out, our brains continually adjust them until they do. Such a model of brain function has been used to explain experimental results in perceptual illusions (i.e., placing strong prior on the principle of “only one object can exist in one place at a time”) and psychological phenomena such as schizophrenia (i.e., misplaced priors).

哲学家安迪•克拉克(Andy Clark)著名地指出, 人脑的流量主要是通过生成的概率密度函数进行的 ,当先前的期望遇到新的感觉刺激时,该函数会通过基于神经元的梯度下降的形式进行更新。 我们的大脑参与了低水平感觉输入的一种预测编码形式,当这些预测无法实现时,我们的大脑会不断对其进行调整,直到达到预期为止。 这种大脑功能模型已用于解释感知错觉的实验结果(即,在“一次只能在一个地方存在一个物体的原则上”强先验)和诸如精神分裂症(即错位)的心理现象先验)。

The above formulation can account for the connection between perception and action as well. The philosopher of cognition Jakob Hohwy put things as follows in his 2014 book The Predictive Mind:

上面的表述也可以解释感知和行动之间的联系。 认知哲学家雅各布·霍威(Jakob Hohwy)在其2014年的著作《预测的心智》中 写道

The basic idea would be that the brain uses a specific hypothesisabout the world, predicts what the sensory input would be like were thishypothesis true, and then actively samples the world selectively to get this predicted sensory input. In other words, the brain generates a fantasy, a set of predictions that now do not fit with the current sensory input. This induces a prediction error that can be minimized by turning the fantasy into reality, that is, by acting to bring oneself into the predicted situation (pg. 78, emphasis mine).

基本思想是,大脑使用关于世界的特定假设,预测假说为真时的感觉输入,然后选择性地主动采样世界以获取此预测的感觉输入。 换句话说,大脑会产生幻想,即一组现在不适合当前感觉输入的预测。 这导致可以通过转动幻想变成现实,那就是,通过作用把自己放入预测情况(第78页,重点煤矿)最小化的预测误差。

To this, I would add that religious prior beliefs are some of our deepest and most fundamental beliefs. For a religious person, the entirety of their experience is seen through the lens of these beliefs. Their religious beliefs color their experience in a distinctive way that makes right clear from wrong, good clear from the bad.

对此,我还要补充说,先验的宗教信仰是我们最深层和最根本的信仰。 对于一个宗教人士来说,通过这些信仰的视角可以看到他们的全部经历。 他们的宗教信仰以一种独特的方式为他们的经历增色添彩,使对与错,对与错变得清晰。

In the language of Bayesian belief updating, the effects of a religiously-informed prior totally swamp that of the likelihood, leaving us with a posterior more or less in the same shape of the prior. The religious person is therefore less likely to change her mind about any issue where her religious beliefs place can be modeled with peaked prior.

用贝叶斯信念更新的语言,一个有宗教知识的先验的影响完全淹没了可能性的影响,使我们或多或少地处于与先验相同形状的后验。 因此,对于任何可以用先验高峰来模拟其宗教信仰地点的问题,宗教人士都不太可能改变主意。

As an example, if we were to imagine the answer to a single question (e.g., “How old is the earth?”) as a discrete probability distribution with the elements of the set as answers, nearly all the mass would be centered over the received wisdom of our religion. We would see nearly all the probability mass over “6,000 years ago,” and essentially none over the hypothesis “4.5 billion years ago.” Such is the power of religious belief in shaping our experience.

例如,如果我们将单个问题的答案(例如,“地球几岁?”)想象成一个离散的概率分布,并且以集合中的元素作为答案,那么几乎所有质量都将集中在接受了我们宗教的智慧。 我们几乎可以看到“ 6000年前 ”的所有概率质量,而在“ 45亿年前 ”的假说中几乎看不到。 这就是宗教信仰塑造我们经验的力量。

The image below illustrates how these prior distributions might look when modeled with a Beta distribution with different alpha and beta parameters. Note that the image is of a continuous distribution, but our example concerns a discrete one: the purpose is just to highlight the various shapes such distributions could take. The point is that the shape of prior will differ drastically for religious and non-religious people. In the extreme, the skeptic will assign a uniform prior.

下图说明了使用具有不同alpha和beta参数的Beta分布建模时,这些先验分布的外观。 请注意,图像是连续分布的,但我们的示例涉及的是离散的图像:其目的只是突出显示这种分布可以采用的各种形状。 关键是,先验的形式对于宗教和非宗教人士将有​​很大的不同。 在极端情况下,怀疑者会分配统一的先验。

Image for post
Religious absolutists would tend to assign a very high probability mass to one single belief on a topic. Skeptics and relativists will tend to have flat, uniform distributions of probability mass over their sets of beliefs (relativism). Source: Wikicommons.
宗教专制主义者倾向于将非常高的概率分配给一个话题的单个信念。 怀疑论者和相对论者倾向于在他们的信念集合上相对平坦地分布概率质量(相对论)。 资料来源:维基百科。

天生的信徒和故意立场 (Born Believers and the Intentional Stance)

Although religious beliefs today might seem to be antiquated in some ways (when was the last time you followed the commands in Leviticus 19:19 not to wear fabrics made of two kinds of fibers?), they are a surprisingly fertile means of making useful inferences concerning the behavior of the natural world. An emerging stream of research on the cognitive science of religion may provide us with an explanation of why religious beliefs are so easy to acquire and difficult to abandon.

尽管今天的宗教信仰似乎在某些方面过时了(当您最后一次遵循利未记19:19的命令时不穿两种纤维制成的织物吗?),但它们却是令人惊讶的肥沃手段,可以做出有用的推论。关于自然世界的行为。 关于宗教认知科学的新兴研究可以为我们解释为什么宗教信仰如此容易获得和难以放弃。

Cognitive scientist Justin Barrett claims we are born believers in gods (and God) just as human children are “born walkers” or talkers. Barrett, on the strength of his research data collected from children all over the world, contends that mere indoctrination by parents is not enough to explain the human predilection for religious thinking. Children apparently come into the world with a belief structure already attuned to perceive intentional agency and impose order and design on the world.

认知科学家贾斯汀·巴雷特(Justin Barrett)声称我们是神(和神)的天生信徒 ,就像人类的孩子是“天生的步行者”或健谈者一样。 巴雷特凭借他从世界各地儿童那里收集的研究数据的力量,认为仅仅父母的灌输不足以解释人类对宗教思想的偏爱。 孩子们显然已经以一种信念结构进入了世界,这种信念结构已经过调整,可以感知故意的行为,并在世界上施加秩序和设计。

Why would this be? I suggest it’s to provide comfort during times of chaos. And on an evolutionary scale of time, humans have had little understanding of the underlying causes behind much of what they experience in nature. Barrett and Lanman (2008) explain the connection between religion and complex natural environments as follows:

为什么会这样呢? 我建议在混乱时提供舒适感。 而且,随着时间的推移,人类几乎不了解他们在自然界中经历的许多根本原因。 Barrett和Lanman(2008)解释了宗教与复杂自然环境之间的联系,如下:

Religion springs naturally from the way ordinary human cognitive systems interact with ordinary human social and natural environments… Religious beliefs arise (at least in part) because they are natural outputs of human cognitive systems solving ordinary problems.

宗教自然源于普通人类认知系统与普通人类社会和自然环境的相互作用……宗教信仰的产生(至少部分是因为)是解决普通问题的人类认知系统的自然产物。

So early humans hit upon one way of making complex natural processes interpretable and thus less cognitively uncomfortable: imposing intentional agency on them. In other words, they ascribed a folk psychology of mind to natural events by understanding them in terms of easily understandable beliefs, desires, and goals. Because we too possess the same intentional agency, we can use this analogical cognitive structure to make inferences about both the causes and future behavior of natural processes. We can thus bootstrap our innate ‘theory of mind’ to explain and predict various natural phenomena that we might otherwise not understand. Daniel Dennett has termed this approach the “intentional stance,” as it gives us an inferential framework for the interpretation of complex behavior in persons.

因此,早期人类想到了一种使复杂的自然过程可解释且因此在认知上不那么舒服的方法:对它们施加故意的作用 。 换句话说,他们通过容易理解的信念,欲望和目标来理解自然事件,从而将民间心理归因于自然事件。 因为我们也拥有相同的故意代理,所以我们可以使用这种类比的认知结构来推断自然过程的原因和未来行为。 因此,我们可以引导我们与生俱来的“心理理论”来解释和预测我们可能无法理解的各种自然现象。 丹尼尔·丹尼特(Daniel Dennett )将这种方法称为“故意立场”,因为它为我们提供了一种解释人的复杂行为的推断框架。

As Justin Barrett again writes:

正如贾斯汀·巴雷特(Justin Barrett)再次写道:

By virtue of being agents with special knowledge and power, gods may be used to reason about great fortune and misfortune, and be connected with human moral failings or triumphs as causes of otherwise inexplicable fortune or misfortune.

由于是具有特殊知识和能力的代理人,神可被用来推理巨大的运气和不幸,并与人类道德上的失败或胜利联系在一起,这是原本无法解释的财富或不幸的原因。

Image for post
An example of ascribing intentional agency to the moon: The moon “wants” to cover the sun. Photo by Jongsun Lee on Unsplash
赋予月球故意代理的一个例子:月亮“想要”覆盖太阳。 李宗孙在《 Unsplash》上的 照片

现在怎么办? 培养实践推理能力 (What Now? Developing Your Capacity for Practical Reasoning)

So, to summarize, here’s the connection between recommender systems and religion: they’re both solutions to cognitive overload. Religions reduce uncertainty by placing strong priors on specific moral and metaphysical beliefs, while recommender systems, thanks to the economic interests (e.g., maximize engagement, click-through rate, and time on site) of their designers, effectively place strong priors on what they want to you to do, see, and buy. These priors have the effect of filtering and shaping our experience of reality in ways we may not be aware of.

因此,总而言之,这是推荐系统与宗教之间的联系:它们都是认知超负荷的解决方案。 宗教通过在特定的道德和形而上学信念上放有先验优势,从而减少了不确定性;而由于其设计者的经济利益(例如,最大限度地提高参与度,点击率和现场停留时间),推荐系统有效地将了先验优势放在了他们的设计上希望您去做,看,买。 这些先验会以我们可能不知道的方式过滤和塑造我们对现实的体验。

Your experience of digital reality is thus a reflection of the goals and interests of the designers of digital platforms. Your possibilities for action in this digital reality are limited by the assortment of digital affordances on these platforms. Your limited capabilities for action therefore constrain the possibilities of who or what you can become. In digital reality, what you believe to be your most unique and defining features as an individual may, in the end, count for nothing. When intentions are ignored, all behavior looks the same.

因此,您对数字现实的体验反映了数字平台设计人员的目标和兴趣。 在这些数字现实中,您采取行动的可能性受到这些平台上各种数字能力的限制 因此,您有限的行动能力会限制您成为谁或成为什么样的人的可能性。 在数字现实中,您认为最独特,最具有个人特色的功能最终可能一无是处。 当意图被忽略时,所有行为看起来都是一样的。

This naturally raises the question of whether persons might one day have finer, personalized control over the parameters of the recommender systems they engage with. If so, they will need to be able to answer questions about what kind of person they wish to be, the priority of certain personal values, and which kind of actions are better and worse than others.

这自然提出了一个问题,即人们是否有一天可能会对与其使用的推荐系统的参数进行更好的个性化控制 。 如果是这样,他们将需要能够回答有关他们希望成为什么样的人,某些个人价值观的优先级以及哪种行动比其他行动更好和更坏的问题。

As I’ve tried to explain above, there is good reason to believe that we humans have a fundamental psychological need for certainty that could explain our inherent predilection to religiosity. As the European “Dark Ages” can attest to, the apparent trade-off for having intense religious “priors” was hundred of years of scientific stagnation. Is it mere coincidence that formal notions of probability appeared right as religious thinking (in the West at least) began to decline? Even today, the average person has a hard time understanding the nature of probability. The Monty Hall problem stumped some seriously bright people. Statisticians still fight over what probability actually is. The great Einstein was loathed to believe “God played dice.”

正如我在上面试图解释的那样,有充分的理由相信我们人类对确定性具有基本的心理需求,这可以解释我们固有的宗教信仰倾向。 正如欧洲“黑暗时代”可以证明的那样,拥有强烈的宗教“先验”的明显权衡是数百年的科学停滞。 当宗教思想(至少在西方)开始衰落时,形式的概率概念正确地出现只是偶然吗? 即使在今天,普通人仍然很难理解概率的本质。 蒙蒂·霍尔 ( Monty Hall)问题困扰了一些非常聪明的人。 统计人员仍在争辩实际的可能性多少。 伟大的爱因斯坦不愿相信“上帝扮演骰子”。

But as religiosity in the West appears to be in decline, the use of recommender systems in digital environments appears to be growing — with no end in sight. Millions of people, especially young people, use recommender systems every day on platforms and apps like Youtube and Instagram. We’ve read news reports about young people falling prey to extremist ideology after entering a filter bubble of extremist Youtube content, for instance. Reducing information overload comes with a price. That price may be a weakening of an invaluable capacity for critical thinking, necessary for a healthy and functioning democracy.

但是,由于西方的宗教信仰似乎正在下降,因此在数字环境中推荐系统的使用似乎正在增加,而且还没有结束。 每天都有数百万人,尤其是年轻人,在Youtube和Instagram等平台和应用上使用推荐系统。 例如,我们读到有关年轻人进入极端主义Youtube内容的过滤泡后沦为极端主义思想的新闻报道 。 减少信息过载是有代价的。 这种代价可能削弱了批判性思维的宝贵能力,而批判性思维对于健康和运转的民主制度是​​必不可少的。

Kant understood the Enlightenment to be embodied by the phrase “dare to know!”

康德将启蒙运动理解为“ 敢于知道!”。

By outsourcing opportunities for practical reasoning about what kind of persons to become and what kind of actions to engage in, we are limiting the development of one of our most fundamental human capacities.

通过将机会推理的机会外包出去,以了解要成为什么样的人以及要从事什么样的行动,我们就限制了我们最基本的人的能力之一的发展。

This isn’t to say that you shouldn’t use recommender systems. They’re quite handy and convenient. I use them all the time, too. Rather, this is merely a warning that in doing so, you risk potentially losing your ability to think, reason, and face uncertainty in a way that does not come naturally to humans. Hundreds of thousands of years passed by before anyone had the idea to think that thinking itself, removed from dogma and superstition, might be a valuable capacity worth protecting. For older adults born way before the digital age, this may not matter. But for digital natives, such capacities for critical thinking and practical reasoning may seriously be at stake.

这并不是说您不应该使用推荐系统。 它们非常方便和方便。 我也一直使用它们。 相反,这仅是警告,这样做有可能导致您丧失思考,推理和面对不确定性的能力,而这是人类不自然的。 数十万年之前,任何人都没有想到过,摆脱教条和迷信的思想本身可能是值得保护的宝贵能力。 对于在数字时代之前出生的老年人来说,这可能并不重要。 但是对于数字原生代来说,这种批判性思维和实践推理的能力可能会受到严重威胁。

My recommendation: don’t be so quick to let corporations rob you of a chance to exercise your capacity for practical reasoning. Part of what it means to be a human is to possess the capacity to choose what kind of person to be and what kind of acts to perform.

我的建议 :不要这么快就让公司抢走您进行实际推理的能力。 成为一个人意味着什么的一部分就是拥有选择什么样的人和什么样的行为的能力。

翻译自: https://medium.com/towards-artificial-intelligence/religious-beliefs-and-recommender-systems-solutions-to-the-same-problem-e07b603afd8c

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值