先讲一个起名的故事。

1950 年代初，美国兰德公司有位年轻数学家，叫 Richard Bellman。他碰上的麻烦，今天写经费申请的人都不陌生：他想做的事，名字不好听。

Bellman 晚年在自传里回忆，美国国防部长 Charles Wilson 特别讨厌“研究”（research）这个词。谁在他面前说这个词，他就会涨红了脸，甚至发火。兰德公司拿的是空军的钱，空军归国防部管。“研究”都不能提，“数学”就更别提了。

所以 Bellman 得给自己的工作起个名字。要求有三条：听上去像在干正事；别让人看出是在搞数学；最好谁都挑不出毛病。

他先挑了 programming。在当时的军队里，这个词的意思是“排计划、定日程”，和写代码没关系。前面再加一个 dynamic。这个词能让人想到“分好几步、随时间推进的决策”，而且几乎没法用在坏的意思上。照他自己的说法，这是一个“连国会议员都没法反对”的名字。

它就是程序员面试必刷的“动态规划”。

说来好笑。一个本来和编程无关的名字，后来成了编程面试的头号难关。而那个“动态”，在今天最常见的动态规划题里（背包问题、最长公共子序列、编辑距离），一点随时间变化的影子都没有。你做的事情，就是把图 1 那样的一张表填满。

这个故事本身还有个 bug。Bellman 说起名是在 1950 年秋天，可 Wilson 1953 年 1 月才当上国防部长，1950 年的国防部长是马歇尔。而 Bellman 用“动态规划”作题目的第一篇论文，1952 年就发表了。《人工智能：一种现代方法》第 3 版的两位作者 Russell 和 Norvig 专门指出过这一点。

如果你觉得这只是个例外，可以把大学的专业目录翻开看看。计算机科学不太研究计算机，信息论不管信息是什么意思，统计学最早是在给国家称量幸福，形而上学一开始只是书架上的一张标签。

这些名字，大多不是起名的人犯傻。相反，它们刚起出来的时候往往挺聪明，有的还是精心设计的。问题出在别处：名字从来不是写给学科自己看的，是写给某个观众看的。观众一换，名字就不准了。

下面按“观众”分成四类，一件一件看。

旧时候乡下有个习俗，给孩子起名叫“狗剩”“铁蛋”。据说这样阎王小鬼看不上，孩子好养活。这种名字真正的读者不是孩子，也不是亲戚，而是可能来抓孩子的小鬼。

Bellman 干的是同一件事，只是方向反过来。乡下人把名字起得够难听，让小鬼嫌弃；Bellman 把名字起得够好听，让部长挑不出刺。办法相反，用意一样：名字是给那个手里有权的旁观者看的。

差不多同一时间，五角大楼里还有一个同款名字：线性规划。Dantzig 当时在美国空军工作，任务是给训练、补给、调度排出最好的计划。军方管这种计划叫 program。后来计算机出现了，programming 有了新意思：写程序。

这下就乱了。几十年里，很多学生对着“数学规划”（mathematical programming）这门课发愣：这跟编程有什么关系？答案是没关系。

到了 2010 年，国际数学规划学会干脆改名，叫“数学优化学会”。学术组织因为名字容易让人误会而主动改名，很少见。中文译成“规划”，反倒躲开了这个坑。

1955 年，McCarthy 写达特茅斯会议的提案，用了一个新词：人工智能。他后来说，起这个名字，一个原因是想和当时很火的“控制论”分开。他既不想把控制论的创始人维纳奉为师父，也不想跟他吵架。一个改变了世界的名字，有一部分是为了躲开一位前辈。

这门手艺到今天还在用。磁共振成像刚进医院的时候，“核”字容易让病人想到核弹和辐射，英文里就把 nuclear 悄悄删了，只叫 MRI。中文医学上的规范名称也是“磁共振成像”。可病人们到现在还是照样说：“去拍个核磁。”

第二类事故没有谁设计，全是时间造成的。起名的时候，名字是准的。后来词义变了，学科也变了，只有名字留在原地，像一块化石。

英文 computer 几百年来指的都是人，就是专门做计算的职员。

19 世纪末，哈佛天文台雇了一批女性“计算员”，处理天文照片上的数据。NASA 和它的前身也雇了一大批人工计算员，一直干到 1970 年代。1962 年绕地球飞行前，宇航员 John Glenn 还特意点名，让计算员 Katherine Johnson 把电子计算机算出的轨道手算核对一遍。电影《隐藏人物》讲的就是她和同事们。

后来机器接过了这份工作，也接过了这个名字。再后来，研究计算的学科又用这台机器来命名，叫“计算机科学”。

MIT 的 Hal Abelson 在经典课程 SICP 的第一课上说得很直接：“计算机科学”这个名字很糟糕。第一，它不太像科学，更像工程，甚至像魔法。第二，它也不太关乎计算机，就像几何学并不真的关乎测量工具。在他看来，数学讲的是“什么是对的”，计算机科学讲的是“怎么一步步把它做出来”。

还有一句流传更广的话：“计算机科学和计算机的关系，就像天文学和望远镜的关系。”很多人说这是 Dijkstra 讲的，其实查不到出处。能找到的最早的相近说法，出自 1986 年的一本书，作者只说这是业内人士爱说的话。连吐槽名字的名言都署错了名，也算是一种幽默。Dijkstra 本人倒确实更爱叫它 computing science。丹麦人干脆叫它 datalogi（Peter Naur 1966 年造的词），意思是“数据学”。

中文也没好到哪里去。它既叫“计算机”，又叫“电脑”。一个把它看小了，只当是算数的机器；一个又把它看大了，直接叫成了“脑”。

Abelson 拿几何打比方，是有道理的，因为几何的名字更老、也更不准。geometry 的字面意思是“丈量土地”。希罗多德记过一件事：尼罗河发大水冲走了谁家的地，国王就派人去重新量，好给这家减税。他认为希腊人的几何就是从这儿学来的。可今天的几何学家研究弯曲的空间、方程的零点，没有谁需要下地丈量。

中文的“几何”更有意思。1607 年，徐光启和利玛窦一起翻译欧几里得，书名叫《几何原本》。“几何”在古汉语里是“多少”的意思，比如曹操的“对酒当歌，人生几何”。常有人说“几何”是 geo 的音译，这个说法找不到证据。利玛窦在序里用“几何”指的就是“量”，也就是“多少”。照字面理解，几何学就是“多少学”。

经济学的事故，要从一个讽刺说起。economics 来自希腊语，意思是“持家的办法”。古希腊作家色诺芬有一本书，中译本就叫《经济论》。翻开一看，讲的是怎么打理庄园，怎么教刚过门的妻子管家。亚里士多德在《政治学》里分得更清楚：管家是正当的；为了钱而没完没了地追钱，是违背自然的。

也就是说，给经济学留下名字的祖师爷，反倒看不起赚钱。

中文“经济”的出身更高。它是“经世济民”的缩写，杜甫《上水遣怀》里有一句：“古来经济才，何事独罕有。”说的是治国安邦的本事。后来日本学者拿它翻译 economy。严复嫌它意思太宽，在《原富》里坚持译成“计学”，梁启超也试过“生计学”，最后都没争过“经济”。

然后这个词就一路往下掉：经济实惠，经济适用房，经济舱。从“经世济民”一直掉到飞机最后几排，大概是汉语里最惨的一次降级。

现代经济学研究的，既不是持家，也不是治国，更不是省钱。它研究的是：东西有限、想要的很多的时候，人怎么选，又怎么被激励。所以学经济的人最怕被问：“那你说，现在该买哪只股票？”

统计学的出身比经济学还大。18 世纪，德国学者 Achenwall 让 Statistik 这个词流行起来，意思是“关于国家的学问”，词根就是 state（国家）。18 世纪末，苏格兰人 John Sinclair 让它在英语里流行起来，下的定义很动人：调查一个国家的状况，弄清楚这里的居民享有多少幸福，将来又该怎么改进。

统计学最早，是在给一个国家称量幸福。

中文“统计”则把它请进了会计室：统而计之，就是汇总、数数。所以学统计的人在饭桌上常被这么介绍：“这孩子学统计的，做表可厉害了。”

现代统计学的核心，既不是国家，也不是数数，而是在不确定中做推断：手里只有一千个样本，凭什么对一亿人下结论？这个问题，名字里一个字也没提。

第三类事故最不容易发现。为了让人一听就懂，学科常常从日常语言里借一个词，把原来的意思掏空，再填进一个精确的定义。这种名字是写给直觉看的。它确实能让你马上有感觉，可惜常常是错的感觉。

1948 年，香农发表了信息论的开山论文。题目叫《通信的数学理论》，标题里没有“信息”两个字。论文第一页就说清楚了：消息通常是有意义的，但意义这一层，和工程问题无关。

在香农那里，信息量量的是“意外程度”，也就是不确定性。由此有个很反直觉的结论：一只猴子在键盘上乱敲，每个字母的信息量都比莎士比亚的原文高。英语很有规律，下一个字母常常猜得到；猴子的下一个字母，谁也猜不到。

后来信息论火遍了生物学、心理学、经济学，香农自己反倒不安了。1956 年他写了一篇只有一页的短文，题目叫 The Bandwagon（大意是“跟风的花车”），劝大家冷静：这个理论被吹得超过了它实际做到的。

信息论里最核心的那个量，名字来得更随便。据热力学家 Myron Tribus 转述，香农当时犹豫，是叫它“信息”还是“不确定性”。冯·诺依曼建议他叫“熵”，理由有两条：第一，统计力学里早就有这个公式，也叫这个名字；第二，更重要的是，没人真正知道熵是什么，所以辩论起来你总能占上风。

可这个故事只有 Tribus 一个来源。香农本人 1982 年接受采访时说，他很确定这事没有发生在他和冯·诺依曼之间。所以它多半靠不住。不过它能流传这么广，也许正因为大家都觉得：熵这个名字，确实像是用来让对手闭嘴的。

混沌理论的开山论文之一，是气象学家 Lorenz 1963 年发表的，题目叫《确定性非周期流》。“确定性”三个字就写在标题里。把 chaos 这个词用在这类系统上，要等到 1975 年李天岩和 Yorke 的一篇论文。

混沌系统里没有一点随机。起点定了，未来就全定了。它之所以没法长期预测，不是因为上帝在掷骰子，而是因为起点哪怕差一根头发丝，后面的结果也会差得很远，而那根头发丝，我们永远量不准。

连那个最有名的比喻，也是起名起出来的。据 Lorenz 自己后来回忆，1972 年他去做报告，没按时交题目，会议召集人就替他拟了一个：“巴西的一只蝴蝶扇动翅膀，会不会在得克萨斯州引起一场龙卷风？”而在 1963 年的另一篇论文里，Lorenz 用的比喻是一只海鸥。要是那位召集人没那么有文采，我们今天说的可能是“海鸥效应”。

概念层面的事故就更多了，随手举三个。

“正态”分布，让人以为别的分布都“不正常”。大力推广这个名字的统计学家 Karl Pearson 后来自己也承认，这个名字有个缺点：会让人以为其他分布都是“反常”的。

“回归”分析，听上去像在研究“回到原处”。其实这个词来自 Galton 1886 年对身高的研究：高个子父母的孩子往往没那么高，矮个子父母的孩子往往没那么矮，都在“向平庸回归”。今天的回归分析，早就和身高、和“回归”都没关系了。

“虚数”一点也不虚。这个名字可以追到笛卡尔，他管这类根叫“想象的”，带点看不起的意思。高斯抱怨过：要是当初把 +1、−1、√−1 叫作“正向单位”“反向单位”“侧向单位”，就不会有那么多神秘感了。今天，量子力学的薛定谔方程里堂堂正正写着一个 i。描述世界最基本的方程之一，离不开一个“虚”的数。

上面这些事故，中文读者还要再碰一遍。每个学科进到中文里，都要被重新起一次名。翻译的人面对的，是他自己那个时代的读者。

最典型的是进化论。

英文 evolution 来自拉丁语，本义是“展开”，像展开一卷书。18 世纪，这个词被胚胎学家用来说一种早就被推翻的学说：胚胎里本来就藏着一个缩小的成体，发育只是把它“展开”。也就是说，evolution 最早的意思是“按事先定好的样子展开”，和达尔文的理论正好相反。

达尔文自己显然对这个词有戒心。《物种起源》第一版从头到尾没用 evolution，他说的是“带有变异的传承”（descent with modification）。只有一个例外：全书最后一个词，evolved。evolution 要到 1872 年的第 6 版才出现在书里。

可词有自己的命运。斯宾塞把 evolution 推成了“进步”的同义词，“适者生存”这个说法也是斯宾塞发明的。1898 年，严复翻译赫胥黎的书，把 evolution 译成“天演”，“物竞天择，适者生存”一下子传遍全国。

胡适在《四十自述》里回忆，那时候很多人拿这些词给自己和孩子起名，“竞存”“天择”很常见。他的表字“适之”，就是二哥从“适者生存”里挑的“适”字，后来他干脆用“胡适”作了名字。用他的话说，他的名字是“这种风气底下的纪念品”。

一个本来就容易让人误会的词，经过一次翻译，给中国现代最重要的思想家之一起了名字。

后来“天演”输给了“进化”。“进化”是日本人造的汉语词，严复书里其实也用过，随着大批日文著作译进来，它最后压倒了“天演”。一个“进”字，把斯宾塞的“进步”观牢牢焊进了汉语。从猿到人排着队往前走的那张图，“进化成更高级的生命”这种说法，都是这个字留下的。

可现代生物学的看法是：自然选择没有预定的方向，只是适应眼前的环境。洞穴里的鱼丢了眼睛，鲸丢了后腿，也都是演化。所以越来越多的学者主张改叫“演化”。

最有意思的是数码宝贝和宝可梦。它们的“进化”，是一只个体按事先定好的路线变成另一种样子。这其实比达尔文更接近 evolution 在 18 世纪的本义。一代小孩在动画片里学到的“进化”，无意中是两百多年前的老意思。

形而上学的经历，是一出中西两边都有的荒诞剧。

先看西方。亚里士多德自己管这门学问叫“第一哲学”。按传统的说法，公元前 1 世纪，编者安德罗尼柯整理他的著作，把这批讨论“存在”本身的书排在《物理学》后面，标签就写“物理学之后的那些书”（ta meta ta physika）。后来的人只看字面，把 meta 理解成“超越”，这门学问从此就带上了神秘色彩。到今天，欧美书店里标着 Metaphysics 的那排书架上，常常摆着水晶、塔罗牌和占星书。

生于书架标签，又回到了书架标签。

再看中文。1881 年，井上哲次郎等日本学者编《哲学字汇》，从《易传》“形而上者谓之道”里取了“形而上学”这个译名，非常典雅。可在中文教科书里，它用的是恩格斯的意思，指“孤立、静止、片面地看问题”，和辩证法相对。所以一个中国学生说“我研究形而上学”，听的人第一反应可能是：这孩子看问题的方法有问题。

更巧的是，中国本土的形而上学也走了同一条路。魏晋玄学是中国哲学史上最讲思辨的一段，王弼、郭象这些人讨论的是“有”和“无”。而今天“玄学”最常见的用法是：抽卡玄学，考试玄学，“这个 bug 有点玄学”。

东西方两门最古老、专门追问“存在”的学问，到了今天，一个成了水晶货架，一个成了抽卡口头禅。

在亚里士多德那里，修辞术是一种能力：在每一种情况下，找到能用来说服人的办法。雅典公民要在议会里辩论打不打仗，要在法庭上替自己辩护，会说服人是基本功。中世纪的大学把它列为“三艺”之一，和语法、逻辑并列。

到了英语里，rhetoric 慢慢变成了贬义词。说一句话“只是 rhetoric”，意思是空话、漂亮话。到了中文里，它又缩成了“修辞手法”：比喻、拟人、排比、夸张，语文卷子上的填空题。

最讽刺的是中文“修辞”的出处。《周易》里说“修辞立其诚”，意思是说话要以真诚为本。一门名字出自“真诚”的学问，到了英语里成了“不真诚”的代名词。

代数的事故温和一些，但很好玩。algebra 来自九世纪巴格达数学家花拉子米的书名。al-jabr 的本义是“复位”：把方程一边减掉的项挪到另一边，让它归位。这个词在西班牙语里还走了另一条路：algebrista 一度指正骨的大夫，《堂吉诃德》里就这么用过。那时候受了伤去找 algebrista，你找的不是数学家，是接骨大夫。

中文的“代数”，是靠晚清李善兰和伟烈亚力合译的《代数学》（1859）流行开来的，意思是“以字代数”，用字母代替数字。这对中学代数很贴切。可到了大学的“近世代数”，讲群、环、域，学生会发现这里研究的是运算本身的结构，根本没有要“代”的“数”。

“近世代数”的“近世”，来自范德瓦尔登 1930 年那本名著 Moderne Algebra（《近世代数》）。有意思的是，作者自己从 1955 年的新版起删掉了 Moderne，书名只剩 Algebra，因为当年的“现代”早就成了标准内容。原作者都改口了，中文课表上的“近世”却一直留到今天。名字停在了它最年轻的时候。

公平起见，也说说反过来的故事：有些学科到了中文里，反而得到一个比原名更准的名字。

英文 operations research，字面意思只是“对行动的研究”，来自二战时英军研究雷达怎么部署、护航船队怎么编队。这个名字看不出它到底在做什么：在各种限制下找最好的决策。1956 年，许国志等中国学者从《史记》“运筹策帷帐之中，决胜于千里之外”里取了“运筹”两个字。英文名只告诉你“这是研究行动的”，中文名直接告诉你：这是张良干的活。

英文 field 在代数里指一种运算结构，比如有理数、实数这种能加减乘除的系统；在物理里指弥漫在空间里的作用，比如电磁场、引力场。两件事毫无关系，却共用一个词。中文干脆分开叫：代数里叫“域”，物理里叫“场”。台湾的代数课本里还有第三个名字“體”，来自德文 Körper，本义是“身体”。

法国数学家托姆的 catastrophe theory，研究的是：条件一点点平滑地变，系统为什么会突然跳到另一个状态。比如一座结构突然垮掉，或者一条狗从对峙突然扑上去。名字里的 catastrophe（灾难）太有戏剧性，上世纪 70 年代，它被媒体捧成能解释一切的理论，后来又被学界狠狠批评。中文译成“突变论”，冷静多了。

最漂亮的翻译，大概是“熵”。1865 年克劳修斯造 entropy 这个词时，取的是希腊语“转变”的意思，还特意让它长得像 energy（能量）。

1923 年 5 月，德国冷冻工程学家 Rudolf Plank（不是那位物理学家普朗克）在南京讲热力学，当翻译的物理学家胡刚复据说就在那时造了“熵”字：左边一个“火”，右边一个“商”，因为熵的变化正是热量除以温度得到的“商”。一个汉字，把定义直接写在了脸上。

可惜好名字也挡不住命运。今天中文网上最常见的“熵”，出现在讲自律的帖子里：“自律就是对抗熵增”，“人生的本质是熵增”。传说里冯·诺依曼那句玩笑（没人知道熵是什么，所以辩论时你总占上风），几十年后在中文互联网上应验了，方式谁也想不到。

最后说一个反过来的例子。“博弈”在古汉语里是两种游戏：六博（一种掷采走棋的游戏）和围棋。

孔子说：“饱食终日，无所用心，难矣哉！不有博弈者乎？为之，犹贤乎已。”意思是：整天吃饱了什么心思都不用，还不如去下下棋。

game theory 在大陆数学界的审定名，至今还是“对策论”；台湾的经济学界叫它“赛局理论”。可在大陆的经济学界和日常说法里，通行的是“博弈论”。

然后奇妙的事发生了：不是名字误导了学科，而是学科改写了名字。今天“博弈”在日常汉语里，早就不是下棋赌钱，而是策略上的较量：和甲方博弈，和房东博弈，和自己的拖延症博弈。孔子眼里聊胜于无的消遣，两千多年后成了诺贝尔经济学奖的常客。

读到这里你可能会问：这么多名字都不准，为什么不改？

因为改名这件事，本身就是一场博弈。院系、学位、期刊、教材、搜索关键词、经费申请书，全都绑在旧名字上。人人都知道“计算机科学”不准，可谁要是自己单独改名，谁就会从招生简章和搜索结果里消失。博弈论里管这叫“谢林点”：大家选它，不是因为它最好，而是因为大家都知道别人也会选它。数学规划学会那次改名值得一提，正是因为少见。

代价谁来付？每一个需要跟别人解释自己专业的人。学计算机的被叫去修电脑，学统计的被叫去做表，学经济的被问该买哪只股票，学心理学的被要求“猜猜我在想什么”。名字对外面承诺了一件事，学科做的是另一件事，中间的差价，每个学生都要在饭桌上付一遍。

名字还会影响谁出钱。斯坦福的统计学家 Rob Tibshirani 有一张流传很广的对照表，大意是：统计学叫“模型”的，机器学习叫“网络”；统计学叫“拟合”的，机器学习叫“学习”；统计学叫“测试集表现”的，机器学习叫“泛化”。最扎心的一行是：统计学里的“大额经费”是五万美元，机器学习是一百万。

数学家对这件事倒看得很开。据学生 Blumenthal 转述，希尔伯特说过：几何里的“点、线、面”，随时都可以换成“桌子、椅子、啤酒杯”，定理照样成立。在公理的世界里，名字只是个代号，内容全在定义里。

可人不是公理系统。第一次听到一个名字，脑子会自动放出它的日常意思：“进化”就是变高级，“正态”就是正常，“虚数”就是假的，“信息”就是有意义的东西。名字决定了一门学科给人的第一印象，第一印象又决定了谁会去推开那扇门。

把这些名字放在一起看，它们的历史本身就是一部演化史。注意，是演化，不是进化。

每个名字出生的时候，都是对当时环境的一次成功适应。“动态规划”适应了一位讨厌“研究”的部长，“人工智能”适应了一场地盘之争，“经济”适应了日本的译者，“天演”适应了一个急着自强的时代。后来环境变了，名字却留了下来，像人的尾椎骨：祖先靠它摇尾巴，我们留着它，最大的用处是提醒自己从哪儿来。

所以下次碰到一个名不副实的学科，不妨把它的名字当化石读，别当说明书读。它告诉你的，不是这门学问是什么，而是这门学问出生的时候，站在它面前的是谁：一位部长，一个编者，一个译者，一个对手，一个时代。

名字是学科的出生证明，
不是体检报告。

P.S. 严格说来，这篇也不是什么“调查报告”，只是一篇随笔。之所以起这个标题，是因为“一篇关于学科名字的随笔”没人会点开。你看，Bellman 的幽灵一直都在。

Let’s start with a story about a name.

In the early 1950s, a young mathematician named Richard Bellman was working at the RAND Corporation in the United States. The problem will sound familiar to anyone who writes grant proposals today: the work Bellman wanted to do had a name that sounded bad.

In an autobiography written late in life, Bellman recalled that the US Secretary of Defense, Charles Wilson, hated the word “research.” If anyone said it in front of Wilson, the Secretary would go red in the face and might even fly into a temper. RAND was funded by the Air Force, and the Air Force answered to the Department of Defense. If you couldn’t say “research,” you certainly couldn’t say “mathematics.”

So Bellman had to find a name for the work. It had three requirements: it should sound like serious business, it shouldn’t give away that this was mathematics, and ideally nobody should be able to find fault with it.

The first pick was “programming.” In the military of the day, the word meant drawing up plans and schedules; it had nothing to do with writing code. In front of it went “dynamic.” That word suggests decisions made in stages, unfolding over time, and it is almost impossible to use in a bad sense. In Bellman’s own words, it was “something not even a Congressman could object to.”

That is the “dynamic programming” every programmer drills for job interviews.

The funny part is this. A name that originally had nothing to do with programming became the toughest hurdle in programming interviews. And the “dynamic”? In today’s most common dynamic programming problems (knapsack, longest common subsequence, edit distance), nothing changes over time at all. All you do is fill in a table like the one in Figure 1.

The story itself has a bug, too. By Bellman’s account, the name was chosen in the fall of 1950, but Wilson didn’t become Secretary of Defense until January 1953; in 1950 the Secretary was George Marshall. And Bellman’s first paper with “dynamic programming” in its title came out in 1952. Russell and Norvig, the authors of “Artificial Intelligence: A Modern Approach” (third edition), point this out specifically.

If you think this is a one-off, open any university’s list of majors. Computer science doesn’t study computers much. Information theory doesn’t care what information means. Statistics began by weighing a nation’s happiness. Metaphysics started out as a label on a bookshelf.

Most of these names weren’t chosen by fools. Quite the opposite: when they were new, they were often rather clever, and some were carefully designed. The trouble lies elsewhere. A name is never written for the field itself; it is written for some audience. When the audience changes, the name stops fitting.

Below, I sort them into four kinds by audience and take them one at a time.

In the old Chinese countryside, people used to give children names like 狗剩 (“dog’s leftovers”) or 铁蛋 (“iron egg”). The idea was that the King of Hell and the imps of the underworld would find such a child beneath their notice, and the child would grow up safe. The real readers of a name like that were not the child or the relatives. They were the imps who might come to take the child away.

Bellman did the same thing, only in reverse. The villagers made a name ugly enough to put the imps off; Bellman made one pretty enough that the Secretary could find nothing to pick at. Opposite methods, same purpose: the name is for the onlooker who holds the power.

Around the same time, the Pentagon produced another name of the same make: linear programming. Dantzig was then working for the US Air Force, on a job that meant drawing up the best possible plans for training, supply and scheduling. The military called a plan like that a “program.” Then computers came along, and “programming” took on a new meaning: writing programs.

That is where the confusion began. For decades, students stared at a course called “mathematical programming” and wondered what it had to do with coding. The answer: nothing.

In 2010 the Mathematical Programming Society simply renamed itself the Mathematical Optimization Society. It is rare for an academic body to change its name because the name misleads people. The Chinese translation, 规划 (guīhuà, “planning”), dodged the trap from the start.

In 1955, writing the proposal for the Dartmouth conference, McCarthy used a new term: artificial intelligence. McCarthy said later that one reason for the name was to keep clear of cybernetics, which was all the rage then. The new name meant not having to accept Wiener, the founder of cybernetics, as a guru, and not having to argue with Wiener either. A name that changed the world was coined partly to sidestep an elder.

The trick is still in use. When magnetic resonance imaging first came into hospitals, the word “nuclear” made patients think of bombs and radiation, so English quietly dropped it and called the scan simply MRI. The standard Chinese medical name is likewise 磁共振成像 (“magnetic resonance imaging”). Yet patients still say, as they always have, “I’m off to get a 核磁” (hécí, “nuclear-magnetic”).

The second kind of accident was designed by nobody. Time did it all. When the name was given, it was accurate. Then the words changed meaning, and the fields changed too. Only the name stayed where it was, like a fossil.

For centuries, the English word “computer” meant a person: a clerk whose job was to do calculations.

In the late 19th century, the Harvard College Observatory hired a group of women as “computers” to work through the data on astronomical photographs. NASA and its predecessor also employed large numbers of human computers, right up to the 1970s. Before orbiting the Earth in 1962, the astronaut John Glenn asked for the computer Katherine Johnson by name, to check by hand the orbit the electronic computer had worked out. The film “Hidden Figures” is about Johnson and a group of fellow computers.

Then machines took over the job, and the name along with it. Later still, the field that studies computation was named after the machine: “computer science.”

In the first lecture of the classic course SICP, MIT’s Hal Abelson puts it bluntly: “computer science” is a terrible name. First, it isn’t really a science; it is more like engineering, or even magic. Second, it isn’t really about computers, just as geometry isn’t really about surveying instruments. As Abelson sees it, mathematics is about “what is true,” and computer science is about “how to do it,” step by step.

A more widely repeated line goes: “Computer science is no more about computers than astronomy is about telescopes.” It is often credited to Dijkstra, but nobody can find the source. The earliest similar remark on record is in a 1986 book, whose author says only that people in the field like to say it. Even the famous quip about a bad name has the wrong name on it, which is funny in its own way. As it happens, Dijkstra did prefer to call it “computing science.” The Danes simply call it datalogi (a word Peter Naur coined in 1966), meaning “the study of data.”

Chinese does no better. The machine is called both 计算机 (jìsuànjī, “calculating machine”) and 电脑 (diànnǎo, “electric brain”). One name sells it short, as a mere machine for sums; the other oversells it, calling it a brain outright.

Abelson had good reason to reach for geometry, because its name is even older and even less accurate. “Geometry” literally means “measuring the land.” Herodotus tells how, when the Nile flooded and carried off part of someone’s land, the king would send men to measure it again so that the owner’s tax could be cut. Herodotus believed this was where the Greeks learned geometry. But today’s geometers study curved spaces and the zeros of equations. None of them has to go out and survey a field.

The Chinese name, 几何 (jǐhé), is more interesting still. In 1607, Xu Guangqi and Matteo Ricci translated Euclid together and called the book 几何原本 (“the foundations of 几何”). In classical Chinese, 几何 means “how much,” as in Cao Cao’s line 对酒当歌，人生几何 (“sing over your wine: how much is a life?”). People often say 几何 is a phonetic rendering of “geo,” but there is no evidence for this. In the preface, Ricci uses 几何 to mean magnitude, that is, “how much.” Taken literally, geometry in Chinese is “how-much-ology.”

The accident of economics begins with an irony. “Economics” comes from Greek, where it meant “the art of running a household.” The Greek writer Xenophon wrote a book called the “Oeconomicus” (in Chinese, simply 经济论, “On Economics”). Open it and you find out how to manage a country estate, and how to teach a new bride to keep house. Aristotle, in the “Politics,” draws the line more sharply: running a household is proper; chasing money without end, for the sake of money, is against nature.

In other words, the founder who gave economics its name looked down on making money.

The Chinese word, 经济 (jīngjì, “economy”), comes from an even grander family. It is short for 经世济民, “ordering the world and relieving the people.” Du Fu has a couplet in the poem “Going Upstream, Unburdening My Heart”: “Since ancient times, why have men with a gift for 经济 alone been so rare?” By 经济, Du Fu meant the skill of governing a state and keeping it at peace. Later, Japanese scholars took the word to translate “economy.” Yan Fu thought its meaning too broad and, translating “The Wealth of Nations” as 原富, insisted on 计学 (“the study of reckoning”); Liang Qichao tried 生计学 (“the study of livelihood”). In the end neither could beat 经济.

After that, the word kept sliding downhill: 经济实惠 (cheap and good value), 经济适用房 (affordable housing), 经济舱 (economy class). From ordering the world and relieving the people all the way down to the back rows of the plane: probably the worst demotion any word in Chinese has ever suffered.

Modern economics studies neither household management nor statecraft, and certainly not how to save money. It studies how people choose, and how incentives move them, when things are scarce and wants are many. That is why anyone who studies economics dreads the question: “So, which stock should I buy right now?”

Statistics was born into an even grander family than economics. In the 18th century, the German scholar Achenwall made the word Statistik popular. It meant “knowledge about the state,” and its root is the same as “state.” At the end of the 18th century, the Scotsman John Sinclair brought the word into wide use in English, with a moving definition: an inquiry into the condition of a country, to find out how much happiness its inhabitants enjoy, and how it could be improved in future.

Statistics began as a way of weighing a nation’s happiness.

The Chinese name, 统计 (tǒngjì), moved it into the accounts office: 统而计之, gather it all up and count it, which is to say total and tally. So at family dinners, a statistics student often gets introduced like this: “This one studies statistics. A real whiz at spreadsheets.”

The heart of modern statistics is neither the state nor counting. It is inference under uncertainty: with only a thousand samples in hand, what entitles you to a conclusion about a hundred million people? The name doesn’t say a word about it.

The third kind of accident is the hardest to spot. To be understood at once, a field often borrows a word from everyday language, hollows out its old meaning and fills it with a precise definition. Names like these are written for intuition. They really do give you an instant feeling for the idea. Unfortunately, it is often the wrong feeling.

In 1948, Shannon published the paper that founded information theory. It was called “A Mathematical Theory of Communication”; the word “information” is not in the title. The first page makes things clear: messages often have meaning, but that layer of meaning is irrelevant to the engineering problem.

For Shannon, the amount of information measures surprise, that is, uncertainty. This leads to a very counterintuitive conclusion: a monkey bashing at a keyboard produces more information per letter than Shakespeare’s original text. English is highly regular, so the next letter can often be guessed. The monkey’s next letter, nobody can guess.

Later, as information theory swept through biology, psychology and economics, its founder grew uneasy. In 1956 Shannon wrote a one-page note called “The Bandwagon,” urging everyone to calm down: the theory had been hyped well beyond what it had actually achieved.

The name of the theory’s central quantity came about even more casually. According to the thermodynamicist Myron Tribus, Shannon was unsure whether to call it “information” or “uncertainty.” Von Neumann suggested “entropy,” for two reasons. First, the same formula already existed in statistical mechanics under that name. Second, and more important, nobody really knows what entropy is, so in a debate you will always have the advantage.

But Tribus is the story’s only source. In a 1982 interview, Shannon was quite sure that no such conversation with von Neumann had ever taken place. So the story probably doesn’t hold up. Still, perhaps it spread so widely because everyone feels the same way: “entropy” really does sound like a name made for shutting your opponents up.

One of the founding papers of chaos theory was published in 1963 by the meteorologist Edward Lorenz, under the title “Deterministic Nonperiodic Flow.” The word “deterministic” is right there in the title. The word “chaos” was not applied to systems like this until a 1975 paper by Tien-Yien Li and James Yorke.

There is no randomness at all in a chaotic system. Fix the starting point, and the whole future is fixed. It can’t be predicted far ahead, not because God plays dice, but because a hair’s breadth of difference at the start leads to wildly different results later on, and that hair’s breadth is something we can never measure exactly.

Even the most famous metaphor came out of someone choosing a name. As Lorenz recalled later, the title for a 1972 talk didn’t go in on time, so the session organizer made one up: “Does the flap of a butterfly’s wings in Brazil set off a tornado in Texas?” In another paper from 1963, Lorenz’s image had been a seagull. Had the organizer been less of a wordsmith, we might be talking about the “seagull effect” today.

At the level of single concepts, there are even more accidents. Here are three, picked at random.

The “normal” distribution makes people think every other distribution is not normal. Karl Pearson, the statistician who did most to spread the name, later admitted that it had a drawback: it leads people to believe that all other distributions are “abnormal.”

“Regression” analysis sounds like the study of going back to where you started. In fact the word comes from Galton’s 1886 study of height: the children of tall parents tended to be less tall, and the children of short parents less short; all were “regressing towards mediocrity.” Today’s regression analysis has long since lost any connection with height, or with going back.

“Imaginary” numbers are not imaginary at all. The name goes back to Descartes, who called such roots “imaginary,” with a touch of contempt. Gauss complained that if +1, −1 and √−1 had been called “direct,” “inverse” and “lateral” units, there would never have been so much mystery about them. Today an i sits openly in the Schrödinger equation of quantum mechanics. One of the most basic equations describing the world can’t do without an “imaginary” number.

Chinese readers run into all these accidents a second time. Every field that enters Chinese has to be named all over again, and translators write for the readers of their own day.

The classic case is the theory of evolution.

The English word “evolution” comes from Latin, where it first meant “unrolling,” as you would unroll a scroll. In the 18th century, embryologists used the word for a theory that has long since been overturned: that the embryo already holds a miniature adult, and development merely “unrolls” it. In other words, the earliest meaning of evolution was “unfolding according to a plan laid down in advance,” the exact opposite of Darwin’s theory.

Darwin was clearly wary of the word. The first edition of “On the Origin of Species” never uses “evolution”; what Darwin talks about is “descent with modification.” There is one exception: the very last word of the book, “evolved.” The word “evolution” doesn’t appear in the book until the sixth edition of 1872.

But words have fates of their own. Herbert Spencer turned “evolution” into a synonym for “progress,” and the phrase “survival of the fittest” was also Spencer’s invention. In 1898, Yan Fu translated a book by Huxley and rendered “evolution” as 天演 (tiānyǎn, “heaven’s unfolding”), and the slogan 物竞天择，适者生存 (“creatures compete, heaven selects, the fittest survive”) swept the whole country almost overnight.

In “Autobiography at Forty,” Hu Shi recalled that in those days many people took these words as names for themselves and their children; 竞存 (“compete to survive”) and 天择 (“heaven selects”) were common. The courtesy name 适之 (Shìzhī) was chosen by Hu Shi’s second brother, who took the character 适 (“fit”) from “survival of the fittest.” Later 胡适, Hu Shi, simply became the name in use. In Hu Shi’s own words, the name was “a souvenir of that fashion.”

A word that was easy to misread to begin with went through one translation and ended up naming one of the most important thinkers of modern China.

Later, 天演 lost out to 进化 (jìnhuà, “advancing change”). 进化 was a word coined in Japan from Chinese characters. Yan Fu’s book actually used it too, and as Japanese works poured into Chinese translation, it finally beat 天演. The single character 进, “advance,” welded Spencer’s idea of progress firmly into the Chinese language. The picture of apes marching in line toward a human, and phrases like “evolving into a higher form of life,” are both that one character’s legacy.

The view of modern biology, though, is that natural selection has no set direction; it only adapts to the environment at hand. Cavefish lost their eyes and whales lost their hind legs, and that is evolution too. So more and more scholars argue for calling it 演化 (yǎnhuà, “unfolding change”) instead.

Best of all are Digimon and Pokémon. Their “evolution” is one individual turning into a different form along a path fixed in advance. That is actually closer to what “evolution” meant in the 18th century than to Darwin. The “evolution” a generation of kids learned from cartoons was, without anyone intending it, a meaning more than two hundred years old.

The story of metaphysics is a farce played out in both the West and China.

Start with the West. Aristotle’s own name for this subject was “first philosophy.” According to the traditional account, in the 1st century BC the editor Andronicus, putting Aristotle’s works in order, placed the books about “being” itself after the “Physics” and labeled them “the books after the Physics” (ta meta ta physika). Later readers took the label literally, reading “meta” as “beyond,” and the subject has carried an air of mystery ever since. Today, in bookstores in Europe and America, the shelf marked Metaphysics is often stocked with crystals, tarot cards and astrology books.

Born as a shelf label, it ended up as a shelf label again.

Now the Chinese side. In 1881, Inoue Tetsujirō and other Japanese scholars compiled a “Dictionary of Philosophy,” and took the term 形而上学 from a line in the commentaries on the Book of Changes: 形而上者谓之道, “what is above form is called the Way.” A most elegant name. But in Chinese textbooks it carries Engels’s meaning: seeing things in isolation, at a standstill and from one side only, the opposite of dialectics. So when a Chinese student says “I study 形而上学,” the listener’s first thought may well be: there’s something wrong with how this kid thinks.

By an even stranger coincidence, China’s homegrown metaphysics went the same way. The 玄学 (xuánxué, “study of the profound”) of the Wei and Jin dynasties is the most speculative chapter in the history of Chinese philosophy; thinkers like Wang Bi and Guo Xiang debated “being” and “nothingness.” Yet today 玄学 is most often heard in phrases like 抽卡玄学 (the dark arts of gacha pulls), 考试玄学 (exam superstition), and “this bug is a bit 玄学” (this bug is a bit voodoo).

The two oldest disciplines, East and West, devoted to asking about “being” have ended up, one as a shelf of crystals, the other as a gacha catchphrase.

For Aristotle, rhetoric was “the faculty of observing in any given case the available means of persuasion.” Athenian citizens had to debate war and peace in the assembly and defend themselves in court, so being able to persuade was a basic skill. The medieval universities made it one of the three arts of the trivium, alongside grammar and logic.

In English, “rhetoric” gradually turned into a dirty word. To call something “just rhetoric” is to call it empty talk, fine words. In Chinese it shrank into 修辞手法 (“rhetorical devices”): metaphor, personification, parallelism, hyperbole, the fill-in-the-blank questions on a language exam.

The greatest irony is where the Chinese word 修辞 (xiūcí) comes from. The Book of Changes says 修辞立其诚: when you craft your words, build on sincerity. A discipline whose name comes from “sincerity” has become, in English, a byword for insincerity.

The accident of algebra is milder, but great fun. “Algebra” comes from the title of a book by al-Khwarizmi, a ninth-century mathematician in Baghdad. Al-jabr originally means “restoring”: moving a term subtracted on one side of an equation over to the other side, putting it back where it belongs. In Spanish the word also took another road: an algebrista once meant a bonesetter, and “Don Quixote” uses it that way. Back then, if you got hurt and went looking for an algebrista, you weren’t after a mathematician. You wanted someone to set your bones.

The Chinese name, 代数 (dàishù), caught on through 代数学 (“Algebra,” 1859), translated by Li Shanlan and Alexander Wylie in the late Qing. It means 以字代数, letting letters stand in for numbers. That fits school algebra very well. But in the university course in “modern algebra,” with its groups, rings and fields, students find that the subject is the structure of operations themselves. There are no “numbers” left for anything to “stand in for.”

The 近世 (“modern”) in the Chinese 近世代数 comes from van der Waerden’s famous 1930 book “Moderne Algebra.” Funnily enough, from the 1955 edition on, van der Waerden dropped the “Moderne,” leaving just “Algebra,” because what had been “modern” back then had long since become standard material. Even the author had moved on, yet 近世 is still there on Chinese course timetables today. The name stayed frozen at its youngest.

To be fair, here are some stories that go the other way: fields that got a more accurate name in Chinese than the one they started with.

“Operations research” literally means just “research on operations.” It comes from the British military’s work in World War II on how to deploy radar and how to arrange convoys. The name gives no hint of what the field actually does: find the best decision under all kinds of constraints. In 1956, Xu Guozhi and other Chinese scholars took the two characters 运筹 (yùnchóu, “to plan strategy”) from a line in the “Records of the Grand Historian”: 运筹策帷帐之中，决胜于千里之外, “plan strategy inside the command tent, and win battles a thousand miles away.” The English name tells you only that it is research on operations. The Chinese name tells you straight out: this is a job for a strategist like Zhang Liang.

In algebra, the English word “field” means a structure of operations, a system like the rationals or the reals where you can add, subtract, multiply and divide. In physics it means something spread through space that acts on things, like an electromagnetic or gravitational field. The two have nothing to do with each other, yet they share one word. Chinese simply gives them different names: 域 (yù, “domain”) in algebra and 场 (chǎng, “open ground”) in physics. Algebra textbooks in Taiwan have a third name, 體 (tǐ), from the German Körper, which originally means “body.”

René Thom, the French mathematician, founded catastrophe theory, which asks why a system can suddenly jump to a different state while its conditions change smoothly, little by little. Think of a structure that suddenly collapses, or a dog in a standoff that suddenly lunges. The word “catastrophe” in the name is so dramatic that in the 1970s the media hailed it as a theory that could explain everything, and then scholars came down hard on it. The Chinese name, 突变论 (tūbiànlùn, “theory of sudden change”), is much calmer.

The most beautiful translation is probably 熵 (shāng), for entropy. Clausius coined the word “entropy” in 1865, taking it from the Greek for “transformation” and deliberately making it look like “energy.”

In May 1923, the German refrigeration engineer Rudolf Plank (not Planck the physicist) gave lectures on thermodynamics in Nanjing. Plank’s interpreter, the physicist Hu Gangfu, is said to have coined the character 熵 right then: 火 (fire) on the left and 商 (quotient) on the right, because a change in entropy is exactly the quotient you get by dividing heat by temperature. A single character, with its definition written on its face.

Sadly, even a good name can’t escape its fate. On the Chinese internet today, 熵 turns up most often in posts about self-discipline: “Self-discipline means fighting entropy,” “The essence of life is rising entropy.” The joke in the von Neumann legend (nobody knows what entropy is, so you will always win the argument) came true decades later on the Chinese internet, in a way nobody could have foreseen.

One last example, going the other way. In classical Chinese, 博弈 (bóyì) meant two games: liubo (a game of throwing sticks and moving pieces) and weiqi, the game of Go.

Confucius said: “To eat your fill all day long and give your mind to nothing at all, that is hard indeed! Are there not the games of bo and yi? Even playing those is better than doing nothing.” In other words: rather than spend the whole day well fed and not using your mind, you might as well play a board game.

For game theory, the approved term in mainland Chinese mathematics is still 对策论 (duìcèlùn, “theory of countermoves”); economists in Taiwan call it 赛局理论 (“theory of contests”). But among mainland economists, and in everyday speech, the usual name is 博弈论 (bóyìlùn).

Then something wonderful happened. Instead of the name misleading the field, the field rewrote the name. In everyday Chinese today, 博弈 long ago stopped meaning board games and gambling; it means a contest of strategy. You 博弈 with your client, with your landlord, with your own procrastination. The pastime Confucius rated as better than nothing became, more than two thousand years later, a regular guest at the Nobel Prize in economics.

By now you may be asking: if so many names are wrong, why not change them?

Because changing a name is itself a game, in the game theorist’s sense. Departments, degrees, journals, textbooks, search keywords and grant applications are all tied to the old name. Everyone knows “computer science” is inaccurate, but whoever changes the name on their own vanishes from the admissions brochures and the search results. Game theory calls this a Schelling point: people choose it not because it is best, but because everyone knows everyone else will choose it too. The Mathematical Programming Society’s renaming is worth mentioning precisely because it is so rare.

Who pays the price? Everyone who has to explain their major to other people. Computer science students get asked to fix computers, statistics students get asked to make spreadsheets, economics students get asked which stock to buy, and psychology students get asked to “guess what I’m thinking.” The name promises the world one thing, the field does another, and every student pays the difference at a dinner table, at least once.

Names also affect who gets the money. The Stanford statistician Rob Tibshirani has a widely shared glossary that goes roughly like this: what statistics calls a “model,” machine learning calls a “network”; what statistics calls “fitting,” machine learning calls “learning”; what statistics calls “test set performance,” machine learning calls “generalization.” The line that stings most: a “large grant” in statistics is $50,000; in machine learning it is $1,000,000.

Mathematicians take a relaxed view of all this. Hilbert once said, as Blumenthal (a former student) tells it, that the “points, lines and planes” of geometry could be replaced at any time by “tables, chairs and beer mugs,” and the theorems would still hold. In the world of axioms, a name is just a placeholder; everything is in the definitions.

But people are not axiom systems. The first time we hear a name, our brains automatically play back its everyday meaning: “evolution” means getting more advanced, “normal” means ordinary, “imaginary” means fake, “information” means something meaningful. A name decides a field’s first impression, and the first impression decides who will push open its door.

Put all these names side by side, and their history is itself a history of evolution. Evolution as 演化, mind you, not 进化.

At birth, each name was a successful adaptation to its environment. “Dynamic programming” adapted to a Secretary who hated “research,” “artificial intelligence” to a turf war, 经济 to Japanese translators, 天演 to an era in a hurry to make the country strong. Then the environment changed, but the names stayed, like the human tailbone: our ancestors used it to wag their tails, and we keep it mainly as a reminder of where we came from.

So the next time you meet a field whose name doesn’t fit, try reading the name as a fossil, not as an instruction manual. What it tells you is not what the field is, but who stood before it when it was born: a Secretary of Defense, an editor, a translator, a rival, an era.

A name is a field’s birth certificate,
not its medical report.

P.S. Strictly speaking, this isn’t an “incident report” at all, just an essay. I gave it that title because nobody would click on “An Essay about the Names of Academic Fields.” See? Bellman’s ghost is still with us.