Content uploaded by Zihao Zhang
Author content
All content in this area was uploaded by Zihao Zhang on Jan 05, 2021
Content may be subject to copyright.
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
010 011
VOLUME 8 / ISSUE 4 / AUGUST 2020
https://doi.org/10.15302/J-LAF-1-020032 收稿时间 RECEIVED DATE / 2020-05-30 中图分类号 / TU413.6, Q811.3 文献标识码 / A
编辑 田晓劼 翻译 张书瑞 张子豪 刘浔 田晓劼
EDITED BY TIAN Xiaojie TRANSLATED BY ZHANG Shurui ZHANG Zihao LIU Xun TIAN Xiaojie
本文引用格式 / PLEASE CITE THIS ARTICLE AS
Zhang, Z., & Liu, X. (2020). Control and Uncertainty: Towards a Paradigm of Prototyping.
Landscape Architecture Frontiers, 8(4), 10-25. https://doi.org/10.15302/J-LAF-1-020032
摘要
本文探索了一种不同于模型设计的原型思维
新范式。在将控制论与景观设计进行并置分析后
发现，自20世纪至21世纪初，景观设计学科的发
展规律反映了控制论在同一时期的发展进程。早
期关于系统动力学的决定论和线性理解，逐渐被
基于演生论和开放式的主张所取代。在演生论的
框架下，笔者强调一类可被理解为原型的特殊模
型不能被基于“建模－预测－控制”的传统框
架所包容。原型并不像传统建模过程那样，用
模型代表一个开放的生命系统，原型本身就可
以被理解为一个独立的生命体系。原型思维可以
取代模型设计，并展现出三方面的特质：1）原型
本身即具有“生命力”，这是设计与创造力的根
本；2）原型的真正意义在于其未被定义的状态，
而不在于其可被定义的最直接的用途；3）这种未
被定义的状态能够提供无限可能，并将人们与未
来之间的关系从过去的偶然性和预测转变为期
待与展望。
关键词
原型；景观；控制论；系统思维；不确定性；控制
ABSTRACT
This paper maps out a new paradigm of prototyping
that acts as an alternative to the model-making
paradigm. By juxtaposing the cybernetics
movement with landscape design, the authors have
mapped out a development in landscape discourse
that mirrors the movement of cybernetics in the
20th century and early 21st century. The early
deterministic and linear understanding of systems
dynamics is replaced by an emergent and open-
ended view. Taking on a framework of emergence,
the authors highlight a special type of model
that does not fit within the conventional model-
predict-control framework. Rather than models
that represent another living system, these models
are living systems in themselves with autonomy
and lives. This special type of model can be
understood as prototypes. Prototyping replaces
model-making and exhibits three distinctive
qualities: 1) A prototype has a life of its own, which
serves as the basis for design and creativity;
2) The real usefulness of a prototype lies in its
undefined identity rather than its defined and direct
application; And 3) the identified quality provides
a wide range of possibilities, thus changing our
relationship with the future from chance and
prediction to anticipation and hope.
KEYWORDS
Prototype; Landscape; Cybernetics; Systems
Thinking; Uncertainty; Control
1 Introduction: Model, Control, and Uncertainty
“There is always a tension in landscape between the reality
and autonomy of the nonhuman and its cultural construction,
between the human impulse to wonder at the wild and the
compulsion to use, manage, and control.”[1] Designing with
living systems and materials, landscape designers always have to
deal with unexpected outcomes. What insights can the idea of
prototype provide to address the issue of uncertainty?
Oxford English Dictionary
defines the term of “prototype”
as “[t]he first or primary type of a person or thing; an original
on which something is modeled or from which it is derived; an
exemplar, an archetype”[2]. The term was first popularized in
industrial design and software engineering to denote the first
machine or the first piece of program that was built to prove
the concept. A prototype possesses a twofold quality: On the
one hand, a prototype is a type of model, which has a kind of
metaphorical relationship between one thing and another; On
the other hand, a prototype also carries a sense of realism that
is beyond a metaphor. This twofold quality makes prototype a
special type of model that can transform people’s understanding
of the tension between control and uncertainty.
However, rooted in the western means-end chain relationship,
prototypes have always been overlooked theoretically and
historically. In the western philosophical tradition, one envisages
an ideal form (eidos) as a model, and then the model can serve as
a goal (telos) and, at the same time, an end that calls for actions.
As French Philosopher Francois Jullien puts forward that “with
our eyes fixed on the model that we have conceived, which we
project on the world and on which we base a plan to be executed,
we choose to intervene in the world and give a form to reality.”[3]
Based on this line of reasoning, theory can be differentiated from
practice, the former being the basis for the model, and the latter
being a set of operations that make the model turn into reality.
Most importantly, with the means-end relationship, the idea of
effectiveness and measure can be tied into this habitual reasoning.
Once a range of possible tools and actions are at hand, their
effectiveness will be evaluated. Efficacy becomes the concept that
ties the “means” and the “end” together. Finally, although people
want to project an ideal model on the world and develop means
to achieve this end, unexpected circumstances will always rise
to undermine the action and control regime of any plan. Thus
uncertainty denotes those events that are outside the predictions
allowed by the conceived model[3].
This means-end chain can be observed in every aspect
of contemporary environmental practices. For example, the
past few years have seen a new round of model-making in the
牛津英语词典将“原型”（ｐｒｏｔｏｔｙｐｅ）定义为“人或事物的第一
个或主要类型；基于此本源可以模拟或衍生出其他事物；是一种典范
（ｅｘｅｍｐｌａｒ）或初型（ａｒｃｈｅｔｙｐｅ）”［２］。原型一词首先被普遍使用于工
业设计和软件工程中，用来指代以概念测试为目的的第一台机器或第
一个程序。原型概念具有双重属性：一方面，它是一种表明某一事物
与另一事物之间存在着隐喻关系的模型；另一方面，原型还具有一种
超越隐喻的真实感。原型的双重属性使其成为一种特殊的模型，通过
它可以反思控制与不确定性之间的矛盾。
然而，由于西方文化中根深蒂固的“手段－目的链”思维方式
（即以手段达到目的），原型理论在历史上被长期忽视。在西方哲学
传统中，人们往往用一种理想形态（ｅｉｄｏｓ）作为模型，并将其设为目
标（ｔｅｌｏｓ）采取行动。正如法国哲学家弗朗索瓦・朱利安所说，“我
们（西方人）将目光集中在构想出的模型上，并将其投射在现实世界
中，以此来指导我们的行动；从这个意义上讲，我们选择去干预周遭
世界并赋予现实以形式。”［３］这种思考模式将理论与实践区分开来：理
论成为模型的基础，而实践则成为将模型变为现实的一系列行动。最
重要的是，行动是否有效以及可否被测量，也受到这种思维习惯的深
刻影响：人们习惯于评价工具和措施的功效性。由此可见，功效性决
定了手段－目的链的思维逻辑。最后，尽管人们想把一个理想的模型
投射在周遭世界中，并试图研究出各种方法来实现这一模型，但过程
中总会出现无法预测的情况，影响计划的控制和实施。因此，所谓不
确定性是指那些在构想出的模型中无法预测的状况［３］。
在当代的环境实践中，手段－目的链体现在方方面面。例如，
在过去几年间，环境实践中出现了新一轮对建模的讨论。包括智慧城
市在内的许多研究都提出要把传感网络融入环境中，从而产出更多数
张子豪
弗吉尼亚大学建筑学院构建环境博士候选人、景观系讲师
刘浔
弗吉尼亚大学建筑学院构建环境博士研究生、景观系教师
ZHANG Zihao*
PhD Candidate in Constructed Environment and Lecturer in
Landscape Architecture, School of Architecture, University of
Virginia
LIU Xun
PhD Student in Constructed Environment and Teaching Fellow
in Landscape Architecture, School of Architecture, University of
Virginia
控制与不确定性：
对原型思维范式的展望
CONTROL AND UNCERTAINTY:
TOWARDS A PARADIGM OF
PROTOTYPING
*Corresponding Author
Address: Campbell Hall, School of Architecture, 110 Bayly Dr,
Charlottesville, VA 22903, USA
Email: zz3ub@virginia.edu
1 引言：模型、控制与不确定性
“景观的非人类属性及其文化构建之间，以及人类对自然的敬畏
和对其进行利用、管理与控制的冲动之间，总是存在着矛盾。”［１］景
观设计师在对生命系统和物质进行设计时，经常会遇到意料之外的情
况。那么，原型概念将如何帮助设计师应对设计中的不确定性？
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
012 013
VOLUME 8 / ISSUE 4 / AUGUST 2020
据，机器学习（ＭＬ）算法便能利用这些数据来构建人工智能（ＡＩ）系
统，随后将ＡＩ系统（即模型）应用于信息物理系统之中，以管理各种
环境过程（如雨洪控制［４］）。但是，这一建模框架受困于手段－目的链
思维方式，因为在手段－目的链中，不确定性只被理解为模型与环境
间必须要被消除的差异，而这种控制和不确定性间的矛盾只会一直横
亘在人们与未来的关系之间。
在过去几十年间，景观理论与实践不断挑战着手段－目的链以及
模型在环境设计实践中的作用［５］。这些景观实践强调了原型作为一种特
殊模型的价值，而这种价值并不能用手段－目的链思维方式来解释。
本文以景观实践为例，阐明原型在环境设计中的价值，即原型提供了
一种新的方式来思考建模过程：建模不是为了消除不确定性，而在于
探索更广泛的可能性。如此一来，人们与未来的关系也不再受困于控
制与不确定性之间的矛盾，而是转变为一种期待与展望，这对于人类
应对气候变化等诸多环境挑战而言至关重要。
鉴于模型和控制与不确定性之间的矛盾是控制论和系统论的核心
议题，因此，在控制论和系统思维的发展过程中，重新研究如何诠释
原型也就变得尤为重要。如将一系列景观作品和控制论的三次思潮进
行对照，会发现早期控制论中的手段－目的链并不适用于当代系统性
的景观设计；相反，景观设计可以为原型思维提供不同见解。
对原型的新的认知能够帮助解析控制论运动中的原型思维概念，
并发掘了包括诺伯特・维纳的“蛾”和“臭虫”，以及约翰・冯・诺
伊曼的细胞自动机在内的一系列“反馈生物”①。受手段－目的链思维
方式的影响，这些可以作为控制论运动原型的反馈生物一直为人所忽
视。基于此，这些控制论原型可与一个景观学科的案例
—
哈佛大学
设计研究生院（ＧＳＤ）的响应式环境及人工产品实验室（ＲＥＡＬ）进行
environmental discourse. Myriad of research initiatives, such
as smart cities, propose to embed sensing networks in the
environment to produce more data. Then, the environmental
data feeds into Machine Learning (ML) algorithms to build
Artificial Intelligence (AI) systems. Finally, these AI systems, or
models, are distributed in the cyberphysical systems to manage
all kinds of environmental processes, such as stormwater
control[4]. However, trapped in a means-end relationship,
this framework only conceptualizes uncertainty as model-
environment difference that needs to be reduced one way or
another. In the means-end chain, human’s relationship with the
future always revolves around the tension between control and
uncertainty.
Landscape theory and practice in the past decades have
provided a body of work that challenges the means-end
chain and the role of models in environmental practices[5].
These practices highlight the value of prototypes as a special
type of model that does not fit within the means-end chain.
This paper uses landscape practices to articulate the values
of prototypes in the environmental discourse, and argues
that with prototypes, model-making can be envisaged in such
a way that is no longer on the line of reducing uncertainty
but about providing a wide range of possibilities. Thus, our
relationship with the future is transformed from the tension
between control and uncertainty to anticipation and hope,
which is crucial when humanities are faced with environmental
challenges such as climate change.
Since the idea of model and the tension between control
and uncertainty have always been central to the field of
cybernetics and systems theory, it is crucial to re-examine
how the idea of prototype was articulated in the development
of cybernetics and systems thinking. Contrasting a body of
landscape works and the three waves of cybernetics movement
shows that the contemporary system-based landscape design
framework does not fit within the means-end that is entailed
by early cybernetic principles; Instead, landscape design can
provide an alternative understanding of model-making on the
line of prototyping.
This updated understanding can help parse out the concept
of prototyping in the cybernetics movement and unearth a
family of “cybernetic creatures,”① including Norbert Wiener’s
“the moth” and “the bedbug,” and John von Neumann’s
cellular automaton, as prototypes in the cybernetics movement
which has been overlooked within the means-end chain. These
cybernetic prototypes can be juxtaposed with a case study
in landscape design—a hydromorphology table experiment
conducted at the Responsive Environments & Artifacts Lab
的水文地貌沙床实验
—
进行并置分析。这一案例可以帮助总结出原
型的三个特征，以重新理解环境实践中的不确定性。
2 背景：控制论与景观设计
控制论是自２０世纪４０年代以来在美国发展出的一个交叉学科研究
领域。第二次世界大战后，包括工程师、数学家、人类学家和生态学
家等在内的知识分子基于系统思维的新理论框架对机械系统和生物系
统间的相互控制与交流进行了研究。控制论的核心焦点是如何通过反
馈机制控制不同的系统，以减少不确定性。直到近几年，学者们才开
始关注控制论原理对景观设计的影响［６］［７］。有学者认为，某些控制论理
念是在２０世纪６０年代通过生态科学和艺术等学科渗透到景观设计学科
中来的。伊恩・麦克哈格和劳伦斯・哈普林是在控制论与景观设计之
间建立联系的两位关键性人物［７］。但由于控制论的交叉学科属性，系
统论思想早已盛行于各学科，因此控制论中的概念也早在２０世纪便已
经对学界产生了广泛影响。控制论最初提出的诸多概念（如反馈、耦
合、自生系统论）已被纳入社会学、计算机科学、系统理论、生态学
和人文学科，而这些学科都是景观设计师汲取灵感的土壤。因此，可
以说自２０世纪６０年代以来，控制论以及控制与不确定性等议题已广为
人知。本文通过介绍２０世纪中叶以来控制论与景观设计理论和实践各
自的发展历程，将重点论述原型思维对景观设计理论和实践的影响，
从而帮助景观设计师重新思考不确定性这一议题。
凯瑟琳・海尔斯将控制论的发展概括为三次发生于不同学科领域
的思潮［８］。第一次思潮（通常也被称为一阶控制论）是指１９４６～１９５３
年间召开的关于控制论的“梅西会议”以及此后２０年间的相关研究。
维纳和冯・诺伊曼等早期控制论学者致力于研究稳态系统的反馈机制
和控制策略。第一次思潮后期，包括玛格丽特・米德、格里高里・贝
特森和海因茨・冯・福斯特等在内的人类学家和哲学家，开始对工程
师在建模过程中扮演的观察者身份提出质疑，并提出二阶控制论的新
(REAL) of Harvard Graduate School of Design (GSD). With
the cases, the case study can help map out three characteristics
of prototypes that provide an alternative understanding of
uncertainty in environmental practices.
2 Background: Cybernetics and Landscape Design
Cybernetics is an interdisciplinary field of study that
started in the United States since the 1940s, when a team of
post-war intellectuals, including engineers, mathematicians,
anthropologists, and ecologists, converged on a new theoretical
model based on systems thinking to understand control and
communication between mechanical and biological systems.
The core concern in the field is about how to control different
systems through feedback mechanisms to mitigate uncertainty. It
is only in recent years that scholars have started to pay attention
to the influence of cybernetics principles on landscape design[6][7].
It is argued that cybernetics concepts instilled into the landscape
discipline via ecological sciences and arts in the 1960s, and Ian
McHarg and Lawrence Halprin were two prominent figures who
established the link between cybernetics and landscape design[7].
However, because of the interdisciplinary nature of cybernetics,
its concepts were instilled in every aspect of intellectual life in the
20th century as systems theory has prevailed across disciplines.
Many ideas such as feedback, coupling, and autopoiesis that
were initially developed in the field of cybernetics have been
imported into disciplines such as Sociology, Computer Science,
Systems Theory, Ecology, and Humanities, all of which are fields
where landscape architects have been drawing inspirations.
Thus, it can be simply recognized that, since the 1960s,
cybernetics and the issue of control and uncertainty have been
“in the air.” This paper contrasts the cybernetics movement with
landscape design theory and practice since the mid-20th century
and highlights the idea of prototyping in both fields that help
landscape architects rethink the issue of uncertainty.
Katherine Hayles schematized the development of
cybernetics as three waves of research mobilizing among
different fields of study[8]. The first wave of research, commonly
known as first-order cybernetics, speaks to the Macy
Conferences on cybernetics from 1946 to 1953 and the research
in the following two decades. Early cyberneticians such as
Wiener and von Neumann focused on the feedback mechanism
and control strategies in homeostatic systems. Towards the end
of the first wave of research, anthropologists and philosophers
including Margret Mead, Gregory Bateson, and Heinz von
Foerster problematized the role of engineers as observers in the
model-making process, trying to propose a new framework,
① Cybernetics研究的是
动物和机器之间的控制
与反馈机制，而控制只
是其应用领域，单纯将
cybernetics译为“控
制论”并不准确且易造
成歧义，因此本文将
cybernetic creatures和
cybernetic machine中
的cybernetic译为“反
馈”。
① Cybernetics studies the
control and feedback
mechanism between
animals and machines,
and control is only its
application field. It is
not accurate and easy
to cause ambiguity
to simply translate
cybernetics as "控制
论". Therefore, this paper
translates cybernetics
into "反馈" when it
refers to cybernetic
creatures and cybernetic
machines.
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
014 015
VOLUME 8 / ISSUE 4 / AUGUST 2020
框架
—
这一框架将观察者作为反馈系统的一部分（图１）。这种观念
上的转变引发了２０世纪６０年代后期至７０年代的第二次研究思潮。这期
间，学者们致力于探寻观察者在建模过程中的作用，其中智利生物学
家温贝托・曼图拉那和弗朗西斯科・瓦雷拉提出了“自生系统论”，
以阐释系统如何利用物质和信息流维持自身属性［９］。之后，瓦雷拉开始
涉足人工生命领域，并引发了于２０世纪９０年代初期兴起的第三次控制
论思潮，其中的焦点议题是演生论与复杂性。通过人工生命实验这种
多智能体模拟，学者们认为诸如“智能”和“意识”之类的复杂现象
实际上可以被理解为分布式系统之间相互作用的结果。自２０世纪９０年
代后期以来，第三次思潮开始不断挑战人类例外论，并引发了一系列
后人类主义的思考［８］。
控制论的三次思潮的研究焦点可以与景观研究和实践进行比照
（图２）。麦克哈格的生态学设计方法对于景观的控制可以视为一阶控
制论的一种体现。工程师们通常通过控制熵来维持一个机械系统的稳
态；在热力学中，熵是指系统向着无序状态发展的趋势；麦克哈格在
其著作中大量使用熵的概念，并将其与生态适应性进行类比。在他看
来，熵即不确定性，会对生态系统的完整性造成威胁。因此，他认为
景观设计师应遵循理想的生态演替模型来控制景观变化，最终达到生
态系统“演替顶级”的平衡状态［７］。然而，当代景观理论和实践注重开
放性与演生，而麦克哈格设计框架中的环境决定论和对于生态过程的
线性理解已遭到质疑。［１０］
控制论的第二次思潮可以与大都会建筑事务所（ＯＭＡ）的两个景
观设计项目进行比照。其中一个是ＯＭＡ于１９８２年参加的法国拉・维莱
特公园设计竞赛作品，另一个是于２０世纪９０年代末参加的加拿大当斯
known as second-order cybernetics, to include observers as
part of the cybernetic system (Fig. 1). This conceptual shift
gave rise to the second wave of research from the late 1960s
to the 1970s, in which scholars focused on observing systems
and their role in the model-making process. Chilean biologists
Humberto Maturana and Francisco Varela introduced the
term “autopoiesis” (self-production) to conceptualize how
systems use the material and information flow to maintain their
identities[9].Varela then ventured into the field of artificial life,
thus connecting the third wave of cybernetics research in the
early 1990s. The key concepts are emergence and complexity.
Through the experiment of artificial life—a multi-agent
simulation—scholars posit that complex phenomena such as
intelligence and consciousness are, in fact, results of interactions
between distributed systems. The third wave of research thus
has challenged human exceptionalism and given rise to a range
of posthumanist considerations since the late 1990s[8].
A body of landscape research and projects can be contrasted
with the three waves of cybernetics in terms of their major
concerns (Fig. 2). McHarg’s ecologically inspired design
methodology can be understood as a version of landscape
control through first-order cybernetics. To control a mechanical
system, engineers usually manage the entropy to maintain
the system’s homeostasis. In thermodynamics, entropy refers
to the tendency of a system to move into disorder and chaos.
McHarg used the concept of entropy extensively throughout his
writing and always compared entropy with ecological fitness.
For McHarg, entropy is another term for uncertainty, which
poses a threat to the integrity of the ecosystem. Thus, landscape
designers’ work is to control landscape change by following a
set of ideal models of ecological successions and finally reach
an equilibrium state as expressed in the “climax stage” of an
ecosystem[7]. From a perspective of contemporary landscape
theories and practices that focus on open-endedness and
emergence, McHarg’s design framework has been critiqued for
its inherent environmental determinism and linear understanding
of ecological processes.[10]
The second wave of cybernetics can be contrasted with
two landscape projects by OMA. The first one is the entry for
the Parc de la Villette competition in 1982, and the second is
the entry for Toronto’s Downsview Park competition in the
late 1990s called “Tree City”[11][12]. The two projects can be
understood as a seriation in which the designers deployed the
tactic of “self-production” of the park system as design strategies,
and the urban parks were understood as an autopoietic (self-
producing) system that uses flows of material and information
to reproduce their own organizations. An autopoietic system
维尔公园设计竞赛作品“树城”［１１］［１２］。这两个方案可被看作一个系列，
它们都利用“自生系统”作为设计策略，将公园理解为一个自组织
系统，即利用物质和信息不断复制自己的体系。所谓自生系统，是指
系统通过其内部组织关系来处理信息物质流，从而维持自身特性。例
如，在当斯维尔公园竞赛中，ＯＭＡ用１ ０００条道路构成的路网和植物簇
群作为空间框架，分割出大片未进行功能限定的区域，以满足公园未
来发展需求。乍看之下，ＯＭＡ的“树城”表现出一定的不确定性，因
此是开放的。但是，这种灵活性是基于城市公园系统的自我生产，从
而维持了另一个层面上的稳态，这种对稳态的控制可以由二阶控制论
解释［１３］。景观设计师、理论家克里斯蒂娜・希尔对ＯＭＡ的设计持强烈
的怀疑态度，她指出，尽管ＯＭＡ的文案中描述了一个高度灵活的设计
框架，但这个设计本质上还是一个传统的田园牧歌式的城市公园［１４］。换
言之，场地通过系统运转来维持其作为“田园式公园”的系统组织，
几乎没有为景观演生和未知的发展可能预留任何空间。
由詹姆斯・科纳事务所、斯坦・艾伦与生态学家尼娜－玛丽・利
斯特合作的参赛作品“生态演生”［１５］可以和控制论的第三次思潮进行比
照。第三次思潮注重开放性和演生，而该设计方案的框架将使场地通
maintains its identity by processing the flow of information and
material using its components and relationships. For example, in
the Downsview Park competition, OMA proposed a 1,000-path
network and vegetal clusters separated by vast undesignated areas
as a spatial framework to account for the future development of
the park. It appears that OMA’s Tree City exhibits some level of
uncertainty, thus being open-ended. However, this flexibility is on
the basis of self-production of the urban park as an autopoietic
system, thus maintaining another level of homeostasis by
legitimizing a control regime based on second-order cybernetics[13].
Landscape architect and theorist Kristina Hill was highly critical
towards OMA’s proposal, and she pointed out that despite the
language which renders a flexible framework, OMA essentially
proposed a traditional urban park with pastoral forms[14]. In
other words, the site operates to maintain its system organization
as a “pastoral park,” leaving little room for the consideration of
emergence and unfolding of how the landscape could become.
Another entry for the Downsview Park Emergent Ecologies[15]
from the collaboration between James Corner Field Operations,
Stan Allen, and ecologist Nina-Marie Lister can be contrasted with
1-1
© 张子豪
1-2
© 张子豪
1-1. 一阶控制论
1-2. 二阶控制论
1-1. First-order
Cybernetics
1-2. Second-order
Cybernetics
2. 控制论与景观设计
2. Cybernetics and
Landscape Design
系统
System
观察者
Observer
反馈
Feedback
环境
Environment
系统
System
反馈
Feedback
观察者
Observer
环境
Environment
2
© 张子豪
景观设计
Landscape Architecture
控制论与系统思维
Cybernetics and systems
thinking
生态系统平衡
Ecosystem equilibrium
自生系统论／自组织
Self-production / -organization
生态演生
Emergent ecologies
伊恩·麦克哈格开始教学与
实践，将生态学引入景观
设计领域。
Ian McHarg began
teaching and practice,
introducing Ecology into
the field of landscape
design.
诺伯特·维纳
Norbert Weiner
《控制论：或关于在动物和
机器中控制和通信的科学》
Cybernetics: Or Control
and Communication in the
Animal and the Machine
温贝托·曼图拉那与弗朗西斯科·瓦雷拉
Humberto Maturana and Francisco Varela
《自创生与认知：生命的实现》
Autopoiesis and Cognition: The Realization of
the Living
人工生命模拟
Artificial life simulation
凯瑟琳·海尔斯
Katherine Hayles
《我们何以成为后人类》
How We Became Posthuman
玛格丽特·米德解决了二阶控制论的需求
Margaret Mead addressed the need for second-order
cybernetics
梅西会议，议题为“控制论”
Macy Conferences on Cybernetics
第一次控制论思潮
First wave cybernetics
第二次控制论思潮
Second wave cybernetics
第三次控制论思潮
Third wave cybernetics
大都会建筑事务所
“树城”方案
OMA's Tree City 由詹姆斯·科纳和斯坦·艾伦提出的
“生态演生”方案
Emergent Ecologies,
James Corner + Stan Allen
JCFO景观规划设计事务所改造弗莱士河公园
Freshkills, James Corner Field Operations
《设计结合自然》出版
Design with Nature
伊恩·L·麦克哈格
Ian L. McHarg
1950s
1946~1953 1967
1990s
1948
1973
1999
大都会建筑事务所参加
拉·维莱特公园设计竞赛
OMA's entry for Parc de la
Villette
1982 2000
2000
2003
1969
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
016 017
VOLUME 8 / ISSUE 4 / AUGUST 2020
framework, it starts to articulate prototyping outside the means-
end equation.
3.1 The Moth or the Bedbug, Cellular Automaton, and Artificial
Life
Early cyberneticians built cybernetic machines and simulations
as a way to test out their theoretical considerations. An important
artifact was Wiener’s cybernetic machine known as “the moth”
or “the bedbug.” In
The Human Use of Human Beings
, Wiener
discussed in detail the process of making this “cybernetic robot”
to build a working model of the interaction of two kinds of
feedback mechanisms in the human nervous system: postural
feedback and voluntary feedback (goal-seeking feedback)[19].
Wiener asserted that these two types of feedback mechanisms
exist in human behavior and nervous system disorders, and
Parkinsonism and intention tremors are the cases when these
feedback mechanisms are overloaded and broken down. Wiener
further posited that in Parkinsonism, voluntary feedback regulates
postural feedback: the tremor happens when the patients are
in rest, and when the patients perform a task (goal-seeking),
the tremor subsides or even disappears. Wiener expected to
conceptualize these theoretical explanations in action by building
an apparatus that could act based on the two kinds of feedback
mechanisms.
Wiener and his colleagues built a machine called “the
moth” or “the bedbug.” It was a three-wheeled cart with two
major modes of action, positively photo-tropic (the moth) and
negatively photo-tropic (the bedbug). This machine has two
types of feedback mechanisms with one regulating the other. For
example, to achieve the “the moth” function, the cart carries two
photo-cells in the front, and when there is light, the motor will
turn the wheel towards the direction of the light. This action will
trigger a negative feedback circuit that turns the front wheel in
the opposite direction. So, the cart will move towards the light
in an oscillating pattern (voluntary feedback). If there is no light,
the cart will keep oscillating because there is secondary feedback
between the potentiometer and the motors (postural feedback).
Moreover, when this feedback is overloaded by adjusting the
amplifier in the circuit, the oscillation would grow bigger. This
machine was argued to be an analog to Parkinsonism, in which the
voluntary feedback suppresses the postural feedback (Fig. 3).[17]
“The moth” can be understood as a physical diagram of
feedback mechanism, which was a key concept explored in
the first wave of cybernetic research in the Macy Conferences
era. However, its “usefulness” cannot be conceptualized within
the means-end equation because when the cyberneticians built
it, they did not have a specific goal in mind, such as solving a
期受限于手段－目的链思维方式。通过在演生框架内分析一系列反馈
生物，人们得以跳脱出手段－目的链的逻辑，转而重视原型思维。
3.1 “蛾”或“臭虫”、细胞自动机和人工生命
早期的控制论学者通过设计反馈机和模拟实验进行理论检验。
其中一项重要成果是由维纳设计的反馈机，也被称为“蛾”或“臭
虫”。在《人有人的用处》一书中，维纳详细讨论了如何通过制造反
馈机来构建实体模型，以演示人类神经系统中姿态反馈和自愿反馈
（或寻求目标时的反馈机制）两种机制间的相互作用［１９］。维纳断言，这
两种反馈机制存在于人类行为中，而神经系统疾病（如帕金森症和意
向性颤抖）就是由于这些反馈机制过载崩溃而导致的。维纳进一步指
出，在帕金森症中，自愿反馈会抑制和调节姿态反馈：患者在休息时
会发生颤抖，而当他们执行任务（即寻求目标）时，颤抖便会减弱甚
至消失。维纳系统通过设计一种装置来演示这两种反馈机制。
维纳和他的同事制作了一个叫做“蛾”或“臭虫”的装置，这个
装置像一个小三轮车，有正趋光性（蛾）和负趋光性（臭虫），有两
种可以相互调节的反馈机制。以“蛾”功能为例，小车前端装有两个
光敏管，当有灯光时，电机便会将前轮转向有光的方向。但此动作将
触发另一个负反馈电路，该负反馈电路会使前轮朝相反方向转动。因
此，这个小车便会摇摆着移向光源（有目标的自愿反馈）。当没有灯
光时，由于电位器和电机之间存在次级反馈，小车仍将摇摆着继续前
进（姿势反馈）。而通过调节电路中的放大器使该反馈电路过载时，
摇摆幅度会变大。因此，这个装置可以模拟帕金森症，其中的自愿反
馈可抑制和调节姿势反馈（图３）。［１７］
“蛾”可被看作反馈机制的实体示意模型，而反馈机制是第一
次思潮期间梅西会议中的一个重要概念。然而，手段－目的链不能解
释“蛾”的用途，因为当控制论学者制造它的时候并没有一个具体的
the third wave of cybernetics which focuses on open-endedness
and emergence. The team also proposed a framework that
allows for a long-term generation of the park based on emergent
ecologies. Unlike other finalists which used phasing programs
to build an urban park as an end point, Emergent Ecologies
considered the initial 15-year phasing program as a strategy to
cultivate the long-term potential of the site, providing multiple
possibilities of how the site could be used after its construction.
In other words, constructing a conceivable and stable urban
park was not the goal of this proposal; instead, the aim was to
prepare the Downsview site into a territory with a wide range
of possibilities. The proposal leaves plenty of space to imagine
a place that is wild and vibrant and is fundamentally different
from a pastoral urban park[16].
Contrasting landscape design with the cybernetics movement
highlights how control and uncertainty were conceptualized in
three paradigms of design and research (Fig. 2). Within a first-
order cybernetics framework, uncertainty represents the model-
environment difference that needs to be reduced by gaining
more knowledge about the phenomenon that we study. On the
other side of this reasoning, designers can acquire knowledge
by deploying proper means and close the gap between the
model and the system by building more complex stimulation
mechanisms, and uncertainty can be reduced and managed.
Within the autopoiesis framework (second wave of cybernetics),
uncertainty is instrumentalized as information and material
flow that supports the system to maintain its own identity,
thus legitimizing another kind of control regime that can
only be articulated with second-order cybernetics. Within the
framework of emergence and open-endedness (the third wave),
uncertainty is transformed into potentiality. It is no longer
a negative concept, like entropy in McHargian’s terms, that
needs to be regulated and controlled. Instead, because of the
uncertainties, the future can be articulated to have a wide range
of possibilities, thus becoming resourceful.
3 Cybernetic Creatures and Designers’ Simulation
In the framework of emergence and open-endedness,
prototypes act as an important component to help exploit the
potential embedded in the uncertainty. Since a prototype can
be understood as a special family of models, which should be
situated in the history of cybernetics movement and analyze
how it was conceptualized at large in systems thinking. In a way,
prototypes have always existed in cybernetics movement, but the
habitual means-end reasoning has limited their “usefulness.” By
analyzing a range of cybernetic creatures within the emergence
过生态演生的方式实现长期发展。不同于其他入围方案将建设城市公
园作为“最终目标”的分阶段式设计构想，“生态演生”将为期１５年
的最初阶段作为一种培育场地长期潜力的策略。这一设计将为建成后
的场地发展提供多种可能性。换言之，该方案不是为了建造一个按照
预期发展的、稳定的城市公园，而是意在把场地打造为一片潜力无穷
的区域。该方案为人们预留了足够的想象空间，描绘了一个生机勃勃
的场所，这是与“田园式城市公园”的本质区别［１６］。
通过比较景观设计和控制论运动，可以看出三种不同的设计研究
范式是如何阐释控制与不确定性的（图２）。在一阶控制论框架内，不
确定性表现为模型与环境间的差异，要想消除这种差异则需要不断获
取新知；另一方面，设计师通过采取合适的方法来获得更多的知识，
并通过构建更复杂的模拟机制来缩小模型与环境间的差距，从而消除
或控制不确定性。而在自生理论框架内（即第二次思潮），不确定性
成为了一种信息流和物质流，以及系统维持自身特性的一种手段，这
种控制手段只能通过二阶控制论来解读。而在开放和演生的框架内
（即第三次思潮），不确定性成为了一种潜能，它不再是一个负面的
概念，不再像麦克哈格所认为的熵那样需要加以规范和控制。相反，
正是由于不确定性，未来才充满无限可能。
3 反馈生物与模拟实验
在演生和开放的理论框架内，原型发挥着重要作用，它能够帮助
人们发掘不确定性中蕴含的潜力。作为一类特殊的模型，原型应该被
置于控制论的发展进程中来理解，并分析这一概念在系统思维中的形
成过程。在某种程度上，原型一直存在于控制论运动中，但其效用长
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
018 019
VOLUME 8 / ISSUE 4 / AUGUST 2020
transmit data to each other, then data disappear. Other rules guide
how each cell interacts with its neighbors.[19]
The artificial life experiment is a more complex version of
the cellular automaton. Inspired by autopoiesis theory, biologists
wanted to explore the implication of second-order cybernetics and
how self-production systems can spin off and evolve towards new
directions. One famous experiment is called Tierra—A computer
program designed to replicate and mutate based on simple rules.
The self-replicating program would randomly flip ones and zeros
on the computer memory when copying the data. If the new data
happened to be machine code for another program, the original
program mutated. A “reaper” function would “kill” old or
defective programs when the computer memory became full. By
repeating this simple process, a computer program could evolve
into different “ecosystems” with a wide range of digital species[20].
The artificial life experiment had a high value in the discourse of
posthumanism because it showed that complex behaviors, such
as human thinking, were epiphenomenon of interactions between
distributed systems, such as neural networks, and this realization
allowed posthuman proponents, such as Hayles, to challenge
human exceptionalism.[8]
The interpretation flexibility and undefined identity of these
examples of cybernetic creatures make them prototypes, which
are different from the predictive models used in contemporary
environmental practices such as adaptive management. A
predictive model entails a model-environment difference, as the
model has a metaphorical quality, and it represents a phenomenon
in reality. With predictive models, the gap between the model and
the environment is conceptualized as uncertainty that can only be
conceptualized as negative forces undermining control regimes.
Nevertheless, a prototype has a twofold quality. On the one hand,
like predictive models, a prototype has metaphorical quality,
that is, it uses one thing to refer to another thing; “the moth” is
a working diagram of feedback mechanisms. On the other hand,
it has a sense of realism in it because it exists as part of reality so
that other artifacts can be said to be derived from it. This twofold
quality makes prototypes a special type of model that does not
predict or determine, but inspire and anticipate.
3.2 A Case Study: Hydromorphology Table
In the scientific and engineering practice of physical hydraulic
modeling, large physical river models are built to simulate weather
and flood events and to evaluate the effect of flood control
strategies. Precedents include large-scale hydraulic models built in
the 1950s by the Army Corps of Engineers such as the Mississippi
Basin Model and the San Francisco Bay Model[21]. On the other
spectrum, digital models and simulations are also widely used
相传输信息，那么信息就会被清空。当然，还有其他一些用来决定相
邻细胞间互动的规则。［１９］
人工生命实验比细胞自动机的机制更加复杂。生物学家受到自生
系统论的启发，想要探索二阶控制论的广泛意义，以及自生系统如何
脱离其原有轨迹并朝新的方向发展。其中一项著名实验是Ｔｉｅｒｒａ程序的
编写，该程序可以基于简单的规则进行自我复制和变异，当程序将代
码写入内存条时，会随机反转１和０形成一串新代码。假如新代码恰好
是某个其他程序的机器代码，那么旧有程序就可以说是发生了变异。
一旦计算机内存溢出，“淘汰”功能就会把旧有的或有缺陷的程序
“杀死”。通过重复这一简单过程，一个电脑程序可以进化出五花八
门的“数字物种”，形成各式各样的“生态系统”［２０］。人工生命实验对
于后人类主义的讨论起着重要作用，它证明了复杂行为（如人类的思
考行为）不过是分布式系统（如神经网络）之间的相互作用产生的附
带现象而已，而这样的认知也为海尔斯在内的后人类主义支持者挑战
人类例外论提供了理论支持。［８］
这些反馈生物的范例没有明确定义，可被灵活解读，这些特性
使它们成为了不同于当代环境设计实践（如适应性管理）中使用的预
测性模型的原型。预测性模型具有隐喻性，用于反映某种现实中的现
象，也与现实环境存在差异。因此，在预测性模型中，这种差异被视
为威胁控制策略的不确定性。然而，原型却具有双重属性。一方面，
和预测性模型一样，原型也具有隐喻性，即用一种事物来指代另一种
事物
—
“蛾”即为反馈机制的示意模型；另一方面，原型来自于现
实世界，故而具有现实性，可以不断衍生出其他事物。这种双重特性
使原型成为一类特殊的模型，这种模型不用于预测或决策，而是为人
们带来启发和洞见。
3.2 案例研究：水文地貌模拟沙床
在水力学模型的研究和工程实践中，大型物理河流模型通常用于
模拟气象和雨洪事件，以及评估防洪措施的效果，如２０世纪５０年代由美
国陆军工兵团建造的密西西比河流域模型和旧金山湾区模型［２１］。另一方
面，数字模型和模拟也广泛用于环境设计实践中，以预测系统行为并
problem. Thus “the moth” is not a means to an end. Cyberneticians
were simply interested to see the cybernetic principles in action. In
the framework of emergence, “the moth” can be understood as a
prototype, whose value lies in its uncertainty. Because it functions as
an abstracted diagram of feedback mechanism, it can be interpreted
differently among different fields, and the proliferation of meaning
is the basis of creativity and design. It became a kind of boundary-
object that facilitated the thinking and communication of cybernetic
principles between scholars from different fields and played a
critical role in popularizing the cybernetic principles in its early
stage.[18]
The other two examples—von Neumann’s cellular automaton
and artificial life experiment—can be discussed as a seriation. The
cellular automaton was designed to provide insight into the logical
requirements for self-replicating machines. It can be understood as a
2D grid, and each cell presents a finite-state automaton or machine,
which can be in one of a finite number of states at any given time.
Cells can transmit input to its adjacent cells. In von Neumann’s
model, each cell has 29 different states, which can be understood as
rules for transition signals. For example, if a cell is in the sensitized
state for three cycles, and it receives an input from an adjacent cell,
it will enter a transmission state that will pass the data to the cell to
its north. There are also rules for transmission state. For example,
when an input is given to a cell that is in the ground state (an empty
cell), the cell will enter the sensitized state; if somehow two cells
目标（比如要解决某个问题），所以“蛾”也并非达成目的的手段，
学者们仅想通过它生动地演示控制论原理。在演生框架内，如果将
“蛾”看作原型来理解，那么它的价值恰好在于其不确定性。由于它
是反馈机制的一个抽象的示意模型，因此不同领域会对它产生不同的
解读，而这种发散性思维正是创作和设计的基础。由此，“蛾”成为
了一种存在于不同领域之间的临界物，帮助学者思考和讨论控制论原
理，这对于早期控制论原理的普及有着重要作用。［１８］
另外两个需并置讨论的例子分别是冯・诺伊曼的细胞自动机和
人工生命实验。设计细胞自动机是为了研究“自我复制机”的逻辑需
求。它可被看作一个二维网格，每一个单元格（细胞）都代表了一个
有限状态自动机，在任何特定时间内，每一个单元格都处于一种有限
状态之中，都可向相邻单元格传输信息。在冯・诺伊曼的模型中，每
个单元格都有２９种不同的状态，代表了２９种信息传输规则。例如，若
一个单元格在三个周期内都处于被激活状态，那么当它接收到来自相
邻单元格的信息后，则会进入传输状态，并将信息传输至其上方单元
格中。关于传输状态也有一些规则，例如，当处于基础状态的空单元
格接收到信息时，空单元格便会进入被激活状态。如果两个单元格互
3. “蛾”或“臭虫”
原型（改绘自参考
文献[17]）
3. The prototype
of "the moth"
or "the bedbug"
(Adaptation source:
Ref. [17])
3
© Norbert Wiener
光源
Light source
光源
Light source
光敏管
Photo-cell
放大器
Amplifier
桥接
Bridge
放大器
Amplifier
桥接
Bridge
定位电机1
Positioning motor 1
定位电机2
Positioning motor 2
驱动电机
Driving motor
电位器 Potentiometer
光敏管
Photo-cell 姿态反馈
（光搜索）
Postural feedback
(Light searching)
反馈过载时振幅增大
Oscillation growing bigger
when the feedback is
overloaded
自愿反馈
（无光）
Voluntary feedback
(No light)
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
020 021
VOLUME 8 / ISSUE 4 / AUGUST 2020
器”的系统，该系统由连接到伺服马达的一组亚克力杆组成。每个亚
克力杆均由伺服马达单独驱动，其底部插入沉积物中。当伺服马达转
动时，会驱动亚克力杆上下移动来改变水流形态，从而在沙床的下游
形成不同的地形。随后，通过沙床上方的Ｋｉｎｅｃｔ传感器实时跟踪地形，
形成数字高程模型，以识别出地势的高低。这些信息会反馈到制动系
统，这样就能在高地上借由更多的沉积物来造陆，或引流更多的水体
来侵蚀陆地。埃斯特拉达描述道，水文地貌模拟沙床构建的控制系统
具有实时更新和反馈能力，而这种自主能力是人类无法企及的［２５］。
在另一个示例中，设计师刘浔基于实体界面（ＴＵＩ）的概念开发了
一种混合仿真模型［２６］。ＴＵＩ逐渐取代了传统的图形用户界面（ＧＵＩ），
它结合了物理模型和数字模型的优势，现已被广泛用于增强物理模型
性能以及物理模型的数据可视化，从而实现“比特和原子”的无缝耦
合［２７］。自定义程序使用从Ｋｉｎｅｃｔ传感器感测到的高程数据来生成实时
数字地形模型，该模型可在Ｒｈｉｎｏ和其他软件中用于进行水流、植被分
布、污染物和沉积物等不同类型的数值模拟（图４）。地形变化、景观
turn, they drive the dowels moving up and down to change the
flow pattern, creating different landforms in the downstream
of the table. The topography is then live tracked by the Kinect
sensor above the table, forming a digital elevation model of the
topography so that a series of high grounds and low grounds
can be identified. This information then feeds back to the
actuating system so that it could either build more land in a high
ground by depositing more sand on it or erode the high ground
away by directing more water towards it. The designer described
that the cyborg system built with the hydromorphology table
exhibited a level of live updates and feedback that was beyond
human capacity[25].
In another example, designer Liu Xun developed a hybrid
simulation based on the concept of the Tangible User Interface
(TUI)[26]. TUI is increasingly accepted as an alternative
paradigm to the conventional Graphic User Interface (GUI). It
combines the benefits of physical and digital models in the same
representation and is widely used to enhance the performance
of physical models and visualize the unseen data beneath
the physical model, creating a seamless coupling of “bits and
atoms”[27]. The customized program uses the sensed elevation
data from the Kinect sensor as input to generate a live digital
terrain model, which can be used in Rhino and other softwares
to conduct numerical simulations for different purposes such as
water flow, vegetation distribution, pollution, and sedimentation
(Fig 4). Topographic changes, landscape morphology, water
streamline, and patterns of other simulated agents can be
in environmental practices to predict system behaviors and
evaluate control strategies, such as the South Florida Water
Management Model[22]. These models, physical or digital, are all
envisaged in the line with means-end relationships for prediction
and control. The models begin as an ideal representation of the
environment so that they can be projected and used to give form
to reality. However, as pointed out by Jullien, following this line
of reasoning, there will always be unexpected circumstances
outside the scope of predictions that are allowed by the models.
Thus, one is constantly challenged by the tension between
control and uncertainty. Development in digital tools and
computational methods has provided landscape architects with
advanced techniques to build prototypes that challenge the
predictive models and means-end reasoning. To better illustrate
how prototyping helps with design thinking, this paper presents
a series of case studies based on a hydromorphology table
to demonstrate the qualities of prototypes and the value of
prototyping.
The hydromorphology table was based on the Emriver
stream table and was located at the REAL in Harvard
GSD②. The Emriver stream tables, which were usually used
by hydrologists, were introduced to the field of Landscape
Architecture as a prototyping platform to help designers better
understand hydro-morphological processes, and to facilitate the
production of responsive design strategies. The inputs of both
the sediments and water flow were controlled through 4 material
feeders and a water pump to simulate water flows and behaviors
of sediment. The table allowed for running iterations of the
same hydromorphological system with measurable adjustments,
as well as testing multiple systems with the same flow and
sediment input. Besides the instruments provided by the
Emriver, the prototype platform was also equipped with sensing
and monitoring devices to gather real-time data, including a
Microsoft Kinect above the table and ultrasonic sensors down
the stream. The real-time data then were fed into Rhinoceros 3D
through grasshopper plugins and customized interfaces.
As a prototyping platform, the hydromorphology table
facilitates many research and design projects over the years,
shedding light on the possibility of constructing autonomous
systems that can devise strategies beyond human comprehension
to create “wild” places[23]. For instance, designer Leif Estrada
tested the sensing-processing-actuating responsive framework
in the project Towards Sentience[24]. In one of the design
experiments, the designer proposed an actuating system called
“attuner” that consists of a matrix of acrylic dowels connected
to servo motors. Every servo motor drives a dowel, the bottom
of which sticks into the sediments. When the servo motors
评估控制策略，如南佛罗里达州水管理模型［２２］。物理模型和数字模型都
是通过手段－目的链来进行预测和控制。这些模型代表了某种理想状
态，通过映射在环境中来指导环境改造。但正如朱利安所言，如果我
们沿用手段－目的链思维方式来思考问题，总会遇到模型预测范围之
外的意外情况，那么控制与不确定性之间的矛盾似乎永远无法调和。
数字工具和运算方法的进步为景观设计师提供了有别于预测性模型和
手段－目的链思维方式的新的原型。为了更好地阐述原型如何辅助设
计思维，本文将介绍一系列基于水文地貌模拟沙床（以下简称“沙
床”）的案例，以进一步说明原型的特性以及原型思维的价值。
哈佛ＧＳＤ的ＲＥＡＬ实验室基于Ｅｍｒｉｖｅｒ径流模型建造了一个沙床②。
Ｅｍｒｉｖｅｒ径流模型这一常用于水文研究的模型被作为一种原型思维平台
而引入景观设计学领域，以帮助设计师更好地了解水文地貌变化的过
程，并辅助制定响应式设计策略。设计师可以通过４个进料器和一个水
泵来控制沉积物和水流的输入，用以模拟水流和沉积物的变化。
此外，设计师不仅能够通过调整变量来研究同一水文系统的迭
代，也能在相同流量和沉积物输入的条件下测试不同系统的径流情
况。除了Ｅｍｒｉｖｅｒ模型本身自带的仪器外，ＲＥＡＬ实验室的原型平台还
配备了感应和监控设备（如在沙床上方安装的微软体感设备Ｋｉｎｅｃｔ，
以及位于沙床下游的超声波传感器），以收集实时数据。收集到的
数据通过Ｇｒａｓｓｈｏｐｐｅｒ软件中的自定义插件实时传输到三维建模软件
Ｒｈｉｎｏｃｅｒｏｓ中。
近年来，作为一种原型思维平台，沙床被广泛运用于研究和设计
项目中，这些项目展现了构建自主系统的可能性，而自主系统能提出
超越人类理解范畴的策略并创建“野生”场所［２３］。例如，设计师雷夫・埃
斯特拉达在“走向感知”项目中，测试了“感应－处理－制动”的响
应式框架［２４］。在一项设计实验中，埃斯特拉达提出了一种称为“调制
4
© 刘浔
4. 自定义水文模拟和可视
化程序
4. Customized digital
hydrological simulation
and visualization tool
② 了解更多关于Emriver径
流模型及哈佛GSD REAL
实验室沙床模型的信
息，请访问Emriver径流
模型及REAL实验室官方
网站。
② For more information
about Emriver
stream table and the
hydromorphology table
at the REAL in Harvard
GSD, please visit the
websites of Emriver
stream table and REAL.
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
022 023
VOLUME 8 / ISSUE 4 / AUGUST 2020
形态、水流线，以及其他模拟图像都可通过短程投影仪投射到沙床上
（图５）。如此，基本的地形分析与动态流体模拟被集成到同一个平台
中，可以更好地实现地形数据的可视化，数字模拟和物理模拟也能通过
反馈回路实现耦合。设计师可以依据沙床上的投影信息（如等高线和填
方图）即时手动调整物理模型（图６￣９）。
增强现实的沙床可成为景观设计师新的设计工具，也可被当作一
种合作平台，帮助社区参与中的不同利益相关者进行沟通协商。该原
型并非基于从设计构思到计算机模拟、再到材料处理和结果呈现的线性
设计过程，而是开启了一种新的工作流程，其支持设计过程中的不同步
骤同时或反复进行。本实验中讨论的实体交互概念超越了界面的接触，
是指两种不同类型的模拟的协同与组合：数字模拟可用于验证物理模
拟，物理模拟可用于检测数字模拟，两种模拟相互配合并彼此适应。
通过运用增强现实技术将物理河流模拟与实时的数字河流模拟相整
合，该原型使景观设计师可以直观地利用有形的材料进行设计，并同
步获得计算模拟结果的校验。作为设计师与实体空间之间的动态界面，
实体交互模型的变化性和不确定性对于理解水力形态过程的复杂性和动
态过程非常重要。
需要特别指出的是，沙床模型并没有按照真实景观比例来建造，
所以该原型不宜用来模拟特定地点的水文和生态系统。换句话说，如果
以手段－目的链来理解沙床的概念，那么沙床必须按照真实环境的比例
来建造，这样才能用于预测设计结果并决定相应策略（如密西西比河流
域模型或南佛罗里州达水管理模型）。但是，围绕模型制作的讨论最终
离不开如何用模型“精确”地表达环境的问题，这与模拟参数的不确定
性有关。在建模中，造成不确定性的原因有很多，如结构的不确定性、
算法的不确定性等。其中，参数的不确定性来自模型输入变量的可变
性。对于沙床而言，设计人员可以通过调整介质的大小以及不同介质之
间的比例改变河道“沉积物”的组成，进而调节水流的速度和流量。这
些都是模型本身包含的变量，更改其中的任何变量都将对模拟结果产生
显著影响。如果将沙床视为一个预测性模型，那么设计人员需要计算建
模过程中的所有参数值，以最大程度地减少参数的不确定性。设计人员
不仅必须选择合适的材料尺寸来模拟真实的景观环境，而且需要确定恰
当的材料构成比例，更不用说计算缩放过程中涉及的物理变化，因为理
论上除了水以外，所有事物都被缩小了。为了解决这个问题，设计人员
必须找到另一种具有适当粘滞阻力的流体，以代表流体与要按比例缩放
的固体颗粒之间的相互作用。由此可知，如果遵循手段－目的链来理解
沙床，会不可避免地将它看作一种重构现实世界的理想模型，而这种想
法会导致不确定性永远存在。最重要的是，这种基于手段－目的链框架
visualized in diagrammatic representations. These representations
then can be projected back onto the hydromorphology table with
a short-throw projector (Fig. 5). In such a way, basic terrain
analysis is integrated into this platform together with dynamic
fluid simulation to better visualize the invisible aspects of the
topography, and the digital simulation and the physical
simulation are coupled through a feedback loop. Designers
can manually change the physical model according to the
projected information, such as contour lines and cut-and-fill
maps (Fig. 6 ~ 9).
This augmented hydromorphology modeling table can
be used as a new design tool for landscape designers, as well
as a cooperation platform for community engagement and
communication between different stakeholders. Rather than
developing design in a linear progression from idea to computer-
simulated model, and material manipulation and result, this
prototype inspires an alternative workflow that allows different
steps in the design process to play concurrently or recursively. The
concept of tangibility developed in this experiment is more than
the tangibility of the interface, but the synthesis and composition
of the two different types of simulation: numerical models being
used to manipulate while validating physical models, and vice
versa. Two kinds of simulations feed and adapt to each other.
By integrating physical hydraulic simulation with the real-time
computational fluvial simulation through augmented reality
technologies, this prototype allows landscape architects to design
intuitively with tangible material processes and simultaneously be
informed by computational simulation results. The dynamics
and indeterminacy of the tangible model are important in
terms of understanding the complexity and dynamic process
of the hydro-morphological process. This prototype acts as a
dynamic interface between designers and the physical world.
Particularly, this hydromorphology table as a prototype is
not meant to replicate processes in a site-specific hydrological
and ecological system. The presumed scale of the model is
not built based on a mathematically scaled relationship with
real landscape conditions. In other words, if the table is
understood in the means-end relation, it can only exist as a
scaled simulation of the real environment so that to predict
and determine outcomes of different intervention strategies,
like the Mississippi Basin Model or the South Florida Water
Management Model. However, the discussion about model-
making would eventually revolve around the question of how
“accurate” the model is to represent the environment; that
is, the parameter uncertainty involved in the simulation. In
modeling practices, uncertainties come from various sources
such as structural uncertainty and algorithmic uncertainty.
Parameter uncertainty comes from the variability of input
variables of the model. In the case of the hydromorphology
table, the designer can change the composition of the
“sediment” by adjusting the size of each material and the
percentage of that material in the sediment mixture. The
designer can also vary the flow of water in terms of speed
and quantity. These are all implicit variables involved in the
model itself, and changing any of them could easily alter the
outcome of the simulation. If the table is treated as a predictive
model, then the designer is challenged to minimize parameter
uncertainties by figuring out the values of all the parameters in
the modeling process. The designer has to select not only the
correct size of the material to emulate the sediment on-site but
also the composition of the material that best represents the
sediment. Not to mention that the designer has to calculate the
underlying physics involved in the scaling process itself, since
everything but water is scaled down; and to counter this issue,
the designer might have to switch to another type of fluid with
the right viscosity that best represents the interaction between
the fluid and the solid particles to scale. These examples only
provide a glimpse of the line of reasoning that one inevitably
takes if one follows the means-end equation and understands
the hydromorphology table as a model of an ideal form to
reconstruct the reality. In such conceptualization, uncertainty
always exists. Most importantly, following the means-
end equation, the discussion drifted away from the design
practice and creativity. However, if designers understand the
6
© 刘浔
7
© 刘浔
8
© 刘浔
9
© 刘浔
6. 增强现实沙床地形实验
7. 基于智能体的模拟。设
计师能通过投影，直观
感受设计干预对水文系
统的影响。
8. 现实增强的地形分析
9. 软件操作界面
6. Augmented
hydromorphology table:
landform experiment
7. Agent-based
simulation. Designers
can observe the direct
impact of design
intervention on the
hydrological systems
through projected
simulation.
8. Augmented
topographical analysis
9. Software interface
5. 增强现实沙床控制系统
5. Augmented
hydromorphology table:
control system
5
© 刘浔
投影仪
Projector
体感设备
Kinect
预测数值模拟
Projected numerical simulation
物质模拟
Physical simulation
排口调节器
Outflow adjuster
坡度调节器
Slope adjuster
水泵
Water pump
流量控制器
Flow controller
介质供应器
Media feeders
计算机
Computer
介质／流出物收集器
Media / outflow collector
景观设计学 / 论文 LANDSCAPE ARCHITECTURE FRONTIERS / PAPERS
024 025
VOLUME 8 / ISSUE 4 / AUGUST 2020
下的讨论偏离了设计和创造性。然而，如果将沙床理解为一种原型，
将它视作一个自成体系的环境，一种自主的、不断演化的系统，设计
师就能探索更多的设计可能性（如响应式框架），从而避免陷入控制
和不确定性的无解争取，景观的未来才能拥有更多的机遇。
4 结论与讨论
控制论运动和景观设计的比照揭示了自２０世纪６０年代以来的现代
景观设计发展历程反映着控制论的三次思潮。不确定性概念起初被理
解为模型与环境间必须要被消除的差异，而后来则被视为演生与机遇
的来源。在演生论的框架下，原型被解读为一种特殊的模型，将其与
预测性模型进行对比，会发现二者对于不确定性有着截然不同的理解
思路。利用预测性模型思考问题，会不由自主地通过手段－目的链来
理解不确定性，环境也因此被简化成了一种理想化模型，任何的行为
都必须具有目的性，必须通过手段来实现一个最终目标。不同于预测
性模型所追求的生命系统，原型本身就是一个独立自主的生命系统，
不存在与现实环境的差异，因为原型正是环境的一部分。用原型思维
取代建模思维后，不确定性也具有了新的含义
—
更多的机会与期
待。在原型思维范式中，不确定性将激发演生行为，以及无限的可能
性（图１０）。
原型思维范式具有如下三个特点。首先，原型最初可能只用于测
试某一原理或想法，但一旦被制造出来，便会成为现实的一部分，从
而引发新一轮的观察与建模。这种迭代过程是设计和创造力的基础。
其次，尽管原型可直接用于某些场景，但其真正的用途恰恰在于其未
被限定用途
—
“蛾”、沙床等原型甚至没有专业名称
—
但正由于
不确定性，原型才具有无限的潜力。最后，这种无定义的状态，以及
原型所具有的不确定性和无限潜能将改变人们与未来之间的关系。
hydromorphology table as a prototype, then their relationship
with the future is in the line of emergence and opportunity.
By regarding this model as its own environment, with its own
autonomous and evolving processes, designers can explore
a wide range of possibilities of intervention, such as the
responsive framework, thus bypassing the discussion about
control and uncertainty.
4 Conclusion and Discussion
Contrasting the cybernetics movement with landscape
design reveals that the development in landscape discourse
since the 1960s mirrors the movement of cybernetics.
Uncertainty has transformed from the early understanding that
the disconnection between the model and the environment
has to be reduced to a contemporary understanding that
uncertainty can be the source of emergence and opportunity.
The framework of emergence highlights prototypes as a special
type of model. Prototypes can be contrasted with predictive
models as they entail two different lines of reasoning around
the issue of uncertainty. With predictive models, we are forced
to understand uncertainty within a means-end relationship
that reduces the environment to a simplified representation or
ideal image, and then make it a goal and an end, which calls
for means to achieve it. Rather than predictive models that
represent another living system, prototypes are living systems
in themselves with autonomy and lives, and no longer entail
a model-environment difference because they are part of the
environment. Prototyping replaces model-making and helps
establish a new relationship with uncertainty in the line of
opportunity and anticipation. Uncertainty in the paradigm
of prototyping can be understood as the source for emergent
behaviors and a wide range of possibilities (Fig. 10).
There are three characteristics of the paradigm of
prototyping. First, prototypes might start as a test of a principle
or an idea, but when they are built, they take on a life of their
own and become part of reality that triggers new rounds
of observation and model-making. This iterative process is
the basis for design and creativity. Second, prototypes may
have some direct applications, but their true application lies
in their undefined identity. “The moth” did not even have a
proper name, and neither did the hydromorphology table.
However, just because of this undefined identity, a prototype
has the potential to become anything. Third, because of this
undefined quality of prototypes, they are full of uncertainty,
thus infinite possibility, and our relationship with the future
has changed.
REFERENCES
[1] Spirn, A. W. (1995). Constructing Nature: The Legacy of
Frederick Law Olmsted. In W. Cronon (Ed.), Uncommon
Ground: Toward Reinventing Nature (pp. 91-113). New York:
W.W. Norton and Company.
[2] Prototype. (2000). In Oxford English Dictionary. Oxford:
Oxford University Press.
[3] Jullien, F. (2004). A Treatise on Efficacy: Between Western
and Chinese Thinking (J. Lloyd, Tran.). Honolulu: University
of Hawaii Press.
[4] Sadler, J. M., Goodall, J. L., Behl, M., Bowes, B. D.,
& Morsy, M. M. (2020). Exploring real-time control of
stormwater systems for mitigating flood risk due to sea
level rise. Journal of Hydrology, (583), 124571. https://doi.
org/10.1016/j.jhydrol.2020.124571
[5] Fung, S. (1999). Mutuality and the Cultures of Landscape
Architecture. In J. Corner (Ed.), Recovering landscape:
Essays in contemporary landscape theory (pp. 141-151).
Hudson: Princeton Architectural Press.
[6] Herrington, S. (2016). Landscape Theory in Design. London:
Routledge.
[7] Lystra, M. (2014). McHarg’s Entropy, Halprin’s Chance:
Representations of Cybernetic Change in 1960s Landscape
Architecture. Studies in the History of Gardens and
Designed Landscapes, 34(1), 71-84. https://doi.org/10.1080
/14601176.2013.850313.
[8] Hayles, N. K. (1999). How We Became Posthuman: Virtual
Bodies in Cybernetics, Literature, and Informatics. Chicago:
University of Chicago Press.
[9] Maturana, H. R., & Varela, F. J. (1980). Autopoiesis and
Cognition: The Realization of the Living. Heidelberg:
Springer Netherlands. https://doi.org/10.1007/978-94-
009-8947-4
[10] Steiner, F. R. (2008). The Ghost of Ian McHarg. Log, (13/14),
147-151.
[11] OMA Office Work Research. (n.d.). Parc de la Villette.
Retrieved from https://oma.eu/projects/parc-de-la-villette
[12] OMA Office Work Research. (n.d.). The Downsview Park
competition. Retrieved from https://oma.eu/projects/
downsview-park
[13] Czerniak, J. (2001). Downsview Park Toronto. New York:
Prestel.
[14] Hill, K. (2001). Urban Ecologies: Biodiversity and Urban
Design. In J. Czerniak (Ed.), Downsview Park Toronto (pp.
90-101). New York: Prestel.
[15] Reed, C., & Lister, N. M. (2014). Ecology and Design:
Parallel Genealogies. Places Journal. Retrieved from
https://placesjournal.org/article/ecology-and-design-
parallel-genealogies/
[16] Berrizbeitia, A. (2001). Scale of Undecidability. In J.
Czerniak (Ed.), Downsview Park Toronto (pp. 117-125).
New York: Prestel.
[17] Wiener, N. (1950). The Human Use of Human Beings:
Cybernetics and Society. Boston: Houghton Mifflin.
[18] Star, S. L., & Griesemer, J. R. (2016). Institutional Ecology,
‘Translations’ and Boundary Objects: Amateurs and
Professionals in Berkeley’s Museum of Vertebrate Zoology,
1907-39. Social Studies of Science, 19(1989), 387-420.
[19] Von Neumann, J., & Burks, A. W. (1966). Theory of Self-
Reproducing Automata. Urbana: University of Illinois Press
[20] Ray, T. S. (n.d.). Tierra. Retrieved from https://web.stanford.
edu/class/sts129/Alife/html/Tierra.htm
[21] Cheramie, K. (2011). The Scale of Nature: Modeling the
Mississippi River. Retrieved from https://placesjournal.org/
article/the-scale-of-nature-modeling-the-mississippi-river/
[22] South Florida Water Management District. (n.d.). South
Florida Water Management Model (SFWMM). Retrieved from
https://www.sfwmd.gov/science-data/sfwmm-model
[23] Cantrell, B., Martin, L. J., & Ellis, E. C. (2017). Designing
Autonomy: Opportunities for New Wildness in the
Anthropocene. Trends in Ecology and Evolution, 32(3), 156-
166. https://doi.org/10.1016/j.tree.2016.12.004
[24] Cantrell, B., & Holtzman, J. (2016). Responsive landscapes:
Strategies for responsive technologies in landscape
architecture. New York: Routledge.
[25] Estrada, L. (2018). Towards Sentience. In B. Cantrell &
A. Mekies (Eds.), Codify: Parametric and Computational
Design in Landscape Architecture (pp. 279-288). New York:
Routledge.
[26] Liu, X. (2020). The Third Simulation: Augmented Reality Fluvial
Modeling Tool. Journal of Digital Landscape Architecture, (5),
413-421. https://doi.org/doi:10.14627/537690042
[27] Ishii, H., & Ullmer, B. (1997). Tangible Bits: Towards Seamless
Interfaces between People, Bits and Atoms. In Proceedings
of the SIGCHI Conference on Human Factors in Computing
Systems (pp. 234-241). Retrieve from: http://ebusiness.mit.
edu/sponsors/common/LnchSemExecSum/ishii.pdf
10. 预测性模型与原型思
维关系图
10. The relationship
between predictive
modeling and
prototyping
10
© 张子豪，刘浔
手段－目的链
Means-end chain
理论
Theory
实践
Practice
代表
Representation
现实
Reality
未定义的特性
Undefined identities
未定义的特性
Undefined identities
反复学习
Recursive learning
反复学习
Recursive learning
现实
Reality
隐喻转译
Metaphorical translation
目的驱动的评估
Goal-driven evaluation
广泛的可能性
Wide range of possibilities
感知
Sensing
原型1
Prototype 1
原型2
Prototype 2
原型3
Prototype 3
不确定性=模型／环境差异
Uncertainty = model / environment differences
不确定性=演生与可能性
Uncertainty = emergence and possibility
预测／控制
Predict / control
预测性 模 型
Predictive
models
环境
Environment
演生性范式
Paradigm of emergence
