Frank Knight lived in uncertain times. Born on a farm in Illinois in 1885, between 1900 and 1910, he moved with his family to Colorado, where his father owned a fruit orchard (Knoop, 2010). As small business owners, the Knights likely experienced the adverse effects of the almost biennial recessions, depressions, and financial panics that roiled the U.S. economy at the turn of the 20th century (Knoop, 2010). The first global war—World War I—began in Europe in 1914 (when Knight was 29) and included the United States by 1917. Four of Knight's brothers served in the military in this “war to end all wars”—a war that saw 9 million soldiers and 13 million civilians killed (MacMillan, 2013). And in 1918—when 33‐year‐old Knight revised his dissertation for publication—the world experienced a flu pandemic that killed between 19 and 50 million people worldwide. Some have estimated that the death toll from this pandemic was as high as 100 million (Spinney, 2017).11 Historical information about Knight's family was obtained from 1900 U.S. census, McLean County, Illinois, population schedule, Lawndale, p. 6, dwelling 104, family 104, Winton and Julia Knight; digital image, Ancestry (http://www.ancestry.com: accessed June 16, 2020); citing NARA microfilm publication T623. 1910 U.S. census, Mesa County, Colorado, population schedule, Pomona, p. 14B, dwelling 297, family 308, Winton C. and Julia A. Knight; digital image, Ancestry (http://www.ancestry.com: accessed June 16, 2020); citing NARA microfilm publication T624. “U.S., World War I Draft Registration Cards, 1917–1918,” images, Ancestry (http://www.ancestry.com: accessed June 16, 2020), card for Melvin M. Knight, serial no. 20637, Local Draft Board, Clark University, Worcester, Massachusetts; also card for Frank Hyneman Knight, serial no. 964, Local Draft Board No. 15, Chicago, Illinois. “Colorado, Soldiers in WWI, 1917–1918,” database, Ancestry (http://www.ancestry.com: accessed June 16, 2020), entries for Paul Edwin Knight, Mark Knight, Bruce Winton Knight; citing Roster of Men and Women Who Served in the World War from Colorado, 1917–1918. Colorado, USA; Adjutant General, Colorado National Guard, 1941. “U.S., Department of Veterans Affairs BIRLS Death File, 1850–2010,” database, Ancestry (http://www.ancestry.com: accessed June 16, 2020), entries for Paul Knight, Bruce Knight, Mark Knight.

Thus, it is perhaps not surprising that Knight's first published research focused on the difference between risk and uncertainty and particularly on the implications of uncertainty for businesses (Knight, 1921). For Knight, decision‐making settings were risky when decision makers did not know, ex ante, what an optimal decision was but did know both the possible outcomes of a decision and their probability. Decision‐making settings were uncertain when decision makers did not know, ex ante, what an optimal decision was and also did not know the possible outcomes of a decision nor their probability.

Coase (1988: 1) once observed that his work was “much cited, but little used.” The same can probably be said about Knight's work on uncertainty. The problem with uncertainty, from the point of view of traditional economics, was that it is often not possible to write formal models under conditions of uncertainty, while such models can be written under conditions of risk (Stigler, 1985). So, while acknowledging the existence of uncertainty,22 Some economists deny the existence of uncertainty altogether, arguing that because decision makers always operate with a probability distribution in their mind, decisions are always made under conditions of risk (Savage, 1954). See Alvarez and Barney (2020) for a summary of the behavioral economic critiques of such a model of human decision‐making. economists focused on the risky conditions where they could adopt their preferred predictive methodology.

However, in the last few years, interest in Knightian uncertainty has grown, primarily among management scholars (Alvarez & Barney, 2005; Milliken, 1987; Teece & Leih, 2016). While interest in Knightian uncertainty spanned many areas of management (Burns & Stalker, 1966; Lawrence & Lorsch, 1967; Thompson, 1967), it is the entrepreneurship scholars that have more recently embraced Knightian uncertainty and its implications for research (Alvarez & Barney, 2007; McMullen & Shepherd, 2006; Sarasvathy, 2001). Therefore, it is not surprising that, when “Discovery and Creation: Alternative theories of entrepreneurial action” was published, it generated significant interest. While many scholars had written about uncertainty and its difference with risk, the Discovery and Creation paper was the first to systematically articulate those differences and what they meant for theories of opportunity formation.

Knightian uncertainty is a key assumption of the Creation Theory of the formation and exploitation of entrepreneurial opportunities and is growing in importance (Alvarez & Barney, 2007; Alvarez, Barney, & Anderson, 2013). In contrast to discovery views of entrepreneurship that have assumptions of risk, Creation Theory suggests that entrepreneurs often begin the process of forming opportunities—defined as competitive imperfections in a product or factor market (Alvarez & Barney, 2020)—with limited or no information about the characteristics of the opportunity they may ultimately create. That is, consistent with Knight's definition of uncertainty, entrepreneurs begin the opportunity formation process in conditions of Knightian uncertainty before the outcomes of the creation process can be anticipated even probabilistically. Through an iterative path‐dependent learning process undertaken through a series of actions, entrepreneurs can sometimes end up creating an opportunity that did not exist before they acted to create it.

Putting Knightian uncertainty front and center in understanding the process of forming entrepreneurial opportunities in the 2007 paper seems remarkably prescient given recent events that eerily parallel the setting within which Knight did his original work—including the emergence of the global Covid‐19 pandemic, associated economic upheavals, and growing racial tensions connected to the death of George Floyd and others. All these events were difficult, if not impossible, to anticipate before the actions of individuals set them in motion. Of course, that is what makes human action important—these actions can set Knightian uncertainty into motion with individuals often unaware of doing so. What, if anything, can the concept of Knightian uncertainty do to be helpful in this context?

First, a central feature of the formation of entrepreneurial opportunities under conditions of Knightian uncertainty is that, in the beginning, what is not known is substantially greater than what is known, and what we think is known may not be relevant as human actions bring about change. In the case of the pandemic, we knew very little about the pandemic when it started, but over time, we have learned more about the spread of the virus and the treatment of those infected. Initial efforts to reduce uncertainty were, at best, informed guesses. In the early days, data are often nonexistent or incomplete. As we have witnessed with the current pandemic, conventions and techniques used during stable periods in the treatment of known flu types have been rendered useless. What we knew about medical care and the scientific method was often not relevant as we began to learn more about the situation. Those looking to reduce uncertainty often do not know the right questions to ask, let alone how to answer those questions.

Second, ex post, it is easy to attribute the many missteps taken by individuals in the early stages of Knightian uncertainty to incompetence. And, indeed, some of these steps may be the result of incompetence, but they may also reflect the fundamental uncertainty facing decision makers in the early stages of resolving uncertainty.

Third, the experimentation conducted early in the process of Knightian uncertainty often generates incomplete and even contradictory results. Knightian uncertainty makes it difficult to design optimally informative experiments. What is typical in these conditions is that a person tries one thing, and if it does not work, they try something else. If something works, then they do more of it. As the famous quote of Lord Maynard Keynes replying to a heckler illustrates “When my information changes, I alter my conclusions. What do you do, sir?” This is perhaps most exemplified by the changing mask recommendations. The public was first told that masks did not make a difference in fighting the virus. Then, the public was told that masks helped but only to curb the spread of the virus to others. The most recent information as of this article is that masks help both to curb the spread but also to not catch the virus. In conditions of Knightian uncertainty, what turns out to be incorrect experimental design, ex post, can create ambiguous and contradictory conclusions early on.

Fourth, from the point of view of trying to address challenges in a Knightian uncertain environment, encouraging multiple approaches to addressing these challenges is likely to be a more reasonable strategy then prematurely settling on a single approach. There have been numerous examples of this during the pandemic. The public heard about hydroquinone from the president, the use of steroids from the medical profession, and even something as simple as turning patients on to their stomachs as a way to prevent having to intubate critically ill patients from the medical community. Ex post, it may turn out that many of these approaches were dead ends. But, ex ante, under conditions of Knightian uncertainty, it is not possible to reliably distinguish between fruitful and fruitless paths forward. It is only as the situation and the creation processes unfold that new questions are raised and new answers developed as individuals begin to grasp the context of the condition. The uncertainty is a result of being at the edge of our knowledge and where new knowledge has to be created in order to reduce the uncertainty.

Fifth, decision makers under conditions of Knightian uncertainty use cognitive biases and heuristics when optimization is impossible (Busenitz & Barney, 1997; Gigerenzer, 2008). The lack of historical conventions, knowledge, and techniques render the ability to use traditional data‐driven tools for decision‐making moot. Cognitive biases and heuristics can be used when there are little to no data. Four particularly common biases and heuristics that are often operating in an uncertain setting are the representativeness bias (i.e., the willingness to generalize from small numbers), the overconfidence bias (i.e., having a great deal of confidence in your ability to generalize from small numbers), the confirmation bias (i.e., engaging in experiments that can only confirm your hypotheses), and the persistence bias (i.e., increasing your commitment to a course of action in the face of negative feedback). These biases vary significantly from the standards of rational risk‐based decision‐making, but under conditions of Knightian uncertainty, rational risk‐based decision‐making is not possible. Thus, in order for experiments under Knightian uncertainty to be undertaken, those engaging in these experiments will often manifest these and related biases.

While Covid‐19 and other recent events seem almost surreal to many of us, Frank Knight would have recognized these situations as uncertain. He would have also known that the theories and techniques used for analysis and practice during periods of business, economic, medical, and political stability—tools consistent with discovery theory, scientific methods, rational or boundedly rational analysis, theories, and their related tools that assume the world exists and is static—would be inadequate in dealing with the challenges presented by uncertainty. But Creation Theory tells us how to proceed: Go back to the beginning, when the virus first jumped from animals to humans, when one man died under the knee of another. Go back to the beginning and understand the human actions that started the uncertainty that disrupted the world. By understanding what happened and the specific actions that started the uncertainty, individuals can shape the uncertainty and create opportunities for the betterment of humankind.

弗兰克·奈特（Frank Knight）生活在不确定的时代。他于1885年出生在伊利诺伊州的一个农场，1900年至1910年间，与家人一起搬到了科罗拉多州，科罗拉多州的父亲拥有一个果园（Knoop，2010年）。作为小型企业的所有者，骑士团很可能经历了近两年来的衰退，萧条和金融恐慌的不利影响，这在20世纪初使美国经济陷入瘫痪（Knoop，2010年）。第一次全球战争-第一次世界大战-始于1914年在欧洲（奈特29岁时），并于1917年包括美国。在这场“结束一切战争的战争”中，奈特的四个兄弟在军队中服役。 900万士兵和1300万平民丧生（麦克米伦，2013年）。1918年，当33岁的奈特（Knight）修改其论文发表时，世界经历了一场流感大流行，全世界有19至5千万人丧生。有人估计这种大流行造成的死亡人数高达1亿（Spinney，2017）.11关于奈特家族的历史信息是从1900年美国人口普查获得的，伊利诺伊州麦克莱恩县，人口时间表，劳恩代尔，p.1。6，住宅104，家庭104，温顿和茱莉亚·奈特；数字图像，先祖（http://www.ancestry.com：2020年6月16日访问）；引用了NARA缩微胶片出版物T623。1910年美国人口普查，科罗拉多州梅萨县，人口时间表，波莫纳，第3页。14B，住所297，家庭308，Winton C.和Julia A. Knight；数字图像，先祖（http://www.ancestry.com：2020年6月16日访问）；引用了NARA缩微胶片出版物T624。“美国，第一次世界大战征兵登记卡，1917-1918年”，图片，先祖（http://www.ancestry.com：2020年6月16日访问），Melvin M. Knight卡，序列号。20637年，马萨诸塞州伍斯特市克拉克大学地方草案委员会；还为Frank Hyneman Knight制作的卡片，序列号 964，第15号地方草案委员会，伊利诺伊州芝加哥。“科罗拉多州，第一次世界大战中的士兵，1917-1918年”，数据库，先祖（http://www.ancestry.com：2020年6月16日访问），Paul Edwin Knight，Mark Knight，Bruce Winton Knight的条目；理由是男人和女人的名册谁在科罗拉多第一次世界大战担任，1917年- 1918年的。美国科罗拉多州；副官长，科罗拉多州国民警卫队，1941年“美国退伍军人事务部BIRLS死亡文件，1850年至2010年，”数据库，祖先 （http://www.ancestry.com：进入2020年6月16日），参赛作品包括Paul Knight，Bruce Knight和Mark Knight。

因此，奈特（Knight）的第一篇研究集中于风险与不确定性之间的差异，尤其是不确定性对企业的影响，也许并不奇怪（Knight，1921）。对于奈特来说，当决策者事前不知道什么是最佳决策，但确实知道决策的可能结果及其可能性时，决策环境就会充满风险。当决策者事前不知道什么是最优决策并且也不知道决策的可能结果或概率时，决策环境就不确定。

科斯（1988：1）曾经观察到他的作品“被引用很多，但很少使用”。关于奈特关于不确定性的工作可能也可以这样说。从传统经济学的角度来看，不确定性的问题在于，在不确定性条件下通常不可能编写形式化模型，而在风险条件下却可以编写形式化模型（Stigler，1985）。因此，在承认存在不确定性的同时，22一些经济学家完全否定了不确定性的存在，他们认为由于决策者的思维总是以概率分布进行运作，因此决策总是在风险条件下做出的（Savage，1954）。参见Alvarez和Barney（2020），总结了这种人类决策模型的行为经济学评论。 经济学家专注于风险条件，他们可以采用自己偏爱的预测方法。

然而，在最近几年中，人们对奈特不确定性的兴趣逐渐增加，主要是在管理学者中（Alvarez＆Barney，2005； Milliken，1987； Teece＆Leih，2016）。虽然人们对奈特不确定性的兴趣涵盖了许多管理领域（Burns＆Stalker，1966 ; Lawrence＆Lorsch，1967 ; Thompson，1967），但企业家精神学者最近才接受了奈特不确定性及其对研究的意义（Alvarez＆Barney，2007年； McMullen＆Shepherd，2006年； Sarasvathy，2001年）。因此，毫不奇怪，当“发现与创造：企业家行为的替代理论”出版时，它引起了极大的兴趣。尽管许多学者写过关于不确定性及其与风险的差异的文章，但《发现与创造》论文是第一个系统地阐明这些差异及其对机会形成理论的意义的论文。

骑士式的不确定性是创造和利用企业家机会的创造理论的一个关键假设，并且重要性日益提高（Alvarez和Barney，2007； Alvarez，Barney和Anderson，2013）。与具有风险假设的企业家发现观点相反，创造论认为，企业家通常开始形成机会的过程-定义为产品或要素市场中的竞争缺陷（Alvarez＆Barney，2020年）），关于他们最终可能创造的机会的特征的信息很少或没有。也就是说，与奈特关于不确定性的定义一致，企业家在奈特不确定性的条件下开始机会形成过程，然后才可以甚至概率性地预期创造过程的结果。通过一系列行动来进行依赖于路径的迭代学习过程，企业家有时可以最终创造机会，而他们在创造机会之前就没有机会。

考虑到最近发生的事件与奈特完成其原始工作的环境极为相似，包括全球Covid-19大流行，相关经济的出现，将奈特的不确定性放在首位并集中于了解2007年论文中的创业机会的过程似乎是非常先见的。动荡和种族紧张局势的加剧与乔治·弗洛伊德（George Floyd）等人的去世有关。所有这些事件很难甚至不可能在个人的行动促使它们发生之前就预料到。当然，这就是使人类行动变得重要的原因-这些行动可以使Knightian不确定性动起来，而个人往往不知道这样做。在这种情况下，骑士不确定性的概念有什么作用（如果有的话）？

首先，在奈特主义不确定性条件下形成创业机会的一个主要特征是，一开始，未知的事物远大于已知的事物，而我们认为已知的事物可能与人类行为带来的影响无关。更改。在大流行的情况下，我们对大流行开始时了解的很少，但是随着时间的流逝，我们对病毒的传播和感染者的治疗有了更多的了解。减少不确定性的最初努力充其量只能是明智的猜测。在早期，数据通常不存在或不完整。正如我们在当前的大流行中所看到的那样，在稳定期内用于治疗已知流感类型的惯例和技术已变得毫无用处。随着我们开始更多地了解这种情况，我们对医疗保健和科学方法的了解往往无关紧要。那些希望减少不确定性的人通常不知道要问的正确问题，更不用说如何回答这些问题了。

第二，事后，很容易将骑士在不确定性早期阶段个人所采取的许多失误归因于无能。而且，确实，其中某些步骤可能是无能的结果，但它们也可能反映了决策者在解决不确定性的早期阶段所面临的基本不确定性。

第三，在Knightian不确定性过程的早期进行的实验通常会产生不完整甚至矛盾的结果。Knightian的不确定性使得难以设计最佳信息量的实验。在这种情况下，典型的情况是一个人尝试一件事，如果不起作用，他们会尝试其他事情。如果某事可行，那么他们会做更多的事情。正如梅纳德·凯恩斯勋爵（Lord Maynard Keynes）对a徒的名言所言：“当我的信息改变时，我改变了结论。先生，您要做什么？不断变化的口罩建议可能最能说明这一点。公众首先被告知，口罩在抵抗病毒方面无济于事。然后，人们被告知，口罩是有帮助的，但只能遏制病毒向他人的传播。截至本文为止的最新信息是，口罩不仅可以抑制病毒传播，还可以防止病毒感染。在奈特式不确定性的条件下，事后证明是错误的实验设计会在早期产生模棱两可的结论。

第四，从试图应对奈特式不确定环境中的挑战的观点出发，鼓励过多种应对这些挑战的方法可能是一种较合理的策略，而不是过早地采用单一方法。大流行期间有许多这样的例子。公众从总统那里听到了对苯二酚，医学界对类固醇的使用，甚至像将患者放到肚子上那样简单的方法，以防止必须从医学界向重症患者插管。事后，可能发现其中许多方法都是死胡同。但是，事前，在奈特主义不确定性的条件下，不可能可靠地区分前进的道路和失败的道路。只是随着情况和创造过程的展开，提出了新的问题，并随着个人开始掌握该情况的上下文而产生了新的答案。不确定性是处于我们知识边缘的结果，并且必须在其中创建新知识以减少不确定性。

第五，在不可能进行优化的情况下，在奈特不确定性条件下的决策者会使用认知偏见和启发式方法（Busenitz＆Barney，1997； Gigerenzer，2008））。由于缺乏历史惯例，知识和技术，因此无法使用传统的数据驱动工具进行决策。当数据很少或没有数据时，可以使用认知偏见和启发式方法。经常在不确定的环境中运行的四个特别常见的偏见和启发式方法是代表性偏见（即，愿意从少量数字进行概括），过度自信的偏见（即，对您从少量数字进行概括的能力有很大的信心） ），确认偏差（即从事只能确认您的假设的实验）和持久性偏差（即在面对负面反馈时增加对行动方针的承诺）。这些偏见与理性的基于风险的决策制定标准大相径庭，但是在Knightian不确定性的条件下，基于风险的理性决策是不可能的。因此，为了在奈特主义不确定性下进行实验，从事这些实验的人通常会表现出这些和相关的偏见。

虽然对于我们许多人来说，Covid-19和其他最近发生的事件几乎是超现实的，但弗兰克·奈特（Frank Knight）会意识到这些情况尚不确定。他还应该知道，在商业，经济，医学和政治稳定时期用于分析和实践的理论和技术是与发现理论，科学方法，理性或有限理性分析，理论及其相关工具相一致的工具。假设世界存在并且是静态的，这将不足以应对不确定性带来的挑战。但是创造论告诉我们如何进行：回到开始，当病毒第一次从动物身上跳到人类时，一个人在另一个人的膝盖下死亡。回到最开始，了解人类行为，这些行为开始了扰乱世界的不确定性。