集体心智:社会网络拓扑塑造集体认知
The following article is from 集智俱乐部 Author Ida Momennejad
人类是社会动物,人类的认知也是通过在社会网络中学习和记忆而形成。人类的社会网络与其他物种不同,有着复杂多样的成分和拓扑,这样的社会环境塑造了人类特有的集体认知。这篇综述回顾了支持“社会网络拓扑塑造集体认知及行为”这一假说的相关研究。文章探讨了社会网络拓扑如何影响人类集体记忆、集体信念和行为、文化积累及集体智能,指出人类学习社会、空间与非社会拓扑的潜在的共同机制,并认为当前人工智能可以与脑神经网络及人类社会网络中获得的洞察结合,既加深对人类集体认知的理解,同时促进集体机器智能的发展。本综述回顾的各项研究使用了在图论、统计学、数学、物理学和神经科学中开发的各种跨学科方法。该综述收录于《英国皇家学会哲学通报B》(The Philosophical Transactions of the Royal Society B)2022年1月发布的“集体知识和文化积累在动物、人类及机器中的涌现”主题特刊中。
研究领域:网络拓扑,社会网络,认知神经科学,人工智能
Ida Momenneja | 作者
许逸聪 | 译者
陈斯信 | 审校
邓一雪 | 编辑
论文题目:
Collective minds: social network topology shapes collective cognition
论文链接:
https://royalsocietypublishing.org/doi/full/10.1098/rstb.2020.0315
目录
摘要
引言
网络拓扑协调集体记忆
网络拓扑塑造信念和规范
网络拓扑塑造集体智能
社会网络拓扑塑造神经反应
存在引导社会及非社会拓扑的普适机制?
网络拓扑应用于集体机器智能
结论:人类与机器的集体认知拓扑
目录
摘要
引言
网络拓扑协调集体记忆
网络拓扑塑造信念和规范
网络拓扑塑造集体智能
社会网络拓扑塑造神经反应
存在引导社会及非社会拓扑的普适机制?
网络拓扑应用于集体机器智能
结论:人类与机器的集体认知拓扑
摘要
摘要
图片来源:https://mp.weixin.qq.com/s/LwqdroN3Gd22abGp3ZpvLQ
1. 引言
1. 引言
图1. 网络拓扑的入门知识。社会及非社会网络可以通过图进行分析。(a)展示了一个集体记忆研究 [2] 的网络拓扑示意图。节点(Node),即社群中的个体,用人形表示。边(Edge),即两个节点(个体)之间的直接联系,用线表示。集群(cluster),包括桥接关系(bridge ties,节点间没有公共连接)和集群关系(cluster ties,的节点之间则存在多条公共关系)也在图中标出。节点的度(degree),即节点的关系数,用绿色标记。不同图结构的标准参数包括:随机性(b)、聚类关系(a,c)、网络直径(最大路径长度)和平均路径长度。著名的图拓扑包括:网格结构图、随机环状图和集群社区结构图(c)。
图片来源:https://mp.weixin.qq.com/s/LwqdroN3Gd22abGp3ZpvLQ
2. 网络拓扑协调集体记忆
2. 网络拓扑协调集体记忆
图2. 研究网络拓扑对集体记忆的作用。(a)在实验室形成的交流网络中,度量集体记忆的实验设计 [1,34]。在受试者不知情的情况下,研究者给实验的每个环节都分配了一个10人构成的拓扑。有三个阶段:个体学习并回忆测试阶段、会话阶段、以及会话后个体回忆测试阶段。(b)计算记忆相似程度(随会话而变)的方程。数字代表网络中的成员,例如P6代表6号受试者。
图片来源:https://mp.weixin.qq.com/s/LwqdroN3Gd22abGp3ZpvLQ
通过上述实验范式(图2),作者研究了网络的结构或拓扑对会话后受试者记忆融合程度(这是集体记忆融合的一个衡量标准[1])的影响。数据是从10名受试者构成的网络中收集的,在该网络中,每名受试者都进行了3次会话,这些会话按照集群或非集群网络拓扑有序进行。作者测量了8个集群的10人网络和8个非集群的10人网络(图2)。 该研究的假说是,这些成对受试者的记忆会根据其分离程度进行协调,直接对话的受试者的协调程度最高,而会话距离最远的受试者的协调程度最低(图 3)。该假设在协调的行为学测量中得到了证实(图 3)。成对的结果还表明,大型网络的拓扑会对集体记忆产生影响。与非集群网络相比,集群拓扑的网络具有更大的网络直径,即网络最远成员间具有更长的路径。结果表明,具有较小直径的网络(非集群拓扑)会比具有集群图结构或拓扑的网络更容易发生记忆融合。 Momennejad及其合作者在另一项实验 [2] 中发现,在一个固定的交流网络拓扑中,交流随时间逐渐展开的顺序,决定了记忆的协调程度。也就是说,他们发现,具有弱或桥接关系的个体间先进行信息交换,会加剧集体记忆的融合,这与Granovetter对于弱关系强度的观点相符 [37]。这些个体有直接的联系,但与其他个体没有任何共同的联系。(要直观地了解他们的网络状态,可以用一个类比:他们是朋友,有交流,但他们的朋友不是朋友,也不互相交流。)
图3. 网络拓扑对集体记忆的协调作用。该图显示了图2中描述的集群拓扑条件及非集群拓扑条件所对应的记忆相似度假设矩阵。数字代表网络中的成员,例如P6代表6号受试者。相似度范围从0(自身;深蓝色)到5(最大分离度;深红色)。行为学结果表明,平均而言,非集群网络(b)中的会话后记忆协调度比集群网络(b)更高。这一发现可以用记忆相似度假设矩阵来解释——记忆的协调程度取决于个体的分离程度。集群网络拓扑具有较大的分离度(即较长的节点间距或网络直径),会导致更低的融合性。个体间的协调程度取决于其在社会网络中彼此的分离程度(c,d)。
图片来源:https://mp.weixin.qq.com/s/LwqdroN3Gd22abGp3ZpvLQ
上述研究表明,交流网络的拓扑和时间顺序,可以决定网络中个体之间的记忆协调度,即使在他们无直接交互的情况下。在最近的一项计算研究 [36] 中,我们将多主体或基于主体的模拟,与记忆心理学中的提取诱发遗忘模型相结合 [38,39],并成功模拟了这些有关集体记忆的行为学发现。总的来说,这些行为学和计算学研究提供了一种定量方法,可以根据网络拓扑及记忆遗忘原理,测量在集体水平或介观尺度现象的涌现。 这些最新进展,为研究人类认知及行为在微观、介观和宏观尺度之间联系的理论及实验方法开辟了道路。集体实验的图论方法,以及基于实验的带参数的机构模拟[40],可以帮助理解、预测和比较人类网络的不同拓扑的行为。
3. 网络拓扑塑造信念和规范
3. 网络拓扑塑造信念和规范
4. 网络拓扑塑造集体智能
4. 网络拓扑塑造集体智能
5. 社会网络拓扑塑造神经反应
5. 社会网络拓扑塑造神经反应
6. 存在引导社会及非社会拓扑的
普适机制?
6. 存在引导社会及非社会拓扑的
普适机制?
7. 网络拓扑应用于集体机器智能
7. 网络拓扑应用于集体机器智能
8. 结论:人类与机器的集体认知拓扑
8. 结论:人类与机器的集体认知拓扑
参考文献
[1] Coman A, Momennejad I, Drach RD, Geana A. 2016Mnemonic convergence in social networks: the emergent properties of cognition at a collective level. Proc. Natl Acad. Sci. USA 113, 8171-8176.
[2] Momennejad I, Duker A, Coman A. 2019Bridge ties bind collective memories. Nat. Commun. 10, 1578.
[3] Centola D. 2010The spread of behavior in an online social network experiment. Science 329, 1194-1197.
[4] Centola D. 2011An experimental study of homophily in the adoption of health behavior. Science 334, 1269-1272.
[5] Vlasceanu M, Coman A. 2020Network structure impacts the synchronization of collective beliefs. PsyArXiv.
[6] Derex M, Boyd R. 2016Partial connectivity increases cultural accumulation within groups. Proc. Natl Acad. Sci. USA 113, 2982-2987.
[7] Parkinson C, Kleinbaum AM, Wheatley T. 2017Spontaneous neural encoding of social network position. Nat. Hum. Behav. 1, 1-7.
[8] Momennejad I. 2020Learning structures: predictive representations, replay, and generalization. Curr. Opin. Behav. Sci. 32, 155-166.
[9] Perez-Liebana D, Hofmann K, Mohanty SP, Kuno NS, Kramer A, Devlin S, Gaina RD, Ionita D.2019The multi-agent reinforcement learning in MalmÖ (MARLÖ) competition. ArXiv, abs/1901.08129.
[10] Jaques N, Lazaridou A, Hughes E.2019Social influence as intrinsic motivation for multi-agent deep reinforcement learning. In Proc. 36th Int. Conf. on Machine Learning (ICML), Long Beach, CA, PMLR 97.
[11] Vinyals Oet al.. 2019Grandmaster level in StarCraft II using multi-agent reinforcement learning. Nature 575, 350-354.
[12] Bratman ME. 1993Shared intention. Ethics 104, 97-113.
[13] Bratman ME. 2013Shared agency: a planning theory of acting together. Oxford, UK: Oxford University Press.
[14] Bratman ME. 2015Précis of shared agency: a planning theory of acting together. J. Soc. Ontol. 1, 1-5.
[15] Davidson D. 1992The second person. Midwest Stud. Phil. 17, 255-267.
[16] Tuomela R. 1995The importance of us: a philosophical study of basic social notions. Palo Alto, CA: Stanford University Press.
[17] Gilbert M. 1987Modelling collective belief. Synthese 73, 185-204.
[18] Gilbert M. 1989On social facts. Oxford, UK: Routledge.
[19] Boyd R, Richerson PJ. 2009Culture and the evolution of human cooperation. Phil. Trans. R. Soc. B 364, 3281-3288.
[20] Tomasello M, Melis AP, Tennie C, Wyman E, Herrmann E. 2012The interdependence hypothesis. Curr. Anthropol. 53, 673-692.
[21] Tomasello M. 2009The cultural origins of human cognition. Cambridge, MA: Harvard University Press. Crossref.
[22] Tomasello M, Carpenter M, Call J, Behne T, Moll H. 2005Understanding and sharing intentions: the origins of cultural cognition. Behav. Brain Sci. 28, 675-691; discussion 691–735.
[23] Zahavi D, Satne G. 2015Varieties of shared intentionality: Tomasello and classical phenomenology. In Beyond the analytic-continental divide (eds Bell JA, Cutrofello A, Livingstone PM), pp. 313-334. Oxford, UK: Routledge.
[24] Tomasello M, Carpenter M. 2007Shared intentionality. Dev. Sci. 10, 121-125.
[25] O'Madagain C, Tomasello M. 2021Shared intentionality, reason-giving and the evolution of human culture. Phil. Trans. R. Soc. B 377, 20200320.
[26] Peysakhovich A, Nowak MA, Rand DG. 2014Humans display a ‘cooperative phenotype’ that is domain general and temporally stable. Nat. Commun. 5, 4939.
[27] Gray K, Rand DG, Ert E, Lewis K, Hershman S, Norton MI. 2014The emergence of ‘Us and Them’ in 80 lines of code: modeling group genesis in homogeneous populations. Psychol. Sci. 25, 982-990.
[28] Bahrami B, Olsen K, Latham PE, Roepstorff A, Rees G, Frith CD.2010Optimally interacting minds. Science 329, 1081-1085.
[29] Barabási AL. 2003Linked: the new science of networks. Am. J. Phys. 71, 409-410.
[30] Watts DJ, Strogatz SH. 1998Collective dynamics of ‘small-world’ networks. Nature 393, 440-442.
[31] David E, Jon K. 2010Networks, crowds, and markets: reasoning about a highly connected world. New York, NY: Cambridge University Press.
[32] Barabási AL. 2013Network science. Phil. Trans. A Math. Phys. Eng. Sci. 371, 20120375.
[33] Watts DJ. 2004Six degrees: the science of a connected Age. New York, NY: W. W. Norton & Company.
[34] Hirst W, Yamashiro JK, Coman A. 2018Collective memory from a psychological perspective. Trends Cogn. Sci. 22, 438-451.
[35] Olick JK, Vinitzky-Seroussi V, Levy D. 2011The collective memory reader. New York, NY: Oxford University Press.
[36] Sievers B, Momennejad I. 2019SAMPL: the spreading activation and memory plasticity model. bioRxiv, 778563.
[37] Granovetter MS. 1973The strength of weak ties. Am. J. Sociol. 78, 1360-1380.
[38] Anderson MC, McCulloch KC. 1999Integration as a general boundary condition on retrieval-induced forgetting. J. Exp. Psychol. Learn. Mem. Cogn. 25, 608-629.
[39] Norman KA, Newman EL, Detre G. 2007A neural network model of retrieval-induced forgetting. Psychol. Rev. 114, 887-953.
[40] Momennejad I, Sinclair S, Cikara M. 2019Computational justice: simulating structural bias and interventions. bioRxiv.
[41] Schacter DL, Scarry E. 2001Memory, brain, and belief. Cambridge, MA: Harvard University Press.
[42] Friedkin NE, Proskurnikov AV, Tempo R, Parsegov SE. 2016Network science on belief system dynamics under logic constraints. Science 354, 321-326.
[43] Vlasceanu M, Morais MJ, Duker A, Coman A. 2020The synchronization of collective beliefs: from dyadic interactions to network convergence. J. Exp. Psychol. Appl. 26, 453-464.
[44] Banerjee A, Chandrasekhar AG, Duflo E, Jackson MO. 2013The diffusion of microfinance. Science 341, 1236498.
[45] Paluck EL, Shepherd H, Aronow PM. 2016Changing climates of conflict: a social network experiment in 56 schools. Proc. Natl Acad. Sci. USA 113, 566-571.
[46] Prentice D, Paluck EL. 2020Engineering social change using social norms: lessons from the study of collective action. Curr. Opin. Psychol. 35, 138-142.
[47] Wu SJ, Paluck EL. 2020Participatory practices at work change attitudes and behavior toward societal authority and justice. Nat. Commun. 11, 2633.
[48] Wu SJ, Paluck EL. 2021Designing nudges for the context: golden coin decals nudge workplace behavior in China. Organ. Behav. Hum. Decis. Process 163, 43-50.
[49] Wu SJ, Paluck EL. 2021Having a voice in your group: increasing productivity through group participation. psyarxiv.com.
[50] Woolley AW, Chabris CF, Pentland A, Hashmi N, Malone TW. 2010Evidence for a collective intelligence factor in the performance of human groups. Science 330, 686-688.
[51] Woolley AW, Aggarwal I.2015Collective intelligence and group performance. Curr. Dir. Psychol. Sci. 24, 420-424.
[52] Megías-Robles A, Gutiérrez-Cobo MJ, Cabello R, Gómez-Leal R, Baron-Cohen S, Fernández-Berrocal P. 2020The ‘Reading the mind in the Eyes' test and emotional intelligence. R. Soc. Open Sci. 7, 201305.
[53] Mason W, Watts DJ. 2012Collaborative learning in networks. Proc. Natl Acad. Sci. USA 109, 764-769.
[54] Mao A, Mason W, Suri S, Watts DJ. 2016An experimental study of team size and performance on a complex task. PLoS ONE 11, e0153048.
[55] Dunbar RIM, Shultz S. 2007Evolution in the social brain. Science 317, 1344-1347.
[56] Parkinson C, Kleinbaum AM, Wheatley T. 2018Similar neural responses predict friendship. Nat. Commun. 9, 332.
[57] Sievers B, Welker C, Hasson U, Kleinbaum AM, Wheatley T. 2020How consensus-building conversation changes our minds and aligns our brains. PsyArXiv.
[58] Pérez-Barbería FJ, Shultz S, Dunbar RIM. 2007Evidence for coevolution of sociality and relative brain size in three orders of mammals. Evolution 61, 2811-2821.
[59] Kirby S, Tamariz M.2021Cumulative cultural evolution, population structure, and the origin of combinatoriality in human language. Phil. Trans. R. Soc. B 377, 20200319.
[60] Romano V, Lozano S, Lopez de Pablo J. 2021Reconstructing social networks of Late Glacial and Holocene hunter-gatherers to understand cultural evolution. Phil. Trans. R. Soc. B 377, 20200318.
[61] Migliano A, Vinicius L. 2021The origins of human cumulative culture: from the foraging niche to collective intelligence. Phil. Trans. R. Soc. B 377, 20200317.
[62] Garland EC, Garrigue C, Noad MJ. 2021When does cultural evolution become cumulative culture? A case study of humpback whale song. Phil. Trans. R. Soc. B 377, 20200313.
[63] Gruber T, Chimento M, Aplin LM, Biro D. 2021Efficiency fosters cumulative culture across species. Phil. Trans. R. Soc. B 377, 20200308.
[64] Whiten A, Harrison RA, McGuigan N, Vale GL, Watson SK. 2021Collective knowledge and the dynamics of culture in chimpanzees. Phil. Trans. R. Soc. B 377, 20200321.
[65] Wild S, Chimento M, McMahon K, Farine DR, Shelden BC, Aplin LM. 2021Complex foraging behaviours in wild birds emerge from social learning and recombination of components. Phil. Trans. R. Soc. B 377, 20200307.
[66] Williams H, Lachlan R. 2021Evidence for cumulative cultural evolution in bird song. Phil. Trans. R. Soc. B 377, 20200322.
[67] Schapiro AC, Rogers TT, Cordova NI, Turk-Browne NB, Botvinick MM. 2013Neural representations of events arise from temporal community structure. Nat. Neurosci. 16, 486-492.
[68] Momennejad I, Russek EM, Cheong JH, Botvinick MM, Daw ND, Gershman SJ. 2017The successor representation in human reinforcement learning. Nat. Hum. Behav. 1, 680-692.
[69] Garvert MM, Dolan RJ, Behrens TE. 2017A map of abstract relational knowledge in the human hippocampal-entorhinal cortex. Elife 6, e17086.
[70] Gershman SJ, Cikara M. 2020Social-structure learning. Curr. Dir. Psychol. Sci. 29, 460-466.
[71] Russek EM, Momennejad I, Botvinick MM, Gershman SJ, Daw ND. 2017Predictive representations can link model-based reinforcement learning to model-free mechanisms. PLoS Comput. Biol. 13, e1005768.
[72] Schapiro AC, Turk-Browne NB, Botvinick MM, Norman KA. 2017Complementary learning systems within the hippocampus: a neural network modelling approach to reconciling episodic memory with statistical learning. Phil. Trans. R. Soc. B 372, 20160049.
[73] Saffran JR, Aslin RN, Newport EL. 1996Statistical learning by 8-month-old infants. Science 274, 1926-1928.
[74] Fiser J, Aslin RN. 2002Statistical learning of higher-order temporal structure from visual shape sequences. J. Exp. Psychol. Learn. Mem. Cogn. 28, 458-467.
[75] Romberg AR, Saffran JR. 2010Statistical learning and language acquisition. Wiley Interdiscip. Rev. Cogn. Sci. 1, 906-914.
[76] Friederici AD. 2005Neurophysiological markers of early language acquisition: from syllables to sentences. Trends Cogn. Sci. 9, 481-488.
[77] Tompson SH, Kahn AE, Falk EB, Vettel JM, Bassett DS. 2019Individual differences in learning social and nonsocial network structures. J. Exp. Psychol. Learn. Mem. Cogn. 45, 253-271.
[78] Karuza EA, Thompson-Schill SL, Bassett DS. 2016Local patterns to global architectures: influences of network topology on human learning. Trends Cogn. Sci. 20, 629-640.
[79] Karuza EA, Kahn AE, Bassett DS. 2019Human sensitivity to community structure is robust to topological variation. Complexity 2019, 8379321.
[80] Lynn CW, Bassett DS. 2020How humans learn and represent networks. Proc. Natl Acad. Sci. USA 117, 29 407-29 415.
[81] Stachenfeld KL, Botvinick MM, Gershman SJ. 2017The hippocampus as a predictive map. Nat. Neurosci. 20, 1643-1653.
[82] Chen J, Hasson U, Honey CJ. 2015Processing timescales as an organizing principle for primate cortex. Neuron 88, 244-246.
[83] Brunec IK, Momennejad I.2019Predictive representations in hippocampal and prefrontal hierarchies. BioRxiv.
[84] Momennejad I, Otto AR, Daw ND, Norman KA.2018Offline replay supports planning in human reinforcement learning. Elife 7, e32548.
[85] Momennejad I, Haynes JD.2012Human anterior prefrontal cortex encodes the ‘what’ and ‘when'of future intentions. Neuroimage 61, 139-148.
[86] Momennejad I, Haynes JD. 2013Encoding of prospective tasks in the human prefrontal cortex under varying task loads. J. Neurosci. 33, 17 342-17 349.
[87] Burgess PW, Dumontheil I, Gilbert SJ, Okuda J, Schölvinck ML, Simons JS.2008On the role of rostral prefrontal cortex (area 10) in prospective memory. In Prospective memory: cognitive, neuroscience, developmental, and applied perspectives, (ed. Kliegel M), pp. 235-260. New York, NY: Taylor & Francis Group/Lawrence Erlbaum Associates.
[88] Bunge SA, Wendelken C, Badre D, Wagner AD. 2005Analogical reasoning and prefrontal cortex: evidence for separable retrieval and integration mechanisms. Cereb. Cortex 15, 239-249.
[89] Gilbert SJ, Burgess PW. 2008Social and nonsocial functions of rostral prefrontal cortex: implications for education. Mind Brain Educ. 2, 148-156.
[90] Roca M, Torralva T, Gleichgerrcht E, Woolgar A, Thompson R, Duncan J, Manes F. 2011The role of Area 10 (BA10) in human multitasking and in social cognition: a lesion study. Neuropsychologia 49, 3525-3531.
[91] Gilbert SJ, Williamson IDM, Dumontheil I, Simons JS, Frith CD, Burgess PW. 2007Distinct regions of medial rostral prefrontal cortex supporting social and nonsocial functions. Soc. Cogn. Affect. Neurosci. 2, 217-226.
[92] Adamic LA, Lukose RM, Puniyani AR, Huberman BA. 2001Search in power-law networks. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 64, 046135.
[93] Allison F, Luger E, Hofmann K. 2018How players speak to an intelligent game character using natural language messages. Trans. Digit. Games Res. Assoc. 4, 1-49.
[94] Dafoe A, Bachrach Y, Hadfield G, Horvitz E, Larson K, Graepel T. 2021Cooperative AI: machines must learn to find common ground. Nature 593, 33-36.
[95] Winfield A, Blackmore S. 2021Experiments in artificial culture: from noisy imitation to storytelling robots. Phil. Trans. R. Soc. B 377, 20200323.
[96] Bredeche N, Fontbonne N. 2021A gentle introduction to social learning in swarm robotics. Phil. Trans. R. Soc. B 377, 20200309.
[97] Hart E, Le Goff L. 2021Artificial evolution of robot bodies and control: on the interaction between evolution, learning and culture. Phil. Trans. R. Soc. B 377, 20210117.
[98] Devlin S, Georgescu R, Momennejad I, Rzepecki J, Zuniga E, Costello G, Leroy G, Shaw A, Hofmann K.2021Navigation Turing Test (NTT): learning to evaluate human-like navigation. arXiv.
[99] Arulkumaran K, Cully A, Togelius J.2019AlphaStar: an evolutionary computation perspective. In Proc. of the Genetic and Evolutionary Computation Conf. Companion, pp. 314-315. New York, NY: Association for Computing Machinery.
[100] Xie T, Langford J, Mineiro P, Momennejad I.2021Interaction-grounded learning. arXiv.
[101] Zintgraf L, Devlin S, Ciosek K, Whiteson S, Hofmann K.2021Deep interactive Bayesian reinforcement learning via meta-learning. arXiv.
[102] Grover A, Al-Shedivat M, Gupta JK, Burda Y, Edwards H. 2018Evaluating generalization in multiagent systems using agent-interaction graphs. In Proc. of the 17th International Conf. on Autonomous Agents and MultiAgent Systems, pp. 1944-1946. Richland, SC: International Foundation for Autonomous Agents and Multiagent Systems.
[103] Battaglia PWet al.. 2018Relational inductive biases, deep learning, and graph networks. arXiv, 1806.01261.
[104] Son JY, Bhandari A, FeldmanHall O. 2021Cognitive maps of social features enable flexible inference in social networks. Proc. Natl Acad. Sci. USA 118, e2021699118.
[105] Perra N, Fortunato S. 2008Spectral centrality measures in complex networks. Phys. Rev. E 78, 036107.
[106] Lancichinetti A, Kivelä M, Saramäki J, Fortunato S. 2010Characterizing the community structure of complex networks. PLoS ONE 5, e11976.
(参考文献可上下滑动查看)往期回顾:
4. 数学证明到底是什么?
编辑 / 姜天海
审核 / 范 杰