Teburin Abubuwan Ciki
1. Gabatarwa
ChatGPT, wani chatbot na AI na zamani (SOTA) wanda ke iya samar da bayanai, ya sami shahara sosai saboda yuwuwar sa na canza ilimi, musamman a fagen rubutun Turanci a matsayin Harshen Waje (EFL). Duk da haka, yin aiki tare da ChatGPT yadda ya kamata yana buƙatar ɗalibai su ƙware a injiniyan umarni—ƙwarewar tsara umarni masu inganci don samun sakamakon da ake so. Wannan takarda ta yi nazarin abubuwan da ke cikin umarni da tsarin ɗaliban EFL na sakandare yayin da suke kammala aikin rubutu tare da ChatGPT a karon farko. Ta hanyar nazarin shari'a na hanyoyi huɗu daban-daban, marubutan sun kwatanta tsarin gwaji da kuskure kuma sun nuna buƙatar koyar da injiniyan umarni a fili a azuzuwan EFL.
2. Binciken Littattafai
2.1 ChatGPT a Rubutun EFL
ChatGPT na iya taimaka wa ɗaliban EFL ta hanyar samar da ra'ayoyi, ba da shawarwarin ƙamus, da kuma gyara nahawu. Duk da haka, ba tare da umarni masu kyau ba, sakamakon na iya zama marasa dacewa ko kuma marasa amfani. Binciken Guo da sauransu (2023) ya nuna cewa ɗalibai sukan yi wahala wajen tsara umarni masu tasiri, wanda ke haifar da mu'amala mara kyau.
2.2 Injiniyan Umarni a matsayin Ƙwarewa
Injiniyan umarni ya ƙunshi fahimtar iyawa da gazawar samfurin. Yana buƙatar gyara ta hanyar maimaitawa, ƙayyadaddun bayanai, da sanin mahallin. Nazarin (misali, Woo da sauransu, 2023) ya nuna cewa masu amfani da ba su da fasaha, ciki har da ɗaliban EFL, sukan shiga gwaji da kuskure ba tare da dabaru na tsari ba.
3. Hanyoyin Bincike
3.1 Mahalarta da Wuri
Mahalarta sun kasance ɗaliban sakandare 12 na EFL (shekaru 15-16) daga Hong Kong. Sun yi amfani da ChatGPT a kan iPads a karon farko don kammala aikin rubutu na bayyanawa: "Ka kwatanta wurin da kuka fi so kuma ka bayyana dalilin da yasa yake da mahimmanci a gare ka."
3.2 Tattara Bayanai
An tattara bayanai ta hanyar rikodin allo na iPad, wanda ya ɗauki kowane umarni da aka rubuta da kuma amsar ChatGPT. Masu bincike sun kuma yi hira bayan aikin don fahimtar tunanin ɗalibai.
3.3 Tsarin Nazari
Binciken ya rarraba umarni ta hanyar abun ciki (misali, neman ra'ayoyi, taimakon nahawu, bita) da yawa (yawan umarni kowane ɗalibi). Hanyoyi huɗu daban-daban sun bayyana daga bayanan.
4. Sakamako: Hanyoyi Hudu na Injiniyan Umarni
4.1 Hanyar A: Umarni Kai Tsaye
Dalibai sun ba da umarni guda ɗaya mai cikakken bayani (misali, "Rubuta sakin layi na kalmomi 200 game da bakin teku da na fi so, gami da cikakkun bayanai na ji"). Wannan hanya ta haifar da sakamako mai karɓuwa amma ta iyakance shigar ɗalibai cikin tsarin rubutu.
4.2 Hanyar B: Gyara Ta Hanyar Maimaitawa
Dalibai sun fara da umarni mai faɗi (misali, "Taimaka mini in rubuta game da wurin da na fi so") kuma suka gyara shi bisa sakamakon ChatGPT (misali, "Ƙara cikakkun bayanai game da sautin raƙuman ruwa"). Wannan hanya ta nuna koyo ta hanyar amsa.
4.3 Hanyar C: Rarraba Ayyuka Tare da Taimako
Dalibai sun raba aikin zuwa ƙananan ayyuka: da farko sun nemi tsari, sannan suka nemi ƙamus, kuma a ƙarshe suka nemi cikakken daftari. Wannan tsari mai tsari ya haifar da sakamako mafi inganci da zurfin fahimta.
4.4 Hanyar D: Gwaji da Kuskure na Bincike
Dalibai sun gwada umarni daban-daban ba tare da wata dabara ba (misali, "Ka ba ni ra'ayoyi", sannan "Ka mai da shi tsayi", sannan "Ka canza yanayin magana"). Wannan hanya ba ta da inganci kuma galibi tana haifar da takaici.
5. Tattaunawa
5.1 Babban Fahimta
Binciken ya nuna cewa yawancin ɗaliban EFL suna yin amfani da gwaji da kuskure a matsayin hanyar da ta saba, ba tare da dabaru na tsari ba. Kaɗan ne kawai (Hanyar C) suka nuna rarraba ayyuka yadda ya kamata, wanda ya yi daidai da ka'idodin taimakon metacognitive (Flavell, 1979).
5.2 Tsarin Ma'ana
Ci gaba daga Hanyar A zuwa D yana nuna nau'ikan ikon ɗalibi da zurfin dabara. Hanyar da ta fi inganci (C) tana kwaikwayon ayyukan injiniyan umarni na ƙwararru: rarraba ayyuka, gyara ta hanyar maimaitawa, da ƙayyadaddun mahallin.
5.3 Ƙarfi da Rashi
Ƙarfi: Binciken ya ba da bayanai masu inganci ta hanyar rikodin allo, yana ɗaukar halayen ɗalibai na gaskiya. Tsarin hanyoyi huɗu yana da sauƙin fahimta kuma yana da amfani ga malamai.
Rashi: Ƙananan adadin mahalarta (n=12) ya iyakance yiwuwar yin amfani da sakamakon gabaɗaya. Binciken bai auna ingancin rubutu ta hanyar ƙididdiga ba. Bugu da ƙari, tasirin sabon abu na amfani da ChatGPT a karon farko na iya canza halaye.
5.4 Shawarwari Masu Amfani
Ya kamata malamai su koyar da dabarun injiniyan umarni a fili, kamar:
- Rarraba ayyuka: Rarraba ayyukan rubutu masu rikitarwa zuwa ƙananan umarni.
- Gyara ta hanyar maimaitawa: Yi amfani da sakamakon ChatGPT a matsayin amsa don inganta umarni.
- Bayar da mahallin: Haɗa da matsayi, masu sauraro, da tsari a cikin umarni (misali, "Kai ɗan rubutun balaguro ne da ke rubuta wa matasa").
6. Cikakkun Bayanai na Fasaha da Tsarin Lissafi
Ana iya kallon injiniyan umarni a matsayin matsala ta ingantawa. Bari $P$ ya zama sararin umarni, $O$ sararin sakamako, kuma $f: P \rightarrow O$ aikin ChatGPT. Manufar ita ce samun $p^*$ kamar haka:
$$p^* = \arg\max_{p \in P} \, \text{Dacewa}(f(p), T)$$
inda $T$ shine aikin rubutu da aka yi niyya. Ana iya kimanta aikin dacewa ta hanyar kamanceceniyar cosine tsakanin haɗaɗɗen sakamako da haɗaɗɗen manufa a cikin sararin ma'ana (misali, Sentence-BERT). A aikace, ɗalibai suna sabunta $p$ akai-akai bisa ga $f(p)$ da aka lura:
$$p_{t+1} = p_t + \alpha \cdot \nabla \text{Maki}(f(p_t), T)$$
inda $\alpha$ shine ƙimar koyo kuma Maki shine ma'aunin inganci na heuristic. Wannan yana kwaikwayon hawan gradient a sararin latent, kodayake ɗalibai suna yin hakan ta hanyar ilhami.
7. Sakamakon Gwaji da Bayanin Hoton
Hoto na 1: Rarraba Hanyoyi
Ginshiƙi mai nuna yawan kowace hanya: Hanyar A (ɗalibai 3), Hanyar B (4), Hanyar C (2), Hanyar D (3). Ginshiƙin ya nuna cewa gyara ta hanyar maimaitawa (B) ya fi yawa, yayin da rarraba ayyuka tare da taimako (C) ya fi ƙaranci amma ya fi inganci.
Hoto na 2: Matsakaicin Yawan Umarni a kowace Hanya
Jadawalin layi: Hanyar A (umarni 1.0), B (4.5), C (6.0), D (8.3). Jadawalin ya nuna cewa yawan umarni ba lallai ba ne ya yi daidai da sakamako mafi kyau; Hanyar C ta yi amfani da ƙananan umarni fiye da D amma ta sami ingantaccen rubutu mafi girma (wanda malaman EFL biyu suka kimanta akan ma'auni 1-5: matsakaicin C 4.2, matsakaicin D 2.8).
8. Misalin Shari'a na Tsarin Nazari
Shari'a: Dalibi S7 (Hanyar C - Rarraba Ayyuka Tare da Taimako)
- Umarni na 1: "Ka ba ni tsari don sakin layi game da ɗakin karatu da na fi so. Haɗa da gabatarwa, cikakkun bayanai na ji, da dalilin da yasa yake na musamman."
- Sakamakon ChatGPT: Yana ba da tsari mai maki 3.
- Umarni na 2: "Fadada maki na 2 (cikakkun bayanai na ji) zuwa jimloli 3 ta amfani da kalmomi kamar 'raɗa', 'ƙura', 'dumi'."
- Sakamakon ChatGPT: Yana samar da jimloli na bayyanawa.
- Umarni na 3: "Haɗa tsarin da jimlolin zuwa sakin layi mai daidaituwa. Yi amfani da yanayin magana na hukuma."
- Sakamako na Ƙarshe: Sakin layi mai tsari mai kyau wanda ya sami maki 4.5/5.
Wannan shari'a tana nuna ingantaccen rarraba ayyuka da ƙayyadaddun mahallin.
9. Aikace-aikace da Hanyoyi na Gaba
Ya kamata bincike na gaba ya bincika:
- Koyarwar umarni ta atomatik: Kayan aikin AI waɗanda ke ba da amsa nan take game da ingancin umarni (misali, "Umarninka ba shi da cikakken bayani. Ka gwada ƙayyadaddun yanayin magana.")
- Injiniyan umarni na harsuna daban-daban: Yadda dabaru suka bambanta ga ɗaliban EFL da masu magana da harshen asali.
- Nazarin dogon lokaci: Bibiyar yadda ƙwarewar injiniyan umarni na ɗalibai ke canjawa cikin lokaci.
- Haɗin kai da manhajojin rubutu: Samar da tsarin darasi wanda ke koyar da injiniyan umarni tare da ƙwarewar rubutu na gargajiya.
10. Bincike na Asali
Wannan binciken ya ba da gudummawa mai dacewa ta hanyar yin taswirar yadda ɗaliban EFL masu sababbi suke mu'amala da ChatGPT, yana bayyana gibin da ke tsakanin gwaji da kuskure na ilhami da injiniyan umarni na dabara. Tsarin hanyoyi huɗu kayan aiki ne mai mahimmanci na koyarwa, amma ƙananan adadin mahalarta da rashin sarrafa bayyanar AI na baya sun iyakance yiwuwar yin amfani da shi gabaɗaya. Sakamakon cewa rarraba ayyuka tare da taimako (Hanyar C) yana haifar da sakamako mafi kyau ya yi daidai da ka'idar nauyin fahimi (Sweller, 1988), wanda ke nuna cewa rarraba ayyuka masu rikitarwa zuwa ƙananan sassa yana rage nauyin fahimi kuma yana haɓaka koyo. Duk da haka, binciken bai magance bangaren ɗabi'a ba: ɗaliban da suka dogara ga ChatGPT don samar da ra'ayoyi na iya yin saɓo ba da gangan ba ko kuma su rasa muryar kansu. Aikin gaba ya kamata ya haɗa horar da ɗabi'ar dijital a cikin manhajojin injiniyan umarni. Bugu da ƙari, tsarin lissafi na inganta umarni (Sashe na 6) yana ba da hangen nesa mai zurfi, amma amfaninsa a aikace a cikin ajujuwa bai tabbata ba. Don ci gaba, dole ne malamai su ɗauki injiniyan umarni a matsayin ƙwarewar karatu ta asali, kamar ƙwarewar binciken injin bincike (Head & Eisenberg, 2010). Sai a lokacin ne ɗalibai za su iya amfani da AI a matsayin abokin aiki maimakon abin dogaro.
11. Manazarta
- Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area of cognitive–developmental inquiry. American Psychologist, 34(10), 906–911.
- Guo, K., Woo, D. J., & Susanto, H. (2023). Exploring EFL students' prompt engineering strategies with ChatGPT. Computers & Education: Artificial Intelligence, 5, 100156.
- Head, A. J., & Eisenberg, M. B. (2010). How today's college students use the Web for research. Project Information Literacy Progress Report.
- Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285.
- Woo, D. J., Guo, K., & Susanto, H. (2023). Cases of EFL secondary students' prompt engineering pathways to complete a writing task with ChatGPT. Journal of Educational Computing Research, 61(4), 789–812.