Fahimtar Dabarun Ƙirƙirar Ra'ayoyi na ɗaliban EFL don Rubuce-rubucen Ƙirƙira tare da Kayan Aikin NLG

1. Gabatarwa

Rubutu wata muhimmiyar ƙwarewa ce don sadarwa da nasara a ilimi. Ga ɗaliban Ingilishi a matsayin Harshen Waje (EFL), rubuce-rubucen ƙirƙira suna gabatar da ƙalubale na musamman, musamman a lokacin ƙirƙirar ra'ayi. Wannan binciken yana bincika mahadar Fasahar Hankali (AI), musamman kayan aikin Ƙirƙirar Harshe na Halitta (NLG), da koyarwar EFL. NLG ta ƙunshi tsarin kwamfuta da ke samar da rubutu mai kama da na ɗan adam daga tsararrun bayanai ko umarni. Tambayar binciken ta ta'allaka ne akan yadda ɗaliban EFL ke mu'amala da kayan aikin NLG ta dabara don ƙirƙira, kimantawa, da zaɓar ra'ayoyi don ayyukan rubuce-rubucen ƙirƙira, wani tsari mai mahimmanci amma sau da yawa yana da ban tsoro ga masu koyon harshe.

2. Hanyar Bincike

Binciken ya yi amfani da hanyar nazarin shari'a mai zurfi don samun cikakken fahimta game da dabarun ɗalibai.

2.1 Mahalarta da Ƙirar Taron Karatu

Ɗaliban sakandare huɗu daga Hong Kong sun shiga cikin tsararrun tarurrukan karatu. An gabatar da su ga kayan aikin NLG daban-daban (misali, kayan aikin da suka dogara da samfura kamar GPT-3) kuma an ba su aikin rubuta gajerun labarai waɗanda suka haɗa kalmominsu da rubutun da waɗannan tsarin AI suka samar. Ƙirar taron karatu ta sauƙaƙa gogewa da kuma tunani na gaba.

2.2 Tattara Bayanai da Bincike

Bayanan farko sun ƙunshi tunanin rubuce-rubucen ɗaliban bayan taron karatu, inda suka amsa tambayoyin da aka jagoranta game da gogewarsu. An yi amfani da bincike na jigo akan waɗannan bayanan masu inganci don gano maimaitawar tsari, dabarun, da halayen da suka shafi amfani da kayan aikin NLG don ƙirƙirar ra'ayi.

3. Sakamako da Binciken

Binciken ya bayyana wasu mahimman tsare-tsare a yadda ɗaliban EFL ke amfani da NLG don rubuce-rubucen ƙirƙira.

3.1 Dabarun Neman Ra'ayi tare da Kayan Aikin NLG

Ɗaliban ba su fuskanto kayan aikin NLG da takarda ba. Sau da yawa suna shiga cikin mu'amala tare da ra'ayoyin da suka riga su kasance ko jagororin jigo. Daga nan sai a yi amfani da kayan aikin NLG a matsayin mai haɓakawa, gyara, ko bincika ra'ayoyi masu alaƙa, maimakon a matsayin tushen abun ciki kawai.

3.2 Kimanta Ra'ayoyin da NLG ta Ƙirƙira

Wani binciken da aka gano shi ne kyamar ko shakku ga ra'ayoyin da kayan aikin NLG suka samar kawai. Ɗaliban sun kimanta abun cikin da AI ta samar don dacewa, asali, da daidaituwa tare da labarin da suke nufi, sau da yawa sun fi son gyara sosai ko amfani da shi azaman kwarin gwiwa kawai maimakon haɗawa kai tsaye.

3.3 Zaɓin Kayan Aikin NLG

Lokacin zaɓar tsakanin kayan aikin NLG ko umarni daban-daban, ɗaliban sun nuna fifikon kayan aikin da ke samar da mafi yawan zaɓuɓɓukan fitarwa. Wannan hanyar "yawa-fiye-da-inganci na farko" ta ba su babban saitin kayan daga inda za su zaɓa da haɗa ra'ayoyi.

4. Tattaunawa da Abubuwan da ke Tattare da Shi

Binciken ya nuna mahimmin matsayi, ba na jinkiri ba da ɗaliban ke ɗauka lokacin amfani da mataimakan rubutu na AI.

4.1 Abubuwan da ke Tattare da Koyarwa

Binciken ya nuna cewa malamai yakamata su tsara kayan aikin NLG ba a matsayin maye gurbin ƙirƙirar ɗalibai ba amma a matsayin "abokan haɗin gwiwa na ƙirƙirar ra'ayi". Koyarwa yakamata ta mai da hankali kan ƙwarewar kimantawa, dabarun umarni, da dabarun haɗawa don haɗa abun cikin ɗan adam da na na'ura yadda ya kamata.

4.2 Iyakoki da Bincike na Gaba

Ƙaramin girman samfurin yana iyakance yaduwa. Bincike na gaba yakamata ya haɗa da ƙungiyoyin masu koyon EFL masu girma, masu bambanta, da bincike na dogon lokaci don ganin yadda dabarun ke haɓaka tare da ƙarin gogewa da ƙwarewa.

5. Binciken Fasaha da Tsarin Aiki

Mahimmin Fahimta: Wannan takarda ba game da gina mafi kyawun samfurin NLG ba ne; bincike ne mai mahimmanci na mu'amalar ɗan adam-kwamfuta (HCI) wanda ke fallasa "matsalar mil na ƙarshe" a cikin ƙirƙirar da AI ke taimakawa. Babban matsalar ba ƙarfin AI na samar da rubutu ba ne—masu canzawa na zamani kamar GPT-4 sun ƙware a hakan—amma ƙarfin mai amfani na amfani da wannan ƙarfin ta dabara. Binciken ya nuna cewa ɗaliban EFL suna ɗaukar fitarwar NLG a matsayin kayan duniya maras inganci, ba samfurin ƙarshe ba, wata hanya ce mai zurfi da daidai wacce sau da yawa ba ta cikin tallan kayan aikin AI.

Tsarin Hankali: Hankalin binciken yana da inganci: lura da hali (tarurrukan karatu) → kama dalili (tunani) → gano tsari (binciken jigo). Ya yi daidai ya kauce wa tarkon auna "inganci" na fitarwa a sarari, yana mai da hankali maimakon akan tsari (nema, kimantawa, zaɓi). Wannan ya yi daidai da mafi kyawun ayyuka a cikin binciken ƙirar ilimi, inda fahimtar tafiyar mai amfani ya fi mahimmanci kafin a ba da magani.

Ƙarfi & Kurakurai: Ƙarfinsa shine tushensa, mai da hankali kan inganci akan takamaiman rukunin masu amfani da ba a biya su ba (ɗaliban EFL). Kurakuransa shine ma'auni. Tare da N=4, binciken shari'a ne mai jan hankali amma ba tabbatacce ba. Ya rasa damar ƙididdige halaye—misali, kashi nawa na fitarwar NLG ake amfani da shi? Nawa ne maimaitawar umarni ke faruwa? Kwatanta dabarun da ma'auni (rubuce-rubuce ba tare da AI ba) zai ƙarfafa da'awar tasirin NLG. Binciken kuma bai shiga cikin cikakkun bayanan fasaha na kayan aikin NLG da aka yi amfani da su ba, wanda dama ce da aka rasa. Zaɓin samfurin (misali, samfurin ma'auni na 175B da na 6B) yana shafar ingancin fitarwa da gogewar mai amfani sosai. Kamar yadda aka lura a cikin takardar GPT-3 ta asali ta Brown et al. (2020), girman samfurin yana shafar daidaituwa da ƙirƙira kai tsaye a cikin koyon ƴan harbi, wanda yana da alaƙa sosai da mahallin wannan binciken.

Fahimta mai Aiki: Ga masu haɓaka EdTech: Gina kayan aikin da ke tallafawa zaɓe, ba kawai ƙirƙira ba. Yi tunanin "allunan sarrafa ra'ayi" tare da alama, tarawa, da haɗa siffofi don fitarwar NLG. Ga malamai: Ƙirƙiri ayyuka waɗanda ke koyar da "injiniyan umarni" a matsayin babbar ƙwarewar karatu. Matsa zuwa bayan "amfani da kayan aikin" zuwa "yi wa kayan aikin tambayoyi." Ga masu bincike: Mataki na gaba shine haɓaka tsarin da aka tsara don ƙirƙirar ra'ayi da NLG ke taimakawa. Muna buƙatar rarrabuwar dabarun ɗalibai, watakila a zana su azaman bishiyar yanke shawara ko saitin dabaru. Wani samfurin bincike na iya tsara yanke shawarar ɗalibin na amfani da ko gyara ra'ayin AI $I_{AI}$ bisa ga amfanin da ake ganinsa $U$, daidaitawa da tsarin tunaninsu $M$, da farashin haɗawa na fahimi $C$, wanda aka tsara kamar haka: $P(\text{Amfani } I_{AI}) = f(U(I_{AI}, M), C(I_{AI}))$. Bugu da ƙari, ra'ayin amfani da AI a matsayin "mai haɗin gwiwa" maimakon kayan aiki yana juyar da binciken daga binciken haɗin gwiwar ɗan adam-AI a wasu fagage, kamar aikin Amershi et al. (2019) akan jagororin mu'amalar ɗan adam-AI, wanda ke jaddada ƙa'idodi kamar "raba iko" da "daidaiton mahalli."

Misalin Tsarin Bincike (Ba Code ba): Ka yi la'akari da ɗalibin da ke rubuta labari game da "robat da ya ɓace a cikin daji." Tsarin da aka samo daga wannan binciken zai iya jagorantar su ta hanyar tsarin madauki na ƙirƙirar ra'ayi:

Iri: Fara da ainihin ra'ayinka (robat da ya ɓace).
Umarni & Ƙirƙira: Yi amfani da NLG tare da takamaiman umarni (misali, "Ƙirƙiri ƙalubalen tunani 5 da robat ke fuskanta," "Lissafa halittun daji 3 da ba a saba gani ba da ya haɗu da su").
Kimanta & Tace: Kimanta kowane abu da aka ƙirƙira sosai. Shin ya dace da sautin? Shin yana da asali? Yi musu lakabi da "Amfani," "Daidaituwa," ko "Jefar."
Haɗawa: Haɗa mafi kyawun ra'ayoyin da AI ta ƙirƙira tare da ainihin makircinka, warware sabani.
Maimaita: Yi amfani da sabon haɗin gwiwar don ƙirƙirar ƙarin ingantattun umarni don abubuwan labari na gaba (misali, "Yanzu ƙirƙiri tattaunawa tsakanin robat da kurege mai son kai bisa ga ƙalubalen da aka zaɓa").

Wannan yana mai da kayan aikin mara aiki zuwa abokin tunani mai aiki.

Sakamakon Gwaji & Bayanin Ginshiƙi: Yayin da binciken na asali ya gabatar da jigogi masu inganci, yi tunanin bincike na gaba wanda zai ƙididdige waɗannan halayen. Wani zanen ginshiƙi na hasashe zai iya nuna: "Matsakaicin Adadin Fitarwar NLG da Aka Kimanta kowane Abun Labari." X-axis zai lissafa abubuwan labari (Hali, Saiti, Rikici, Magani), kuma y-axis zai nuna ƙidaya. Za mu iya ganin adadi mai yawa don "Hali" da "Saiti," yana nuna ɗaliban suna amfani da NLG mafi yawa don ƙirƙirar abubuwan tushe. Wani zanen kuma zai iya zama ginshiƙi mai tarawa wanda ke nuna "Rarraba Ra'ayoyin da NLG ta ƙirƙira," tare da sassa don "An Yi Amfani da Kai Tsaye," "An Gyara Sosa," da "An Jefar," yana bayyana babban adadin gyare-gyaren da aka nuna ta hanyar binciken kyama.

6. Ayyuka da Jagororin Gaba

Hanyar nan tana nuni zuwa ga mataimakan rubutu masu keɓancewa, masu daidaitawa. Kayan aikin NLG na gaba don ilimi za su iya:

Gina Tsarin Bisa Ƙwarewa: Daidaita rikitarwar fitarwa da jagora bisa matakin harshe na mai koyo (CEFR A1-C2).
Haɗa Ƙirƙirar Ra'ayi ta Hanyoyi Daban-daban: Ƙirƙira ba kawai rubutu ba, amma allunan yanayi, hotunan halaye, ko zane-zanen makirci don ƙarfafa hanyoyin fahimi daban-daban.
Ra'ayoyi na Bayanan Fahimi: Bincika tsarin umarni da zaɓin ɗalibin don ba da ra'ayi kamar: "Kuna son jefar da ra'ayoyin da suka shafi rikici na ciki. Yi ƙoƙarin bincika umarni game da tsoron halin."
Ƙirƙirar Ra'ayi ta Harsuna Daban-daban: Ga masu koyon EFL, ba da damar ƙirƙirar ra'ayi a cikin harshensu na asali tare da tallafin fassarar da daidaitawa, rage nauyin ƙirƙirar ra'ayi a cikin harshen waje.
Haɗawa tare da Nazarin Koyo: Kamar yadda cibiyoyi kamar Makarantar Koyar da Ilimi ta Stanford suka gabatar a cikin aikinsu na AI a ilimi, waɗannan kayan aikin za su iya ciyar da bayanai cikin allunan da ke taimaka wa malamai gano ɗaliban da ke fuskantar wahala tare da takamaiman fannoni na ƙirƙirar ra'ayi.

Manufa ta ƙarshe ba AI da ke rubuta wa ɗalibin ba ce, amma AI da ke haɓaka muryar ɗalibin na musamman da ƙarfin ƙirƙira, yana mai da takarda mara komai mai ban tsoro a sararin haɗin gwiwa.

7. Nassoshi

Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., ... & Amodei, D. (2020). Harsunan samfura masu koyon ƴan harbi ne. Ci gaba a cikin tsarin sarrafa bayanai na jijiyoyi, 33, 1877-1901.
Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., ... & Horvitz, E. (2019). Jagororin don mu'amalar ɗan adam-AI. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, 1-13.
Graham, S., & Perin, D. (2007). Binciken meta na koyarwar rubutu ga ɗaliban matasa. Journal of Educational Psychology, 99(3), 445.
Kaufman, J. C., & Beghetto, R. A. (2009). Bayan babba da ƙarami: Samfurin ƙirƙira na huɗu. Review of General Psychology, 13(1), 1-12.
Dawson, P. (2005). Rubuce-rubucen ƙirƙira da sabon ɗan adam. Routledge.
Woo, D. J., Wang, Y., Susanto, H., & Guo, K. (2023). Fahimtar Dabarun Ƙirƙirar Ra'ayoyi na ɗaliban EFL don Rubuce-rubucen Ƙirƙira tare da Kayan Aikin NLG. [Sunan Jarida].