2014年10月24日,星期五

体现还是象征?谁在乎!

I still don't understand the hype over 体现的 认识. It's too abstract a concept for me, I guess.  I need more grounding in the real world. (Am I getting it?) So let's consider a real world example of neural 计算. For the record, this 是 partially excerpted/paraphrased from a discussion in 镜像神经元的神话.

就神经计算而言,在谷仓猫头鹰中的声音定位是相当容易理解的。 两只耳朵的输入汇入脑干 椎板核 具有“延迟线”架构,如图所示:
 


在这种安排下,左耳和右耳信号将在其上的神经元(圆圈) converge 同时 将取决于两者激发之间的时间差 耳朵。如果同时刺激两只耳朵(声音在前面),会收敛 发生在 延迟线。如果 声音首先刺激左耳,在此位置会向右偏远 原理图(左耳刺激较早到达,使其信号进一步向下 右耳讯号之前的连线)。如果右耳刺激到来,反之亦然 早点这种延迟线架构从根本上建立了一个符合检测器阵列 检测巧合的单元格位置 代表信息: 两只耳朵的刺激时间不同,因此声音的位置也不同 资源。然后,您要做的就是将各个单元的输出(点火模式)插入 排列到用于控制头部运动的电机电路中,您就有了神经网络 用于检测声源位置和朝向声源的方向。
 
Question: what do we call this kind of neural 计算?  Is it 体现的? Certainly it takes advantage of body-specific features, the distance between the two ears (couldn't work without !) 和 I suppose we 能够 talk of a certain "resonance" of the external world with neural activation.  In 那 sense, it's 体现的.  On the other hand, the network 能够 be said to 代表信息 in a neural 码--the pattern of activity in network of cells--that no longer resembles the air pressure wave 那 gave rise to it.  In fact, we 能够 write a symbolic 码 to describe the 计算 of the network.  Typical math models of the 处理 use cross-correlation but you 能够 do it with some basic 码 like this:


x =在左耳检测到声音发作的时间 
y =在右耳检测到声音发作的时间 
如果 x = y, 然后写‘straight ahead’
如果
x < y, 然后写‘left of center’
如果 x > y, 然后写‘right of center’ 

虽然谷仓猫头鹰中没有代码’s brain, the 建筑 网络的确 实现程序: x y 是来自左侧的输入信号(轴突连接) 和右耳; x和y之间的关系是 计算的 通过延迟线重合 detectors; 和 “rules”通过连接实现用于生成适当输出的功能 在阵列中的各个单元和电机系统之间(在我们的示例中)。大脑和线条 of 码 能够 indeed implement the same 计算al program. Lines of 码 do it with a particular arrangement of symbols 和 规则, brains do it with particular arrangement of 神经元之间的联系 要么 代表信息. Both are accurate ways of describing the 计算s 那 the system carries out to perform a task.  

Does it matter, then, whether we call this non-representational 体现的 认识 要么 classical symbolic 计算?  I think 不.  If we simply start trying to actually figure out the 建筑s 和 计算s of the system we are studying, the question of what to call it becomes trivial.  

28条评论:

安德鲁说过...

The 体现的 vs symbolic 是sue 是 actually this: what 是 the system actually doing in 要么 der to solve the problem? As you 不e, the owl 是 计算该算法。取而代之的是,它连接了神经元,以便以某种方式做出响应。该系统也不代表该信息。相反'连接到其他具有体系结构的系统'的响应以特定方式使用。

Part of the problem with killing off 计算al, representational stories 是 那 you 能够 add them as a gloss to any activity. This 是 why the 体现的 reanalysis always just sounds like a redescription (but what it actually means 是 那 representations aren'做很多实际的工作)。当两个帐户做出不同的预测而一个帐户比另一个帐户更好时,当然就需要艰苦的工作。重读Louise Barrett'关于Portia Spider和Webb的章节'关于蚱sound的声音本地化的工作;这些都是很好的例子,说明了如何使用计算符号帐户't lead to the right questions, while the 体现的 account does.

当然,神经系统会为行为做出重要贡献。但它's job 是 不 to abstract away from the input, which 是 what representational accounts all assume. This barn owl neural 建筑 是 intimately tied to the nature of the information it interacts with 和 the role it needs to play in shaping behaviour. The solution 是 specific to these, 和 是 implemented as a neural 处理 with very specific dynamic characteristics 那 are tuned to the task. That 处理 能够 be described 计算ally but 是 不 in 和 of itself a 计算. The 体现的 analysis 是 interested in what kind of neural systems 能够 be created 和 maintained by the 要么 ganism'与之互动's environments, 不 what kind of general purpose descriptions 能够 be applied, because from the point of view of the 要么 ganism, the former 是 the only thing 那 might lead to 那 neural 建筑.

格雷格希科克说过...

Thanks for this clarification. 我不't think I disagree with you except in terminology. So, 能够 I push you to clarify a couple of additional points?

1.神经元能计算吗?
2.神经元是否处理信息?
3.谁'您在反对自己的象征立场吗?

我问第三个问题,因为即使是古典符号/计算理论家,例如纽厄尔&西蒙(Simon)写了几行符号计算机代码作为解决人类问题的模型,他似乎并没有接受您所反对的立场。

Newell等。写,"Digital computers come into the picture only because they 能够, by appropriate programming, be induced to execute the same sequences of information 处理es 那 humans execute when they are solving problems. Hence, as we shall see, these programs describe both human 和 machine problem solving 在 the level of information 处理es."

让我也将您推到另一点。你写的是神经系统's "工作不是从输入中抽象出来。"在这一点上您是否会坚持到底,即使抽象会增强 "它在塑造行为中需要发挥的作用"?还是在这种情况下抽象可以吗?

沃利·斯通说过...

大家好,

很棒的帖子。我想知道我们是否应该认真对待劳伦斯·夏皮罗'试图理解猫头鹰解释时体现假设之间的区别'合理的本地化(以及所有人类行为)。他特别建议区分'替代假设'这就意味着不必进行计算和表示的讨论来描述诸如声音本地化之类的行为。我认为有很多人赞成这种方法,尤其是像Anthony Chemero这样的人所提出的概念。我认为这些论点在很大程度上是认识论性质的,而不是关于本体论的。在跟随安德鲁's posts, I think he 是 sympathetic to the 替代 idea?

夏皮罗(Shapiro)也将'概念化假设',其中的计算和表示对于认知很重要,但是通过感官运动系统中的模拟来表示事物。这更符合Barsalou'的想法。猫头鹰的描述's sound localization circuitry 能够 easily exist in this type of 体现的 system. The prediction 是 那 when the owl 认为 of previously localized sounds, 要么 how it might localize some future sounds, a simulation in this circuitry underlies her ability to do so.

我觉得直到我们开始区分这些类型为止(我们是否都采用了Shapiro's 要么 不), progress will be difficult. The general idea of 体现 (i.e. the ways in which the body 和 environment shape 和 constrain behavior) 是 a powerful tool for investigating human psychology, 和 it would be a shame to over interpret it 要么 dismiss it too abruptly.

你怎么看?

格雷格希科克说过...

But what progress are we trying to make? We already 知道 how barn owls localize sound in terms of neural 计算. Why should we care whether we call it 体现的 要么 symbolic?

我们当然可能想问其他问题,例如(如您所述)猫头鹰时是否重新激活同一脑干电路"thinks"关于位置。但是请注意,假设关于位置的思考涉及"simulating"原始的感官体验并没有't *本身*解释什么'发生了。为此,我们需要了解原始的感觉过程(延迟线,"convergence neurons" serving as 码s for locations, etc.). When we have the low-level stuff worked out, then we 能够 say, with substance, 那 thinking 关于 location involves re-activating neurons 那 码 location in the brainstem delay-line neural networks. Otherwise, it'只是一个模糊的假设,是寻找答案的一种启发,而不是一种解释。

现在,假设合理的本地化和对本地化的思考都像这样,是体现的还是象征性的?我不'小心。但是,我*会*认为系统正在处理信息。它正在获取有关两只耳朵的声音到达时间(或水平)差异的信息,并将其转换为位置代码。这是标准的信息处理心理学,因为自认知(即信息处理)革命以来已在模型中实现了这种心理学。什么's new with 体现的 认识, as far as I 能够 tell, 是 the claim 那 low level 信息处理 是 important for, indeed part of, "high-level"信息处理(例如,' thinking 关于'或概念化)。它's an interesting new way of modeling 高水平 处理es, one 那 needs empirical verification, but 不 a fundamental shift in paradigm.

多米尼克·卢克说过...

All of this discussion leaves me a bit puzzled as to what happened to the actual fundamental 体现 of human semantics. All of our meaning making 是 based on a human scale experience. For example, if I say "我将头撞到天花板上,当我抬起头时,我无法达到。" I immediately perceive the paradox. I talk 关于 things with respect to my field of vision (get out of my sight), have events back to back, etc. I 能够 tell the difference between The Normans conquering Britain 和 The Beatles conquering America. I instinctively 知道 when it makes sense to talk 关于 me standing in front of a Church 和 the Church being behind me.

It 是 this sort rich bodily 知道ledge of the world 那 makes 认识 possible. And as a linguist when I speak of 体现, 那's what I'd请牢记。而且,所有这些都不能作为一系列算法计算来解决。但是在所有讨论中,我不't see any suggestion 那 this 是 being discussed. I first came across 体现的 认识 in the 80s through cognitive semantics (Lakoff, etc.) 和 have yet to find any psychological 要么 even neural account of this. Nothing offered here even hints 在 dealing with the massive complexity 那 是 体现的 language.

However, I do wonder if it does make sense to talk 关于 体现 在 the level of neural 处理ing. But 那 处理ing does need to account for our ability to instantly reflect the world'我们处理语言的方式在人类层面的复杂性。

沃利·斯通说过...

嗨,格雷格,

感谢您的回复!我只想提出几点。

###
...但是请注意,假设关于位置的思考涉及"simulating"原始的感官体验并没有't *本身*解释什么'发生了。为此,我们需要了解原始的感觉过程(延迟线,"convergence neurons" serving as 码s for locations, etc.). When we have the low-level stuff worked out, then we 能够 say, with substance, 那 thinking 关于 location involves re-activating neurons 那 码 location in the brainstem delay-line neural networks. Otherwise, it'只是一个模糊的假设,是寻找答案的一种启发,而不是一种解释。
###

I agree with you here. At least from the 概念化 perspective this 是 exactly where we are: we have a hypothesis 关于 how how a cognitive task may be preformed (i.e. it 是 supported by simulations in sensorimotor circuits). 我不't anyone adopting the 概念化 approach would have a problem with the idea 那 we need to identify 和 clearly specify what the low-level neural apparatus 是 doing. And right now, this idea of simulation 是 merely a hypothesis, one 那 能够 easily be falsified. I see this as a great opportunity to generate hypotheses 关于 this circuit.

###
现在,假设合理的本地化和对本地化的思考都像这样,是体现的还是象征性的?我不'小心。但是,我*会*认为系统正在处理信息...这是标准的信息处理心理学,因为自从认知(即信息处理)革命以来,它已在模型中实现。什么's new with 体现的 认识, as far as I 能够 tell, 是 the claim 那 low level 信息处理 是 important for, indeed part of, "high-level"信息处理(例如,' thinking 关于'或概念化)。它's an interesting new way of modeling 高水平 处理es, one 那 needs empirical verification, but 不 a fundamental shift in paradigm.
###

许多人说在高级认知任务中使用基于情态的特定信息这一想法并不新鲜(例如'dual coding'假设有点'embodied', 和 imagery too, etc). What 体现 does, 在 least from the 概念化 perspective, 是 extend this to all cognitive tasks-- problem solving, emotional 处理ing, memory, etc. This 是 a less drastic shift in cognitive psychology than the claims of, for instance, proponents of the 替代 hypothesis who suggest representations are 不 needed to explain behavior. Whether these represent completely different paradigms, 要么 exist on a spectrum of 'embodied'范式是另一个问题...

您问为什么我们应该关心我们如何解释猫头鹰'声音的本地化(象征性或体现性)。我认为在框架中解释这一点很重要:a)生成关于我们在声音定位电路中观察到的与猫头鹰的关系的特定假设's behaviour (e.g. simulations in this circuit are needed for the owl to form memories of previously encountered sounds), b) allows us to identify features of behaviour 要么 the environment 那 are the most relevant for understanding this cognitive skill (e.g. what 关于 the environment makes this specific circuit necessary for keeping the owl alive 和 doing its thing?) 和 c) 那 gives us ideas 关于 the causal role a certain 计算al circuit plays in different cognitive tasks (i.e. do you NEED the modality specific sound localization circuit to think 关于 和 form memories of previously experienced, localized sounds). I think 那 certain subtypes of the 体现 thesis (the 概念化 hypothesis in particular) does these things quite nicely by directing our 在tention to different skills 和 tasks 那 might 不 be acknowledged from a purely symbolic, 计算al, amodal, mental-activity-is-all-in-the-head kind of approach.

安德鲁说过...

1.神经元能计算吗?
I think they 能够 be described as computing, but actually I think they are messy biological systems 那 are 不 just trading activity 那 能够 be 码d as 0 和 1. From my limited 知道ledge of the biology, we now 知道 neurons are in all kinds of continuous contact with each other via electrical 和 chemical signalling, modulated by changing gene expression which reflects current environmental demands etc etc etc. Computation doesn'甚至似乎都无法描述这种非线性动力学 处理.

2.神经元是否处理信息?
是的,但是这些信息的性质尚待掌握。它'不会成为Shannon信息-像该框架一样强大'从现实世界中太抽象了。

3.谁'您在反对自己的象征立场吗?
我问第三个问题,因为即使是古典符号/计算理论家,例如纽厄尔&西蒙(Simon)写了几行符号计算机代码作为解决人类问题的模型,他似乎并没有接受您所反对的立场。
Newell等。写,"Digital computers come into the picture only because they 能够, by appropriate programming, be induced to execute the same sequences of information 处理es 那 humans execute when they are solving problems. Hence, as we shall see, these programs describe both human 和 machine problem solving 在 the level of information 处理es."

这句话很棒,因为它'不仅仅是对实际情况的中立描述;它's dripping with epistemology! Just because you 能够 break down a 处理 into algorithmic steps 和 implement those steps in 码 doesn'不一定告诉您有关生物系统如何真正解决任务的任何信息。路易丝的Portia蜘蛛玩意'这本书就是一个很好的例子:它看起来就像是蜘蛛'must'正在计划,因为它能够追踪猎物'目前无法看到。它甚至在那儿坐了一会儿'在思考事情。但它'没有运行模拟,它's 不 planning, it'转移到样本光流中,即使它会短暂失去视觉接触,也能使其运动到猎物,并且只有当人们停止假设它正在以计算方式执行任务并开始询问它是否在感知上进行操作时,这种情况才会显示出来。
具体的假设是,我们的行为是由我们与世界的知觉接触引起的,而不是世界的内部表征, 和 there are an ever increasing number of examples 那 this 是 the case. This 是 不 a redescription of the same explanation, it's a different explanation 和 it matters (to answer your main concern) if a) the 体现的 explanation 是 better 和 b) the 计算al explanation actively interferes with you asking the right questions to find the better, 体现的 account.

因此,无论我们的神经元混乱多大,'s 不 computing 和 representing. It 是 处理信息, but we need to be more ecological 关于 what we mean by information in 要么 der to have the right job description for the brain.

在这一点上您是否会坚持到底,即使抽象会增强"它在塑造行为中需要发挥的作用"?还是在这种情况下抽象可以吗?
课程之一'replacement' style 体现的 认识 work we typically point to (eg the a-not-b error, etc) 是 那 what looks like abstraction might be no such thing. So I'd需要具体的例子来尝试回答,即使如此,我'd need to 知道 enough 关于 the example to be able to propose the kind of 体现的 analysis 那 you'd需要首先进行操作,以便将抽象抽象化。

心理学家's fallacy 是 to mistake their description of the behaviour for the mechanism of the behaviour. Good 体现的 research tries to avoid this problem by 不 assuming things like the need for abstraction 要么 '嘿,他们只是必须在计划蜘蛛'.

格雷格希科克说过...

因此,神经元可以计算和处理信息。 ew!认知革命的生命! ;-)(但是后来我注意到你与自己矛盾,并且神经元不存在't computing. I'll go with your first answer.) Given this, the 体现的 movement does 不 represent a fundamental change in conceptualizing how the mind works. It 是 still an 信息处理 device 那 transforms information from the senses in various ways to accomplish various tasks. What I will credit 体现 approaches with, though, 是 那 they promote the search for lower-level explanations for 高水平 abilities. Some of my own work would certainly qualify in this respect.

回复:您写的问题3,"具体的假设是,我们的行为是由我们与世界的知觉接触引起的,而不是世界的内部表征"这是一个奇怪的主张。知觉科学家同意,知觉是大脑创造的活跃过程。它不是电影屏幕或录音机。这意味着感知涉及世界的内部表示。射击"coincidence neuron"在简单的声音定位模型中,“空间定位”是如何将空间位置转换为神经代码或表示形式的一个示例。"感性接触世界"这不是理论而是虚无的陈述,除非您建议完全撤退行为主义,以使物理世界直接导致行为。那是你的要求吗?我不't think so.

我觉得您完全不赞成这样的想法"computation"发生它必须是干净的0和1。那'就是数字计算机是如何做到的。湿软件的处理方式与众不同,但杂乱无章'仍在处理信息,即计算某种形式的转换。

An example of abstraction for you: it might be useful to recognize different instances of lions, under various lighting conditions, from various angles, 和 with various bits occluded. It might even be useful to recognize the same animal by sound. 那里 是 evolutionary advantage to abstracting across different sensory events 那 cue the same object 和 appropriate responses. Having learned 那 lion A 是 dangerous, we 能够 then generalize to lion B. Can the mind/brain abstract in such a case? Or are we yoked to the physical environment thus treating each lion encounter as unconnected events?

我不同意您对心理学家认为他们对行为的描述是行为的机制(我怀疑您是指实施)的描述。我们完全意识到大脑不't字面上有代码行或数学符号。 (我们也知道,数字计算机不会't either!)

Finally, the spider example 是 a red herring. It shows 那 a 高水平 theory of how 那 spider solves 那 task 是 wrong 和 a lower-level theory works better. This just means 那 information 是 处理ed 和 transformed 在 a lower level. It does 不 show 那 the spider 是 不 representing 和 处理信息 要么 那 高水平 theories aren'在某些领域正确。

安德鲁说过...

那里'这里有很多东西,但逐点走下去可能会在互联网上引起争论:)让我重新关注一下:

这里的问题是'is the 体现的 description just another way of saying the same thing as the 计算al description?'我声称答案是'no'. 让's think 关于 the sound localisation example from your post (it seems like a similar setup to the grasshopper system Louise Barrett reviews).

允许猫头鹰定位声音的神经体系结构是否具有计算能力?它's 不 clear 那 it's computing anything. The neural system behaves in a certain way depending on information 关于 sound passes through it. Where 是 the 计算 happening? Calling this 计算 是 like saying a hammer 是 computing how to fall when I drop it; sure, I 能够 compute what will happen to it given the relevant equations but this 是 不 what the hammer 是 doing. Physicists describe the motion of the hammer in dynamical terms as a 处理 那 unfolds over time under task specific constraints. (Replacement style) 体现的 认识 makes the same move for the same reason. This means we need to describe the relevant constraints 和 how those are assembled into the local dynamic from which behaviour emerges. We talk 关于 this 在我们的论文中 关于我们的四个关键问题。

Does the behaviour of the network represent location? Actually no; what it does 是 shape behaviour (head 要么 ienting) in a 功能性 manner depending on the parameters of the sound input. Calling its behaviour a representation of location adds exactly 不hing to our understanding of the mechanism, 和 worse makes it sound like it has added something. It'仅仅掩盖了一个本来就很有用的故事,只是一团糟。

People forget 那 representations are solutions to a particular problem, namely poverty of stimulus. No poverty of stimulus, no need to represent anything - just detect information using appropriately calibrated systems. This sound localisation system 是 exactly this: a calibrated device 那, when coupled to sound information from the environment it responds in a task specific manner 那 reflects this calibration. This 是 why 体现的 认识 people talk 关于 Watts steam governors 和 极地计s as technological analogies for how things work, rather than computers. The proposed underlying mechanism 是 不 even slightly the same.

This 是 most useful when the mechanisms are predicted to respond differently so you 能够 empirically tell them apart. Understanding fly ball catching as a calibrated interaction with information rather than a 计算 explains why 外野手在弯曲而不是平直的道路上奔跑; this goes for any kind of prospective (online) vs predictive (computational) control set up. The owl 是 controlling head movements prospectively 和 不 predictively; you might want to talk 关于 this as involving a representation but I have no idea why you would bother given 那 representations are never required when prospective control 是 an option.

请注意,我并没有否认大脑的有趣作用。将信息与行为联系起来显然很重要。不过,最好不要将这项工作描述为具有代表性或计算性,并且's the more radical 体现的 argument. 如果 anything I think this work on the sound localisation systems 是 very 体现的 和 actively points to a non-representational, non-computational story being the best way to go.

格雷格希科克说过...

谢谢你,安德鲁。回到基本原则,这是一个好主意’ve明确列出了您的位置。

在我看来,您正在与Straw Fodor战斗,好吧,也许是真正的Fodor或真正Fodor的版本。您的飞球示例说我们不 ’t使用关于速度和角度的感官信息来计算轨迹(在Fodorian中央处理器中),相反,我们移动身体以使球的感知轨迹是笔直的,这恰好确保我们最终在正确的位置出现在正确的位置时间。精细。我赢了’对此表示怀疑。但是您仍然必须进行大量的信息处理—您似乎可以接受信息处理,所以我’m使用该术语而不是计算—只是为了在背景场景中观察球,感知其运动,感知其运动的直线度并生成用于使其保持直线的运动命令等。您在反对特定的信息处理理论,而不是提出一种全新的信息处理方法考虑大脑如何运作。同上巴雷特’s book.

So we agree on how baseball players run down fly balls 和 we agree on how crickets 和 owls localize sound. Our debate, then, 是 不 关于 content. It’s 关于 what you call it: 体现的 要么 something else. Not a very interesting debate.

Now, if you want to deny 信息处理 和 retreat to full blown behaviorism, then we have something fundamental to argue 关于. As far as I 能够 tell from what you are saying, you are squarely in the non-Fodorian cognitivist (i.e., 信息处理 tradition) where we 能够 talk 关于 the merits of different 信息处理 models.

一种更正:计算机本身不是头脑的类比,它是’s the computer program (i.e., the flow of 信息处理).

安德鲁说过...

您似乎可以接受信息处理,所以我’m使用该术语而不是计算.... You are arguing against a particular theory of 信息处理, 不 proposing a whole new way of thinking 关于 how the brain works
好的-告诉我您所说的信息处理是什么意思。在认知科学中,该短语通常表示支持推理机制所必需的一种处理,该推理机制是克服刺激性贫困所必需的表示形式。这当然不是我想的那样。那么,当您使用该术语时,您认为情况如何?

如果你只是说'输入以某种方式变形以形成输出' well, hmm. You might want to think of this as 信息处理 but I wouldn'不一定要。好老 极地计 是一种将输入(车轮上的旋转)转换为输出(面积的测量)的设备,但是'我什么都没做'd obviously want to call 信息处理. This 是 the point of all the dynamical systems examples 和 是 the essence of the different perspective on what's going on.

此外,所需处理的形式取决于您开始使用的信息的形式。平面仪不进行长度测量,因此没有'实施乘法过程以到达区域。仓n也一样;那神经系统没有'例如,似乎没有实现任何三角函数。因此,至少,这是我们的主要论点之一,如果您想表征大脑在做什么,'d最好从信息而不是神经元开始。

One important thing 关于 behaviourism (besides how often it succeeds - way more than the average cognitive experiment :) 是 那 it placed an emphasis on the 要么 ganism'环境是定义的地方'functional' behaviour. You'对那个斯金纳'他的主要弱点是他没有信息论(即我们如何看待环境),而是主要的见识,即我们应该在假设一切都来自大脑之前先看看世界为行为提供了什么。

格雷格希科克说过...

整个过程中有一点小故障"看看世界提供了什么"理念。一些感性的科学家(请参阅唐·霍夫曼'的工作)得出的结论是,世界与我们所感知的不一样。这个想法是感知是一个用户界面,就像您的计算机一样'的桌面环境,旨在隐藏事实。进化模拟证明了这一主张。如果为真,则环境本身就是计算思想的创造。但是那'完全不同的辩论。

安德鲁说过...

一些感性的科学家(请参阅唐·霍夫曼'的工作)得出的结论是,世界与我们所感知的不一样。
在一个层面上,这自然是一个坏主意。如果没有感性信息't really 关于 the way the world 是, then why don'我们死得更可怕吗?

在另一个层面上,这仅仅是 知觉瓶颈. The world 是 dynamical (i.e. best described in terms of both motions 和 forces). Perceptual information 是 kinematic (motions only). So perceptual information 能够not be identical to the world it 是 information 关于. However, it 能够 be specific to 那 world (see Gibson, Turvey, Shaw, Reed 和 Mace plus Runeson); kinematics 能够 specify dynamics.

当然,这意味着我们不't 要么 ganise our behaviour with respect to the world per se, but with respect to the information 关于 the world we are detecting. (This 是 the important bit behaviourism was missing). This idea has a lot of empirical support; I'例如,我在协调动力学方面做了很多工作,并帮助表明相对于相位信息,协调的节奏运动是有组织的。

长话短说,这一点基本上是正确的,但尽管有趣且重要,但有解决方案。

戴维·波佩尔说过...

I guess I always assumed 那 体现的 认识, most crucially, makes claims 关于 the nature of *concepts*, 和 by extension the nature of 思想, language, etc. Concepts - 要么 so I 思想 - are comprised on such a view of (lists of) sensorimotor features, because of the (in my view flawed 和 even false) assumption 那 sensorimotor 处理ing 是 in some sense 'closer' to the world. Sensorimotor 处理ing 是 perhaps 更接近 to the body/world in some completely pre-theoretical 和 (misleadingly) intuitive sense, but when you look under the hood, the infrastructure of sensorimotor 处理ing 是 highly complex, highly abstract, distal to the 知觉/action surfaces, relentlessly inferential, 和 so on.

In any case, if concept=bundle of sensorimotor features, 和 NOTHING ELSE (if there 是 something else, like a prototype, then you have given up on the basic assumptions 和 introduced an abstraction with a different format ... cause you have to specify the axes over which you average), then these features have to have (a) causal force in inference 和 (b) some way of combining, since a/the key ingredient of 认识 和 language 是 compositionality. This latter criterion 是, to my 知道ledge, very very tricky on the most straight-up 体现的 view.

这一切都非常接近* ooold *"意象图论"这有一些严重的缺点。值得注意的是:组成性。

卡拉马扎(Caramazza)和马翁(Mahon)为此写了很多重要的东西,兰迪·加里斯特尔(Randy Gallistel)'s Memory 和 the Computational Brain 是 also helpful. Jackendoff, in various places, also gets very explcit 关于 the nitty-gritty. One really has to spell out in detail the nature of a representation, 和 in a way 那 那 (conceptual 要么 lexical) representations 能够 do the work they demonstrably do, in terms of storage, inference, sensorimotor interfaciness, etc. How this 是 done in neural tissue, i.e. 在 the implementational level of analysis, 是 completely unclear.

没有争论说感觉运动特征与概念和词语有关,从这个意义上说,'weak 体现.'
但是,作为概念和词语的理论,感觉运动特征列表是否足以满足需要是非常值得怀疑的。

格雷格希科克说过...

大卫- I too 思想 体现 was just a new way of thinking 关于 how concepts are represented. That there may be more involvement of lower-level sensory 和 motor systems than previously 思想. That's an interesting idea worth investigation. Like you, I think it has serious limitations. Then I started reading claims 那 体现 was a "post-cognitive"革命,新的思维模式以及计算方法的替代方法。那's what disturbed me because you 能够'无需某种形式的计算就可以通过视网膜(或单个神经元)。

格雷格希科克说过...

安德鲁-唐'跟我一样实际上,您应该先了解一种理论,然后再对它进行批判。 :-)"感知界面理论"正如霍夫曼所说,我们不'看不见世界。那不't mean 那 we don'看不出它对生存有用。实际上,它认为我们不'不能准确地看到它,因为从残障中看到它对于生存更为有用。他以计算机桌面GUI为例。您桌面上的文件是't *确实*是矩形,蓝色,并且位于屏幕的左上方。它 '只是一个方便的界面,使计算机交互对于每天的快速使用都是可行的。如果我们必须直接处理机器电路,电压,二极管等计算机,那么大多数人和大多数任务将无法使用计算机。霍夫曼(Hoffman)同上(并与进化模拟一起展示),以进行感知。

"Well,"你反对,很多人都对霍夫曼's arguments, "如果那辆车从街上走下来不是'真的在那里,为什么不'你走在前面吗?"只是因为感知不是't veridical doesn't mean it 是n't useful. 你不'出于同样的原因,不要在汽车前出门't将文件图标拖到垃圾桶图标中。您're 不 *really* putting the file inside a trash 能够, but the consequences of the action are real enough: the data will disappear.

沃利·斯通说过...

大家好,
我认为在这场讨论中有一个非常重要的区别,夏皮罗在他的教科书中指出过。大卫'关于感觉运动过程在概念过程中的作用的观点显然与'conceptualization'各种实施方式;安德鲁'的想法似乎与'replacement'品种。这两个假设都暗示着身体和环境优先于解释行为(因此都享有描述性术语)'embodiment'), do so for different reasons. Importantly, 计算, perceptual inference, 和 representation are NOT 在 odds with the 概念化 idea, while they ARE 在 odds with the 替代 idea. This very well may make 概念化 和 替代 hypothesis incompatible. Importantly, while 概念化 ideas 能够 be explicated firmly in neuro-computational language, the 替代 ideas 能够not. This makes 替代 体现 more revolutionary (or radical), while the 概念化 theory 是 不 so revolutionary; it 是 just a way of grounding abstract 'thought' (or more objectively, behaviors 那 we think reflect abstract 思想 such as language) in the sensorimotor experiences of animals (i.e the interaction between brain, behavior, 和 environment).

Importantly, whether we take a 概念化 要么 替代 approach (or neither) has major consequences for how we describe, 和 what predictions we make, of the owl'良好的本地化能力。我认为在这一点上,区别变得很重要。

格雷格希科克说过...

Thanks, Heath. I stick to my 要么 iginal point, though: if we all agree 那 延迟线s in the owl (or any neural 建筑 要么 any neuron) 是 处理信息, then the 替代 variant of 体现 是 just traditional cognitive psychology in new clothes.

沃利·斯通说过...

I think there are certainly strong arguments for 那 perspective 和 I could certainly agree with you; in general, I find explanations with 计算s 和 representations quite useful, too.

I do wonder if the traditional cognitive psychology 和 radical 体现 are so different in their epistemology 那, though they 在tempt to explain the same thing, they are NOT mutually exclusive. Light behaves as particles 要么 waves, depending on how it 是 measured. Do we have a case where a behavioral phenomena (e.g. sound localization) behaves as a 计算al 处理 要么 as a direct 知觉 处理, depending on how you measure it (i.e .form the traditional perspective 要么 the radical one)? 如果 this 是 so, then traditional cognitive psychology could be internally coherent 和 predictive, 和 so could radical 体现, but one does 不 provide the BEST answer (i.e. light 是 不 particles OR waves). I think this line of 思想 takes us afield of the phenomenon in question (i.e. sound localization), 和 it does get pretty thick with a philosophy 那 I am 不 an expert in, but I do wonder if there 是 value to this 不ion for practicing experimental psychologists. (Am I saying 'can'我们都是朋友吗?也许有一点,但我正在考虑科学的实际后果!)

格雷格希科克说过...

Interesting 思想s. First, yes, let'都是朋友!我喜欢辩论,想法很有趣,无论是对是谁,还是在我们之间,'会学到一些东西。那'是我们玩的游戏,而且对我而言从来都不是个人的。有趣的故事:我参观了ASU和Art Glenberg's "认知认知实验室。 "在与他和他的实验室聊天时,我正在告诉他们我认为他们错了的所有原因以及Art在反驳等等。学生们变得慌张,生气,沮丧。 Art和我度过了愉快的时光。在我的部门谈话之后,Art邀请我去他家,甚至提出要我过夜。他'是个好人,即使我认为他's dead wrong, I'd和他一起开心地闲逛。

无论如何,我发现这很有趣,并告诉那些实际上试图理解猫头鹰如何定位声音或人类检测视觉运动的人't care whether the system 是 体现的 要么 symbolic 要么 whatever you want to call it. They simply try to figure out how it works. The debate I'm having with 安德鲁 seems to be moving toward more 关于 what to call 延迟线s 不 whether there are 延迟线s. 如果 那'在我们最终到达的地方,哲学上的区别是虚无。

安德鲁说过...

Couple of things: Heath 是 exactly right 关于 the conceptualisation vs 替代 hypotheses 和 the fact they are distinct. The former 是 关于 接地表示 后者不是。我之所以进入替换营,是因为 我和萨布丽娜在我们的论文中争论, we think allowing any 体现 removes any need for representation by definition.

希思,我不'认为代表性和非代表性帐户就像是粒子/波动事物-我认为它们完全不同且不兼容(并且前者是错误的:)

无论如何,我发现这很有趣,并告诉那些实际上试图理解猫头鹰如何定位声音或人类检测视觉运动的人't care whether the system 是 体现的 要么 symbolic 要么 whatever you want to call it.
我认为他们应该关心,因为即使他们不这样做't explicitly talk 关于 it, their work 仍然受到理论框架的指导。如果那个框架是错误的,那么它们将陷入死胡同;如果他们不这样做'不承认理论在指导他们的经验工作中的作用,那么他们就赢了't 知道 how to get out of the dead end. 没有有效的理论自由科学。

我不'当前同意猫头鹰系统正在处理信息,至少不会以'处理信息'什么的解释's going on.

为了记录我也没有't take these arguments personally; if my work 能够'不能接受批评,然后我'我做错了。 (和 I've最近还与Glenberg进行了文学交流;我认为他'也是完全错误的,但我们进行了很好的辩论,他在整个过程中都非常好。一世've还与同事进行了精彩的排练,吓坏了学生! :)

安德鲁说过...

Re Hoffman:这种方法的问题在于,您需要根据一些规则集来约束您对世界的假印象,这些规则集实际上可以提高您的生存能力。该规则集从何而来?在您的环境的短期和长期变化中,它的稳定性如何?

The GUI example actually reveals all these problems. Software interface design 是 actually an amazingly hard problem because the environment you are controlling has few if any regularities, let alone any laws. 那里 are lots of ways to make the GUI, some work for a while 要么 for a limited use, 和 they are all very fragile 和 能够 get screwed over easily.

What you need to 知道 to enhance survivabilty 是 what'现在世界上正在发生。您需要的信息非常接近'about' 那 world as you 能够 get. One large class of information we use 是 那 which 能够 specify action-relevant dynamical properties of the world. It exists 和 we use it because it 是 very stable 和 extremely 功能性 - it 是 entirely 关于 what you need to 知道 because of the lawful 处理 那 created it (event dynamics interacting with energy arrays such as light).

任何说明我们系统地使用输入以使其更有用的理论都需要一个关于我们如何'know'如何使输入更有用,以及在哪里'knowledge' came from (scare quotes just indicating you 能够 read 'knowledge' any way you like, 不 just some explicit rule set). Our Frontiers paper actually argues 那 as soon as you have a workable way to answer those questions, you actually end up with such good access to the environment 那 this kind of representational, inferential futzing 是 no longer required. This 是 why we think radical 体现 是 actually kind of inevitable.

我不同意我'm随随便便驳回了这一点;一世'我以理论为由驳回了这一点 心理学家通常不习惯。我认为理论是一种优势,而不是劣势,但是对于想要数据的心理学家来说,这听起来总是错误的。数据很好,但是有时候's 不 the point.

格雷格希科克说过...

关于霍夫曼

安德鲁·威尔逊(Andrew Wilson)说,"该规则集从何而来?"

查尔斯·达尔文说:"natural selection."

唐·霍夫曼(Don Hoffman)同意达尔文的观点。

格雷格希科克说过...

Regarding 信息处理--

安德鲁说"I don'当前同意猫头鹰系统正在处理信息,至少不会以'处理信息'什么的解释's going on."

好。你不't like "computation." 你不't like "信息处理."告诉我您喜欢什么术语并提供定义。

您说计算神经科学家需要一个理论:"没有有效的理论自由科学。" I agree completely. But they have a theory. Delay lines function as coincidence detectors such 那 the location of a stimulus 能够 be represented by the activity of the network. You 能够 build a functioning model 那 does this. 实际上,智能手机中是劳埃德·瓦特(Lloyd Watt)开发的芯片,该芯片使用这种神经计算机制来改善设备的声音处理能力。 It 是 a strong theory with proven real world applicability. You just don'不喜欢他们用来描述正在发生的事情的术语。

威廉·马汀说过...

好。你不't like "computation." You don't like "信息处理."告诉我您喜欢什么术语并提供定义。

如果 I had to channel the 体现的 认识 perspective, I think they would say 那 their term 是 physics. I think they assume everything in psychology reduces to physics, so they think 那 labels like "信息处理" 要么 "computation" 要么 "cognition" 要么 even "perception"只是组织现象的便捷方法,不要't really exist.

安德鲁说过...

实际上,智能手机中是劳埃德·瓦特(Lloyd Watt)开发的芯片,该芯片使用这种神经计算机制来改善设备的声音处理能力。
I should hope the 计算al solution works in my phone - it'电脑!这种成功与猫头鹰大脑是否在计算无关。

我其实不't 知道 what term to use any more. I'我不反对在一个层面上进行信息处理,尽管我特别考虑猫头鹰系统,'看不到正在执行的计算'尚不清楚正在进行处理。我看到了一种具有动态功能的设备,该设备可以与信息交互,从而为猫头鹰产生功能性行为。这是一种处理,但是'这不是该术语通常的应用方式。没有人说锤子在执行时正在处理信息'落在引力场', for example.

William 是 right to an extent; the (replacement style) 体现的 term 是 much more grounded in physics. It'不能归结为物理学(这是一个我不懂历史的词组)'t think applies here) but it 是 关于 treating nervous systems as physical devices rather than 计算al ones.

神经系统确实将信息与行为联系在一起。一世'随着线程的不断发展,我变得更加确信'信息处理'这不是正确的构架方法。一世'我仍然在想。

格雷格希科克说过...

The point 关于 the smart phone chip 是 那 the 信息处理 principles, abstracted away from its implementation, has explanatory value 和 real world applicability. This shows 那 the 信息处理 level of description 是 doing explanatory work.

The point 那 underlies traditional cognitive science/信息处理 approaches 是 那 with the right arrangements of matter (e.g., neurons) a system 能够 be induced to *transform* energy (i.e., information such as ITDs) in such a way to accomplish a task (an 要么 ienting head turn). Honestly, I think it 是 fine if you prefer to describe what 是 going on 在 a physical level: cochlear waves mechanically stimulate hair cells which induces currents 那 integrate which causes action potentials. Probably you'我们需要将其简化为分子过程,以更接近您所设想的因果世界。那'它被称为分子生物学和生理学,它非常有价值,但在建模认知过程时却是一个非常笨拙的层次。

What cognitive psychologists have 不iced 是 那 you 能够 capture a lot of generalizations if you work 在 the level of 信息处理 (aka, 计算) 和 make a lot of progress understanding visual object recognition, hearing, language, motor control, etc. 如果 it really 是 more a level of description thing then 替代 体现 是n'一个替代的计算认知模型'寻找底层的实现。在这种情况下,您需要切换领域并研究生理学,解剖学,分子神经生物学或物理学。 ;-)

Incidentally, those of us working on cognitive neuroscience take our job to be the effort to uncover 生物学与计算模型之间的联系. We value both 和 want to understand their relation. To my mind this 是 much more productive than being exclusively 计算al 要么 exclusively physiological. You 能够 learn from both levels of description.

多米尼克·卢克说过...

@格雷格,我想在那里'这是将建模成功与解释力相混淆的问题。一个不相关字段的示例是'理性选择理论' in economics. It'实际上,它非常擅长对人口行为建模(在一定程度上)。但是,实际上,它甚至无法近似真实的个人做出决策的方式。

在许多情况下,建模成功不一定意味着您've提供了解释。它'记住'所有模型都是错误的,但其中许多有用' dictum. So, I'd argue 那 there'发现之间的区别"生物学与计算模型之间的联系"对生物现象的观察进行计算建模。我认为后者非常有用,但是前者充满了将模型与现实融合的巨大潜力。正如我尝试的那样,语言学充满了这类问题 在这篇文章中显示.

@Andrew Wilson I tend to agree with you 那 信息处理 是 too tied in the symbolic 计算 model. But if you go back to the Shannonian basics, it 能够 be a very powerful way of thinking 关于 signal transmission. It allows for some mathematical modeling of simple interactions (maybe even hammer hitting a nail). However, most people think it means 处理ing a series of if-then transformations between strings of bi-polar symbols. And it's very unlikely 那 this 是 a very good picture of what the brain 是 doing (as far as I 能够 tell from afar) 和 definitely 不 what goes on in language (as I 能够 tell from very up close).