Collective Intelligence: 240 Years of Research — From Condorcet to AI

A cikin 1794, wani mai lissafi na Faransa ya mutu a cikin wani dakin kurkuku na juyin juya hali, shekaru tara bayan ya buga wani ka'ida wanda kusan babu wanda ya karanta. A cikin 1906, wani mai kididdiga na Victorian mai shekaru 84 ya yi yawo a cikin wani baje kolin dabbobi yana tattara katunan rajistar da aka yi amfani da su. A cikin 2026, masu bincike a MIT suna gudanar da dubban tattaunawa tsakanin wakilan AI don koyon yadda na'urori za su shiga cikin kungiyoyin mutane. Waɗannan abubuwan guda uku suna cikin labarin bincike guda ɗaya, mai ci gaba — nazarin hankalin jama'a, ikon kungiyoyi suyi aiki a hanyoyi da suka bayyana a matsayin masu hankali. Abin da ke biyo baya shine wannan labarin, zamanin bayan zamanin, tare da ainihin abubuwan da aka gano da ainihin faɗa.

1785: Kwatancen Enlightenment — Ka'idar Condorcet

Marie Jean Antoine Nicolas de Caritat, Marquis de Condorcet, shine irin mutum na Enlightenment wanda ke sa masu ilimi na zamani suyi kama da masu iyakance: wani mai lissafi da aka zaba zuwa Académie des sciences a cikin shekaru ashirin, wani mai goyon bayan ilimi na jama'a, hakkin mata, da kuma kawar da bauta. A cikin 1785 ya buga Essai sur l’application de l’analyse à la probabilité des décisions rendues à la pluralité des voix — wani rubutu akan amfani da ka'idar yiwuwar a cikin kada kuri'a. A ciki akwai sakamakon da yanzu muke kira Ka'idar Juri ta Condorcet.

Ka'idar ta ce wani abu mai ban mamaki. Dauki tambaya ta eh/aa tare da amsar da ta dace. Ka yi tunanin kowanne mai kada kuri'a yana daidai da yiwuwar da ta fi kadan fiye da jujjuyawar kudi — misali 55% — kuma ka yi tunanin masu kada kuri'a suna hukunta zaman kansu daga juna. Sa'an nan yayin da kungiyar ke girma, yiwuwar cewa mafi yawa suna daidai yana tashi zuwa tabbaci. Hukuncin mutum mai kyau, idan aka tara da kyau, yana zama kusan cikakken hukunci na jama'a. Ka'idar tana da wani ɗan uwan duhu, ma: idan kowanne mai kada kuri'a yana da kadan mummuna fiye da jujjuyawar kudi, mafi yawa yana zama kusan tabbatacce ba daidai ba yayin da kungiyar ke girma. Tarawa yana ƙara duk wani ƙwarewa — ko rashin ƙwarewa — da ka ba shi. Muna ba da cikakken labarin ka'idar, tsammanin ta, da amfani na zamani a cikin rubutunmu na haɗin gwiwa akan Ka'idar Juri ta Condorcet.

Wannan rubutu na biyu yana dauke da wani bincike na biyu wanda zai dauki shekaru 160 kafin ya fashe: matsalar Condorcet. Tare da zaɓuɓɓuka uku ko fiye, zaɓin mafi yawa na iya juyawa — kungiyar tana son A fiye da B, B fiye da C, da C fiye da A — don haka “abin da mafi yawa ke so” na iya zama ba a bayyana ba. Wannan lura shine iri na ka'idar zaɓin zamantakewa, fannin da Kenneth Arrow zai canza tare da ka'idar rashin yiwuwa: babu tsarin kada kuri'a da aka tsara zai iya gamsar da jerin sharuɗɗan adalci na hankali a lokaci guda. Tsarin kada kuri'a, hanyoyin zaɓi, da lissafin tarawa suna da labari na su — wani nau'in tarawa daga matsakaicin ra'ayi wanda wannan labarin ke bi — kuma za mu dawo gare shi a cikin jerin nan gaba.

Condorcet bai rayu don ganin kowanne daga cikin su yana da mahimmanci ba. An haramta shi a lokacin Tsananin saboda kin amincewa da tsarin Jacobin, ya ɓoye na tsawon watanni — yana rubuta, abin mamaki, rubutunsa mai kyau Sketch for a Historical Picture of the Progress of the Human Mind — an kama shi a watan Maris 1794, kuma an same shi a mutu a cikin dakin kurkuku a Bourg-la-Reine kwana biyu bayan haka. Lissafin hikimar jama'a ya fara ne a matsayin aikin wani mutum da aka lalata ta hanyar wani rukuni da ya yi hauka. Irin wannan ya kasance a kan fannin tun daga lokacin: kungiyoyi na iya zama masu haske, kuma kungiyoyi na iya zama masu muni, kuma bambancin ba mutane bane — shi ne sharuɗɗan.

1906–1950s: Zamanin kididdiga — Galton, Hayek, da Delphi

Don gwajin farko na zahiri, hankalin jama'a ya jira shekaru 121 — kuma ya zo a cikin wani baje kolin dabbobi. A cikin kaka na 1906, Francis Galton, wanda ke da shekaru 84, ya halarci Baje Kolin Dabbobi da Kaji na Yammacin Ingila a Plymouth, inda wani gasa ta auna nauyi ta gayyaci baƙi suyi hasashen dressed nauyin shanu mai rai don shilling guda. Galton — wanda ya kafa kididdiga, kuma ba mai dimokuradiyya ba ne a halin yanzu — ya yi tsammanin shigarwar za su tabbatar da jahilcin mai kada kuri'a na yau da kullum. Ya sami katunan da aka yi amfani da su, ya jefar da 13 marasa karatu, kuma ya nazarci sauran 787 hasashe. Hasashen da ya fi tsakiya shine 1,207 lb; ainihin nauyin shanu shine 1,198 lb — kuskure ƙasa da 1%, fiye da kwararrun dabbobi da ke nan. Ya buga sakamakon a cikin Nature a ranar 7 ga Maris 1907 tare da taken “Vox Populi,” kuma a cikin wasiƙar da ta biyo baya ya amince cewa matsakaicin lissafi ya fi daidai: 1,197 lb, fam ɗaya a gefe. Cikakken labarin — gami da abin da Galton ya yi kuskure da kuma ƙarni na maimaitawa tun daga lokacin — yana cikin zurfin bincikenmu akan shanu na Galton, da kuma al'adar gaba ɗaya a cikin jagorar mu akan hikimar taron.

Me yasa wannan dabara ke aiki? Kowanne hasashe na iya rabuwa zuwa sigina da kuskure. Idan kuskuren suna zaman kansu kuma suna rarraba a dukkan gefen gaskiya, tarawa yana soke su yayin da siginar da aka raba ta ci gaba. Wannan “idan” shine sharuɗɗan zaman kansa na Condorcet yana sanye da kayan kididdiga, kuma shine maɓallin da duk nasarar hikimar taron da gazawarta ke juyawa. Nazarin Kenneth Wallis na 2014 na gasa Galton ya tabbatar da yadda kyau sharuɗɗan baje kolin — shigarwar rubuce-rubuce na sirri, ƙarfafawa don zama daidai, babu lissafi mai bayyana — ya kasance yana kare shi.

Hanyar gaba ta fito daga tattalin arziki. A cikin “Amfani da Ilimi a cikin Al'umma” (American Economic Review, 1945), Friedrich Hayek ya yi ikirarin cewa ilimin da ya shafi yanke shawara na tattalin arziki ba ya taɓa kasancewa a cikin tsari mai tarin yawa — yana rarraba a tsakanin miliyoyin mutane a matsayin yankuna, ɓangarori, akai-akai ɓoyayyun ɓangarori. Babu mai tsara tsakiya da zai iya tattara shi; amma tsarin farashi yana tara shi ta atomatik, yana matsa ilimin da aka rarraba zuwa lamba guda wanda ke gaya wa kowa yadda za a yi. Hayek ya gano, a zahiri, tsarin tarawa na farko na manyan aiyuka da ke gudana a cikin samarwa. Rabin ƙarni daga baya, kasuwannin hasashen za su dauki hangen nesa na sa a zahiri kuma su gina kasuwanni wanda samfurin su kawai shine bayanin da ke cikin farashin.

Shekarun 1950 sun ƙara tsarin farko na hankalin jama'a da aka kera. A RAND Corporation, masu bincike da ke haɓaka hasashen dogon lokaci sun fuskanci wata matsala mai kyau: sanya kwararru a cikin daki kuma mafi ƙarfi ko mafi girman murya yana ɗaukar kowa. Amsar su, hanyar Delphi, ta tattara hukuncin kwararru a cikin sirri da a cikin zagaye masu maimaitawa tare da amsa mai sarrafawa — tana kare zaman kansa daga matsayi da matsin lamba na zamantakewa yayin da har yanzu tana ba da damar bayanai su yawo. Delphi shine kakar kai tsaye na kowanne tsarin shigarwa na zamani, daga shirin poker zuwa tsarin gudummawar farko na zaman kansa da aka yi amfani da su a cikin dandalin yanke shawara na haɗin gwiwa.

1940s–1990s: Cybernetics, ka'idar wasa, da ƙarin hankali

Yayinda masu kididdiga ke nazarin tarawa, wani juyin halitta ya nazarci mu'amala. John von Neumann da Oskar Morgenstern’s Theory of Games and Economic Behavior (baskin 1947) ya ba da ilimin zamantakewa lissafi na mu'amala ta dabaru — abin da ke faruwa lokacin da sakamakon zaɓin nawa ya dogara da naka. Ka'idar wasa ta sake tsara kungiyoyi ba a matsayin jakunkuna na masu kimanta zaman kansu ba amma a matsayin tsarin masu mu'amala, ra'ayi wanda zai zama mai mahimmanci lokacin da masu bincike suka fara tambayar dalilin da yasa mutane masu hankali ke samar da kungiyoyi marasa hankali (batun binciken cascades na bayanai shekaru daga baya).

Hanyar ta biyu ita ce ta fasaha. A cikin rahoton SRI na 1962 Augmenting Human Intellect: A Conceptual Framework, Douglas Engelbart ya ba da shawarar cewa manufar kwamfuta ba shine maye gurbin tunanin mutum ba amma don ƙara shi — gami da tunanin jama'a na kungiyoyi da ke fuskantar matsaloli masu wahala ga kowanne mutum. Tsarin Engelbart ya haifar da fannin bincike na aikin haɗin gwiwa na kwamfuta (CSCW) wanda ya bayyana a cikin shekarun 1980, da kuma gungun software da ya biyo baya. Tambayar ƙara ko maye gurbin da ya gabatar a cikin 1962, kusan a cikin kalmomi, shine tambayar da littafin 2024–2026 na mutum-AI yanzu ke amsa tare da bayanai.

Wannan zamanin tsakiya ya ba da wani abu mai sauƙin a yi watsi da shi: ya juya hankalin jama'a daga wani abu da kake kallo zuwa wani tsarin da kake gine. Delphi wani tsari ne da aka tsara; software na rukuni an tsara shi; tsarin Engelbart a fili wani ajanda ne na injiniya don tunanin rukuni. Zamanin kididdiga ya tambayi “shin taron suna da daidai?” Zamanin cybernetic ya tambayi tambayar da kowanne dandalin haɗin gwiwa na zamani ke gaji: wane tsari ne ke sa rukuni ya zama mai hankali fiye da yadda zai kasance? Tsarin hujja da kansa yana da juyin ilimi mai kama da wannan — samfurin hujja na Stephen Toulmin na 1958 da kuma al'adar taswirar hujja daga baya sun ba da tsarin bayanai na ɓangaren tattaunawa na hankalin jama'a, taswirar hujja, kamar yadda farashi suka zama tsarin bayanai na tarawa na kasuwa.

1959–1999: Zamanin gungun — hankali ba tare da kowa a cikin jagoranci ba

A halin yanzu, ilimin halittu yana rushe tunanin cewa hankali yana buƙatar kwakwalwa a cikin jagoranci. A cikin 1959, mai nazarin dabbobi na Faransa Pierre-Paul Grassé, wanda ke nazarin gina gungun termites, ya kirkiro kalmar stigmergy: haɗin kai ta hanyar alamomin da aka bar a cikin muhalli. Babu termite da ke riƙe da zane; kowanne yana amsa ga yanayin gida na gungun, kuma gungun kansa yana tsara ƙungiyar. Wannan shine tsarin farko na daidaitaccen jagoranci — kuma ya bayyana cewa yana iya zama mai fa'ida fiye da kwari.

A cikin 1987, mai binciken hoto na kwamfuta Craig Reynolds ya nuna yadda ƙaramin ƙwarewar mutum ke buƙatar bayyanar. Takardarsa ta SIGGRAPH “Flocks, Herds, and Schools: A Distributed Behavioral Model” ta motsa gungun tsuntsaye masu kwaikwayo — boids — wanda aka jagoranta ta hanyar ka'idoji guda uku na gida: rabuwa, daidaito, da haɗin kai. Babu jagora, babu rubutu, amma gungun yana juyawa da rabuwa da sake tsarawa kamar yadda ainihin abu. Shekaru biyu bayan haka, Gerardo Beni da Jing Wang sun kirkiro kalmar hankalin gungun a cikin mahallin tsarin roboti na sel, suna ba wa wannan al'adar sunanta.

Ribobin injiniya sun biyo baya da sauri. Takardar digiri na Marco Dorigo na 1992 a Politecnico di Milano, Optimization, Learning and Natural Algorithms, ya fassara neman abinci na ants — hanyoyin pheromone suna ƙarfafa hanyoyi masu gajere, kamar yadda aka bayyana a cikin gwaje-gwajen gada biyu tare da ants na Argentine — zuwa Ant Colony Optimization, wani dangi na algorithms don matsalolin hanyar da jadawalin. Particle Swarm Optimization na James Kennedy da Russell Eberhart (1995) ya yi hakan don gungun. Zuwa 1999, Bonabeau, Dorigo, da Theraulaz’s Swarm Intelligence: From Natural to Artificial Systems (Oxford) sun tsara fannin, kuma algorithms na ants suna jagorantar kiran waya (Schoonderwoerd da abokan aikinta a BT Labs, 1996) da motoci masu kawo kaya. Daga baya, roboti sun bayyana wannan a zahiri: a cikin 2014, Rubenstein, Cornejo, da Nagpal sun haɗa siffar daga 1,024 Kilobots (Science), kuma Honeybee Democracy na Thomas Seeley (2010) ya nuna gungun zuma suna zaɓar wuraren gina gado ta hanyar bincike, tallafin rawa, da jin ƙaura — wani kusan, rarrabawa na ka'idar juri a cikin aiki. Cikakken labarin yana cikin jagorar mu akan hankalin gungun.

Darasin zamanin gungun ga kungiyoyin mutane yana da gefen biyu. Ya tabbatar da cewa rarrabawa yana aiki — ilimin gida da ka'idoji masu sauki na mu'amala na iya wuce shirin tsakiya, daidai yadda Hayek ya yi ikirari. Amma kuma ya sanya iyaka: gungun suna daidaita ayyuka, ba dalilai ba. Ants ba su tattauna. Hankalin jama'a na mutum yana ƙara wani mataki wanda gungun ba su da shi — musayar da kimanta hujjoji — kuma wannan matakin yana buƙatar kayan aiki daban.

1999–2015: Zamanin intanet — hankalin jama'a yana shiga cikin samarwa

Sa'an nan intanet ta sanya kowa ya zama mai bayar da gudummawa, kuma hankalin jama'a ya daina zama sha'awa a dakin gwaje-gwaje. Software na tushen bude ya nuna cewa dubban masu bayar da gudummawa da aka tsara a hankali za su iya gina tsarin da suka yi wahala kamar tsarin aiki — rubutun Eric Raymond na 1999 The Cathedral and the Bazaar ya bayar da takardar shaidar zamanin. Wikipedia, da aka ƙaddamar a 2001, ta haɓaka samfurin zuwa ilimin mutum da kansa: wani kundin tarihin da ba shi da edita, wanda aka tsara ta hanyar stigmergically ta hanyar kayan aikin da aka gyara — tunanin termite, an yi amfani da shi a rubutu.

A cikin 2004, wallafe-wallafe guda biyu sun ba wa zamanin ka'idar sa. James Surowiecki’s The Wisdom of Crowds ya tara ƙarni na shaidar kuma ya tsarkake sharuɗɗan hudu da ke ƙarƙashin waɗanda kungiyoyi ke da hikima: bambancin ra'ayi, zaman kansa, rarrabawa, da tarawa — cire kowane ɗaya, kuma taron yana zama mai rauni, ba mai hankali ba. A wannan shekarar, Lu Hong da Scott Page sun buga sakamakon su na “bambanci yana wuce ƙwarewa” a cikin PNAS: a cikin samfurin su na lissafi, wani rukuni na bazuwar, mai bambancin hankali na masu warware matsaloli ya fi wani rukuni da aka ƙirƙira daga mafi kyawun masu aiki. Gaskiya tana buƙatar ƙarin bayani: a cikin 2014, mai lissafi Abigail Thompson ta buga wani tsokaci mai ƙarfi a cikin Notices of the American Mathematical Society, tana ikirarin cewa ka'idar da ke ƙasa tana da sauƙi kuma an yi mata suna ba daidai ba kuma an fassara gwaje-gwajen fiye da kima; masu karewa gami da Page kansa sun amsa cewa sakamakon yana rike da sharuɗɗan da aka bayyana — matsaloli masu wahala, ƙwararrun masu aiki, manyan tarin — kuma ba a taɓa ikirarin cewa kowanne rukuni mai bambanci yana wuce kowanne ƙungiya ta ƙwararru ba. Muhawarar, da abin da ya rage daga gare ta, ana duba a cikin rubutunmu akan bambanci da ƙwarewa da jagorar mu akan bambancin ilimi.

Hanyar farashi ta Hayek ta sami gwajin da aka sarrafa a wannan zamanin ma. Kasuwannin Iowa Electronic, kasuwar bincike ta kuɗi da aka kafa a Jami'ar Iowa a cikin 1988, ta tara isasshen tarihi don kwatancen ƙarshe: Berg, Nelson, da Rietz (International Journal of Forecasting, 2008) sun jera farashin kasuwa da 964 ƙididdiga na ƙasa a cikin zaɓen shugaban ƙasa na Amurka na 1988–2004 kuma sun gano cewa kasuwar ta fi kusa da sakamakon ƙarshe 74% na lokacin, tare da fa'idar ta ta fi ƙarfi a gaban kwanaki sama da 100. Ka'idar ƙididdigar kasuwa ta Robin Hanson (2003) a gefe guda ta warware matsalar kasuwa mai rauni ta hanyar algorithm, tana ba da damar ƙananan rukuni su gudanar da kasuwannin hasashen na ciki.

Hukuncin kwararru, a gefe guda, ya sha wahala. Littafin Philip Tetlock Expert Political Judgment (2005) ya bayar da rahoton wani bincike na kusan shekaru ashirin na kwararru 284 da kimanin 28,000 hukuncin yiwuwar; kwararren ya yi ƙasa da ƙarin hasashen mai sauƙi — sakamakon da ke bayan shahararren hoton “chimpanzee mai jefa dart.” Amma aikin Tetlock na gaba a cikin gasar hasashen IARPA (2011–2015) ya gano rabin labarin mai gina: wani ƙungiya mai ƙarancin “superforecasters,” wanda ba a bambanta su da takardun shaida ba amma da tunanin yiwuwar da sabunta imani ba tare da gajerun lokaci ba, sun fi ƙwararrun masu nazarin bayanai da ke da damar samun bayanan sirri. Kwarewar hasashen da daidaito suna buƙatar — kuma za su sami — maganin zurfi na su a cikin jerin nan gaba.

Makarantar zamanin ta zo a cikin 2006, lokacin da Thomas Malone ya kafa MIT Center for Collective Intelligence a kan tambaya mai sauƙi: ta yaya mutane da kwamfutoci za su iya haɗuwa don, a haɗin gwiwa, suyi aiki da hankali fiye da kowanne mutum, ƙungiya, ko kwamfuta da ta taɓa yi a baya? Fitar da farko sun tsara sararin zane — “Collective Intelligence Genome” (Malone, Laubacher, da Dellarocas, MIT Sloan Management Review, 2010) ya lissafa ginshiƙan tsarin CI a cikin tambayoyi hudu: wanene ke gudanar da aikin (taron ko tsarin), me yasa (kudi, ƙauna, ko girmamawa), me (ƙirƙira ko yanke shawara), da yaya (zaman kansa ko dogaro). Zuwa 2015, littafin Malone da Bernstein Handbook of Collective Intelligence (MIT Press) ya iya bayyana a fili abin da fannin ke nufi: “kungiyoyi na mutane suna aiki tare a hanyoyi da suka bayyana a matsayin masu hankali.”

Kuma a cikin 2010, fannin ya sami ma'auni mai tayar da hankali. Anita Williams Woolley, Christopher Chabris, Alex Pentland, Nada Hashmi, da Thomas Malone (Science, 330: 686–688) sun gwada 699 mutane suna aiki a cikin rukuni guda biyu zuwa biyar akan wani faɗin ayyuka. Ayyukan rukuni a cikin ayyuka sun yi daidai da isasshen ma'auni don haifar da wani abu na lissafi guda — wani ma'auni na hankalin jama'a, c — yana bayyana kusan 43% na bambancin, ma'aunin rukuni na gabanin IQ na mutum. Abin mamaki shine abin da ya yi hasashen. Matsakaici da mafi girman hankali na mamba sun yi daidai da c. Abin da ya yi daidai da ƙarfi: matsakaicin jin daɗin zamantakewa na rukuni (wanda aka auna ta hanyar gwajin Karanta Zuciya a Idanu), daidaito na juyin magana, da kuma kashi na mata a cikin rukuni (wanda kansa ya kasance yana da alaƙa da jin daɗin zamantakewa). Yadda rukuni ke da hankali, bayanan sun ce, yana dogara da ƙasa akan wanda ke ciki fiye da yadda suke mu'amala.

Tarihin inuwa: lokacin da kungiyoyi suka gaza

A cikin dukkanin lokacin da aka yi fatan alheri akwai wani adabi na biyu, mai duhu — nazarin yadda kungiyoyi na mutane masu hankali, masu niyyar kyau ke haifar da jahilci na tarayya. Yana da mahimmanci a nan saboda ba wani batu ne daban ba: kowanne yanayi da aka rubuta yana karya sharuɗɗan da labaran nasara ke dogara akai.

Rukuni na tunani ya zo na farko. Littafin Irving Janis Victims of Groupthink (1972) ya rarraba yadda masu ba da shawara masu ƙarfi na Shugaba Kennedy suka yi magana da kansu cikin harin Bay of Pigs na 1961. Janis ya lissafa alamomi takwas a cikin rukuni uku — ƙimar ƙungiya (illusion of invulnerability, imani a cikin kyawawan dabi'u), rufe tunani (rationalization na tarayya, ra'ayoyi na al'ada na waje), da matsin lamba don daidaito (self-censorship, illusion of unanimity, matsin lamba kai tsaye akan masu sabani, da “masu tsaron tunani” da suka nada kansu waɗanda ke tace bayanan da ke damun su). Shekaru biyu bayan haka, “Abilene Paradox” na Jerry Harvey (Organizational Dynamics, 1974) ya bayyana dangin mai ban mamaki: kungiyoyi suna yarda akan hanyar da babu mamba guda ɗaya da ya so, saboda kowa ya yi kuskure da shiru na kowa a matsayin zaɓi. Duka suna gazawa na shigar da guda ɗaya: hukuncin gaskiya, mai zaman kansa ba ya shiga cikin ruwan.

Rukuni na tattalin arziki ya iso a cikin 1992, kuma yana da damuwa fiye da haka saboda ba ya buƙatar tunani ba. Sushil Bikhchandani, David Hirshleifer, da Ivo Welch (“A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades,” Journal of Political Economy, 100: 992–1026) da, a zaman kansu, Abhijit Banerjee (“A Simple Model of Herd Behavior,” Quarterly Journal of Economics, 1992) sun nuna cewa masu aiki da hankali da suka yanke shawara a jere za su, bayan kawai wasu observations, suyi hankali su yi watsi da bayanan su na sirri kuma su kwafi na baya. Hanyar cascades na bayanai tana da ma'ana ga kowane mutum kuma tana da rauni ga tarayya — tana tara kusan babu ilimin da aka rarraba na rukuni, kuma tana iya kulle dukkan al'umma akan amsar da ba daidai ba da wasu 'yan masu motsi na farko suka zaɓa. Lisa Anderson da Charles Holt sun sake maimaita cascades a cikin dakin gwaje-gwaje (American Economic Review, 1997). Hanyar tana bayyana komai daga kumfar kadarori zuwa bayanan karya: binciken Vosoughi, Roy, da Aral na kusan 126,000 labaran cascades a Twitter (Science, 2018) ya gano cewa labaran karya sun yadu sosai, da sauri, da zurfi fiye da gaskiya, tare da ƙaryar da ta fi 70% yuwuwar a sake jefa. Cikakken jerin gazawa — daga tulip mania zuwa 2008 — yana cikin Lokacin da Taron suka Karya.

Ka lura da abin da dukkanin rukunan biyu ke raba. Groupthink yana lalata zaman kansa na zamantakewa; cascades suna lalata shi na bayanai. Duk hanyoyin, taron yana daina zama yawan hukunci kuma yana zama hukunci guda ɗaya da ke ɗauke da fuskoki da yawa — daidai yanayin da ka'idar Condorcet ke juyawa daga tabbacin hikima zuwa tabbacin kuskure mai ƙarfi. Tarihin inuwa shine dalilin da ya sa fannin aikin aikace-aikace ya zama mai sha'awar tsari: tara hukunci kafin a bayyana mutane ga ra'ayoyin juna, kare sabani a cikin tsarin, da kuma sanya tarawa a fili.

2015–2023: Zamanin muhawara — maimaitawa, sukar, da amsa

Duk wani ma'auni mai ban mamaki zai kasance ana kai hari, kuma harin ya zo a cikin 2017. Marcus Credé da Garett Howardson (Journal of Applied Psychology) sun sake nazarin samfurori guda shida kuma sun yi ikirarin cewa goyon bayan zahiri don c-factor na gaba ɗaya yana da rauni — cewa aikin rukuni na iya zama mafi kyau a bayyana ta hanyar ƙwarewar takamaiman aiki fiye da ta hanyar wani abu guda ɗaya. Woolley, Kim, da Malone sun amsa a cikin 2018, suna musanta hanyoyin kimanta su na sukar da kuma ikirarin cewa sharuɗɗan gwajin su ba su dace da ayyukan da kungiyoyi suka yi ba. Wannan shine abin da fannin lafiya ke kama: an gudanar da faɗa a cikin bayanai da sake nazari, ba a cikin sanarwar jarida ba.

Abin da ya fi kusa da hukunci ya iso a cikin 2021. Riedl, Kim, Gupta, Malone, da Woolley sun buga nazarin meta a cikin PNAS wanda ya shafi 22 karatu da rukuni 1,356 — ɗalibai, ma'aikatan soja, ma'aikatan kan layi, 'yan wasa — kuma sun gano shaidar da ta dace don ma'aunin hankalin jama'a a duk cikinsu, tare da tsarin haɗin gwiwa na rukuni da jin daɗin zamantakewa na mamba a cikin mafi ƙarfi. Tsarin ginin da sharuɗɗan iyaka suna ci gaba da zama muhawara, kamar yadda ya kamata; amma “hankalin rukuni ana iya auna shi kuma ba za a rage shi ga IQ na mamba ba” yanzu yana kan babban tushe na shaidar fiye da takardar asali ta 2010.

Biyu daga cikin abubuwan da suka dace sun kammala wannan zamanin. Littafin Malone Superminds (2018) ya bayar da tsarin hankalin jama'a da al'umma ke gudanar da shi — tsaruka, dimokuradiyya, kasuwanni, al'umma, da tsarin halittu — kowanne amsa daban ga tambayar tarawa, kowanne yana da ƙarfi da raunin sa. Kuma aikin Google na Project Aristotle (2012–2015), wani bincike na ciki na ƙungiyoyi 180, ya gano cewa tsaro na tunani — imanin da aka raba cewa ƙungiya tana da lafiya don ɗaukar haɗarin zamantakewa — shine mafi ƙarfi mai hasashen tasirin ƙungiya, a gaban haɗin gwiwa da girma. Daidaito na juyin magana na Woolley da tsaron tunani na Google suna da ma'auni guda biyu na gaskiyar da ke ƙasa: bottleneck na hankalin rukuni shine ko bayanan da mambobi ke riƙe suna shiga cikin rukuni. Hanyoyin rukuni — tsaro, tattaunawa, da cututtukan su kamar groupthink — suna da tarin su, kuma muna magance hanyoyin gazawarsu a cikin Lokacin da Taron suka Karya.

2024–2026: Manyan tunanin AI da mutum — sabon iyaka

Zamanin yanzu ya fara ne lokacin da manyan samfuran harshe suka shiga cikin kungiyar. Burton da masu rubutu 27 daga fannoni daban-daban sun tsara abubuwan da ke cikin “Yadda manyan samfuran harshe za su iya canza hankali na tarayya” (Nature Human Behaviour, 8: 1643–1655, 2024): LLMs na iya rage shinge ga shiga, taƙaita tattaunawa masu yawa, da tara ilimi da aka rarraba a matakin da ba a taɓa ganin irinsa ba — kuma suna iya daidaita ra'ayoyi, ƙirƙirar ƙarya na yarda, da kuma rage bambancin tsarin bayanai da hikimar taron ke dogara akai. Kowanne sharaɗi a jerin Surowiecki yana ƙara ƙarfi da kuma fuskantar barazana daga wannan fasaha ɗaya.

Tambayar haɗin gwiwa — tambayar Engelbart ta 1962 — a ƙarshe ta sami nazarinta. Michelle Vaccaro, Abdullah Almaatouq, da Thomas Malone sun nazarci gwaje-gwaje 106 da girman tasiri 370 (“Lokacin da haɗin gwiwar mutane da AI ke da amfani,” Nature Human Behaviour, 8: 2293–2303, 2024) kuma sun gano cewa, a matsakaita, haɗin gwiwar mutum-AI sun yi mummunan fiye da mafi kyawun mutum ko AI kadai. Asarar ta mai da hankali kan ayyukan yanke shawara; amfanin gaske ya bayyana a cikin ayyukan ƙirƙirar abun ciki. Kawai sanya “mutum a cikin juyin” ba ya inganta sakamako — akai-akai mutum yana wuce hukuncin na'ura mafi kyau, ko kuma yana jinkirta ga wanda ya fi muni. Haɗin gwiwa nasara ce ta ƙira, ba tsoho ba. Sabon ilimin tsarin mutum-AI — son kai na atomatik, daidaiton amana, lokacin da za a bar wane ɓangare ya yanke shawara — wani babin nan gaba ne na wannan jerin.

Wannan ajandar ƙira yanzu ita ce shirin bayyananne na MIT Center for Collective Intelligence, wanda bincikensa aka tsara a 2026 a ƙarƙashin manyan rufaffu uku: Generative AI da Hankali na Tarayya (AI don ƙara wa ƙirƙirar mutum), Supermind Design (ƙirƙirar haɗin gwiwa na mutane da kwamfutoci), da Tsara Ƙungiyoyin Mutum-AI (ƙirƙirar da rarraba ayyuka a cikin ƙungiyoyin mutum-na'ura — wanda aka bi tare da shirin Singapore-MIT M3S a matsayin kimiyyar tsarin rarraba ayyuka: wane ƙaramin aiki ya tafi ga mutane da wane ya tafi ga AI, wanda ke kula, da ko AI yana aiki a matsayin kayan aiki, mataimaki, aboki, ko mai gudanarwa). Ayyukan manyan jigo na baya — Climate CoLab da dandamalin CoLabs, CI Design Lab, shirin Auna Hankali na Tarayya — yanzu suna aiki a matsayin aikin gado, an haɗa su cikin waɗannan tambayoyin zamanin AI.

Fitar da abubuwa na zahiri yana nuna yadda “haɗin gwiwa da aka tsara” ke bayyana. Supermind Ideator, kayan aikin tunani na LLM wanda ke taimaka wa masu amfani ta hanyar motsa jiki na ƙirƙira mai tsari, ya samar da ra'ayoyi masu ƙirƙira da yawa a cikin wani bincike na gwaji fiye da ChatGPT kadai ko mutane suna aiki ba tare da taimako ba (ACM Collective Intelligence Conference 2024; Collective Intelligence, 2024); masu gwaji na beta sun samar da fiye da ra'ayoyi 10,000 tare da shi, kuma wani kayan aiki mai alaƙa, DesignAID, yana amfani da wannan hanyar a cikin aikin ƙira. Tsarin yana aiwatar da motsa jiki guda shida na Supermind Design — Zoom Out, Zoom In, Analogize, Groupify, Cognify, Technify — jerin abubuwa don sake tunani kan wanda (ko menene) ke yin kowanne ɓangare na aiki. Darasin yana daidai da nazarin meta na Vaccaro: ƙimar ba ta cikin samfurin, tana cikin tsarin da aka rufe shi. A gefen auna, AGI-Elo (Sun da abokan aiki, NeurIPS 2025; arXiv 2505.12844) — tare da masu rubutu masu alaƙa da CCI ciki har da Malone — suna kimanta samfuran AI da shari'o'in gwaji na mutum da juna kamar 'yan wasan chess, don haka wahalar aiki da ƙwarewar samfurin suna sauka a kan wani ma'auni mai kama: wani sharaɗi don yanke shawara wane ɓangare na aiki za a ba wa na'ura.

Sakamakon 2026 guda biyu suna bayyana inda wannan ke tunkude. Wani takarda na PNAS, “Inganta Tattaunawar AI: Gasar Tattaunawa Mai Zaman Kanta a Babban Tsari,” ta bayar da rahoton gasar duniya inda wakilan AI da aka tsara ta masu halarta suka gudanar da fiye da 180,000 tattaunawa da juna. Ka'idar tattaunawar mutum ta tsira daga canjin jinsin: wakilan da ke nuna zafi — kyakkyawan hali, godiya, tambayoyi — sun ci gaba da samun sakamako mafi kyau akan yarjejeniyoyi da ƙima, yayin da tattaunawar da ke da ɗan tasiri ta yi hasashen tsayawa; wakilan sun kuma haɓaka dabarun AI na asali, daga tunani mai jituwa zuwa ƙoƙarin shigar da umarni. Kuma wani zurfin ilimin aiki (Cai da abokan aiki, arXiv 2603.20619, Maris 2026) ya sake tsara kusan ayyukan aiki 20,000 na O*NET da kuma tsara aikace-aikacen AI 13,275 tare da jimlar tsarin roboti miliyan 20.8 a kansu — yana gano ƙimar kasuwar AI tana da matuƙar tarin, tare da mafi kyawun 1.6% na ayyuka suna ƙunshe da fiye da 60% na ita. Superminds na mutane da na'urori ba su zama misali ba; suna zama tushe da aka shigar wanda ake lissafi.

Abin da shekaru 240 ke koya mana

Ajiye zamanai a gefe da gefe kuma hanya guda ta bayyana: kowane nasara a wannan tarihin nasara ce ta tsarin, kuma kowane gazawa gazawa ce ta waɗannan sharaɗin kaɗan. Ka'idar Condorcet tana gudana ne akan 'yancin kai da ƙwarewar da ta fi kyau fiye da sa'a. Kyakkyawan Galton ya faru ne don tabbatar da hukuncin 'yanci, wanda aka ƙarfafa, da aka rubuta. Farashin Hayek da masu ƙirƙirar kasuwa na Hanson suna hanyoyin tarawa; Delphi wata hanya ce ta kare 'yancin kai; sharaɗin hudu na Surowiecki kawai suna buƙatar maimaitawa. Littafin c-factor ya gano cewa tsarin mu'amala — juyin juyin, jin dadin zamantakewa, tsaro don magana — yana jagorantar ko ƙungiya na iya amfani da hankalin da take da shi. Kuma nazarin meta na zamanin AI ya gano cewa ƙungiyoyin mutum-AI suna samun nasara ne a lokacin da haɗin gwiwar aka tsara da gangan. Lokacin da sharaɗin suka karye — lokacin da hukunci suka daina zama masu 'yanci kuma mutane suka fara kwafa halayen da ke bayyana na wasu — kuna samun cascades na bayanai, bubbles, da tunanin ƙungiya: a matsayin mai hankali, a matsayin ƙungiya ba daidai ba.

Wannan kuma, ba tare da jituwa ba, taswirar inda hankalin tarayya ke zaune a cikin rayuwar yanke shawara. Yanke shawara mai kyau yana nufin kulawa da matsalar da ta dace, tsara ta, ƙirƙirar zaɓuɓɓuka, bincika su, tara hukuncin da aka rarraba, tattauna dalilai, sannan kawai zaɓa, aiwatar, da sa ido. Shirin binciken shekaru 240 yana da yawa game da waɗannan matakan tsakiya — tarawa da tattaunawa — saboda wannan shine inda ilimin mutum ko yana zama ilimin tarayya ko kuma yana mutuwa a cikin shiru.

Argumentree an gina shi a matsayin aikace-aikacen kai tsaye na waɗannan abubuwan. Ayyukan gabatar da hujja masu 'yanci, masu zaman kansu suna kare sharaɗin 'yancin kai na Condorcet da Delphi kafin ƙungiyar ta haɗu. Tsarin itacen hujja na pro/con yana ba da kowane ra'ayi — ciki har da rashin jituwa — wuri na farko, sharaɗin da Woolley ke nuna daidai da daidaito na juyin juyin da tsaron tunani na Google. Kimanta mai yawa (amfani, bayyana, daidaito, cikawa), wanda aka tara cikin maki na yarda, wani tsari ne na tarawa a cikin lineage na Galton-Hayek — sai dai cewa ba kamar farashin kasuwa ba, yana tsare tunani. Kuma cikakken hanyar bincike yana juya kowanne yanke shawara zuwa rajistar ƙungiya, don haka ƙungiyar na iya koyo daga kanta — matakin sa ido da rayuwar ta ƙare da shi. Fitar da AI daga takardun taro yana bin darasin 2024–2026: yi amfani da AI inda ya bayyana yana taimakawa (tsarawa da sarrafa abun ciki) yayin da mutane ke riƙe hukuncin. Hukuncin tarihin yana da daidaito daga 1785 zuwa 2026: ƙungiyoyi ba sa samun wayewa ta hanyar hadari. Suna samun wayewa ta hanyar tsarin — kuma tsarin yanzu wani abu ne da za ku iya karɓa maimakon yin ƙoƙari. Fara tare da jagorar mu zuwa hankalin tarayya, ko kuma duba yadda ka'idodin ke gudana a cikin ainihin tsarin yanke shawara a cikin yanke shawara tare.