Nazarin Shari'a: Hanyoyin Injiniyanci na Tambayoyi (Prompt Engineering) na Daliban Sakandare EFL tare da ChatGPT don Ayyukan Rubutu

1. Gabatarwa

Zuwan na'urorin AI masu samar da maganganu na zamani (SOTA) kamar ChatGPT yana kawo dama da kalubale ga ilimi, musamman a cikin koyon harshe. Wannan takarda tana binciken yadda daliban makarantar sakandare na Turanci a matsayin Harshen Waje (EFL), a matsayin masu amfani na farko, suke shiga cikin injiniyanci na tambayoyi (prompt engineering)—ƙwarewar ƙirƙirar umarni don AI—don kammala aikin rubutu. Babbar matsalar ita ce haɗin gwiwa mai inganci tare da ChatGPT ba ta da ma'ana ta halitta; yana buƙatar ƙwarewar da aka koya wadda yawancin ɗalibai ba su da ita, wanda ke haifar da hanyoyin gwaji da kuskure marasa inganci. Wannan binciken yana nufin zayyana hanyoyi daban-daban da ɗalibai ke bi, tare da nazarin abubuwan da ke ciki, inganci, da ci gaban tambayoyinsu don ba da labari game da dabarun koyarwa don haɗa ilimin AI cikin ajin rubutu na EFL.

2. Hanyar Bincike

Wannan binciken yana amfani da hanyar nazarin shari'a mai inganci. An tattara bayanai daga rikodin allo na iPad na ɗaliban sakandare EFL a Hong Kong suna amfani da ChatGPT da makamantansu na SOTA chatbots a karon farko don kammala daidaitaccen aikin rubutu. Binciken ya mayar da hankali kan cikakken bincike na tambayoyin da ɗalibai suka samar, jerinsu (hanyoyi), da kuma sakamakon AI da ya dace. Binciken ya gano hanyoyi huɗu daban-daban na asali bisa tsarin mu'amala, ƙwarewar tambayoyi, da dabarun tsarin aiki.

3. Nazarin Shari'a: Hanyoyi Hudu na Injiniyanci na Tambayoyi

Binciken ya bayyana manyan tsare-tsare huɗu na mu'amala, wanda ke wakiltar matakan shiga da tunani na dabarun daban-daban.

3.1. Hanya A: Mai Ragewa (The Minimalist)

Daliban da ke cikin wannan hanya sun yi amfani da ƴan tambayoyi, sau da yawa ba su da ma'ana (misali, "Ka rubuta makala game da gurɓata yanayi"). Sun nuna ƙarancin shiga cikin tunani (metacognitive engagement), suna karɓar sakamakon farko na AI tare da ƙaramin bita ko ƙayyadaddun bayani. Wannan hanya tana nuna ƙarancin fahimtar iyawar AI da kuma buƙatar takamaiman umarni.

3.2. Hanya B: Mai Gyara A Jere (The Iterative Refiner)

Waɗannan ɗalibai sun fara da tambaya ta asali amma sun shiga cikin tsarin gyara a jere. Dangane da sakamakon farko na AI, sun ba da umarni na biyo baya kamar "ka ƙara tsawonsa," "ka yi amfani da kalmomi masu sauƙi," ko "ka ƙara misali." Wannan hanya tana nuna fahimtar da ke tasowa game da yanayin mu'amala da maimaitawa na haɗin gwiwar mutum da AI.

3.3. Hanya C: Mai Tsara Tsari (The Structured Planner)

Hanya mafi ci gaba inda ɗalibai suka yi ƙoƙarin tsara aikin don AI tun daga farko. Tambayoyin sun haɗa da abubuwa kamar wasan kwaikwayo ("Kai malami ne na rubutu"), umarni mataki-mataki ("Da farko, ka ba ni ra'ayoyi uku. Sa'an nan, ka zayyana ra'ayin farko"), da ƙayyadaddun iyakoki ("Ka rubuta kalmomi 150 ta amfani da lokacin da ya wuce"). Wannan hanyar tana nuna tsare-tsare na dabarun da kuma bayyanannen samfurin yadda ake "shirya" AI ta hanyar harshe.

3.4. Hanya D: Mai Bincike da Gwaji (The Exploratory Tester)

Waɗannan ɗalibai sun yi amfani da yawan tambayoyi iri-iri, sau da yawa na gwaji. Sun gwada iyakokin AI tare da buƙatun ƙirƙira, waɗanda ba su da alaƙa da batu, ko masu rikitarwa don fahimtar ayyukansa kafin su yi amfani da shi ga ainihin aikin. Wannan hanya tana nuna tunanin bincike, mai ƙwarewar fasaha amma ba koyaushe yana haifar da ingantacciyar manufar aikin ba.

4. Sakamako & Bincike

4.1. Tsarin Ingancin da Adadin Tambayoyi

An lura da alaƙa bayyananne tsakanin ƙwarewar tambaya da ingancin sakamako na ƙarshe. Hanya C (Mai Tsara Tsari) a koyaushe tana haifar da rubutun da suka fi dacewa, masu dacewa da aikin, kuma masu wadataccen harshe. Sakamakon Hanya A (Mai Ragewa) ya kasance na gaba ɗaya kuma sau da yawa ba su da manufa. Adadin tambayoyi kawai (mai yawa a Hanya D) bai tabbatar da inganci ba; ingancin dabarun (Hanya C) shine babban abin banbancewa.

Taƙaitaccen Mu'amalar Tambaya

Hanya A (Mai Ragewa): Matsakaicin tambayoyi 2-3; Ƙarancin takamaiman bayani.
Hanya B (Mai Gyara A Jere): Matsakaicin tambayoyi 5-8; Gyara na mayar da martani.
Hanya C (Mai Tsara Tsari): Matsakaicin tambayoyi 4-6; Babban tsari kafin aiki.
Hanya D (Mai Bincike da Gwaji): Matsakaicin tambayoyi 10+; Babban bambancin, gaurayawan dacewa.

4.2. Tasiri akan Abubuwan da aka Rubuta

Samfuran rubutu na ƙarshe sun bambanta sosai. Tambayoyin da aka tsara sun haifar da sakamako waɗanda suka fi dacewa da buƙatun aikin, sun yi amfani da ƙamus mafi dacewa, kuma sun nuna tsari mafi bayyananne. Tambayoyin masu ragewa sun haifar da rubutun da, ko da yake daidai ne a nahawu, ba su da zurfi da keɓancewa, suna kama da abubuwan gaba ɗaya na yanar gizo.

5. Tattaunawa: Abubuwan da suka shafi Ilimin AI

Binciken ya jaddada cewa yin amfani da ChatGPT yadda ya kamata ƙwarewa ce da ake koyarwa, ba iyawa ta halitta ba. Yaɗuwar hanyoyin ragewa da marasa inganci a jere tsakanin masu fara aiki yana nuna babban gibi a cikin ilimin na yanzu. Marubutan suna jayayya don bayyanannen ilimin injiniyanci na tambayoyi (prompt engineering) don haɗa shi cikin manhajojin EFL. Wannan zai motsa ɗalibai fiye da gwaji da kuskure, yana ba su tsare-tsare don ƙirƙirar umarni bayyananne, sanya ayyuka, ƙayyade tsari, da gyara sakamako a jere—canza AI daga akwatin baƙar fata (black-box oracle) zuwa kayan aikin haɗin gwiwa.

Mahimman Fahimta

Injiniyanci na tambayoyi wani sabon nau'i ne na ilimin dijital wanda ya zama dole don zamanin AI.
Hanyoyin ɗalibai zuwa AI sun bambanta, suna buƙatar koyarwa daban-daban.
Ingancin umarni (tambaya) kai tsaye yana ƙayyade ingancin sakamakon taimakon AI.
Ba tare da jagora ba, ɗalibai suna fuskantar haɗarin haɓaka halayen mu'amala marasa aiki ko marasa inganci tare da AI.

6. Tsarin Fasaha & Bincike

Daga mahangar fasaha, injiniyanci na tambayoyi yana mu'amala tare da ayyukan yiwuwar samfurin harshe na asali. Tambaya da aka ƙera da kyau $P$ tana jagorantar samfurin $M$ don ɗaukar samfuri daga yanki mafi ƙuntatawa kuma mafi kyawu na rarraba sakamakonsa $D$ don wani mahallin $C$. Ana iya wakiltar tsarin a matsayin haɓaka yiwuwar yanayin da ake so na jerin sakamako $O$:

$O^* = \arg\max_{O} P(O | C, P, M)$

Inda tambaya mara ma'ana tana ƙara ruɗani (entropy) a cikin $D$, yana haifar da sakamako na gaba ɗaya, takamaiman tambaya tare da ƙuntatawa (rawa, tsari, salo) yana rage ruɗani, yana tuƙi $M$ zuwa ga $O^*$ mafi manufa. Hanyoyin ɗalibai da kyau suna wakiltar dabarun daban-daban don sarrafa wannan yiwuwar yanayin ta hanyar umarnin harshe na halitta.

Misalin Tsarin Bincike

Yanayi: Dalibi yana son ChatGPT ya taimaka masa ya rubuta sakin layi mai gamsarwa game da sake yin amfani da kaya (recycling).

Tambaya Mai Rauni (Babban Ruɗani): "Ka rubuta game da sake yin amfani da kaya."
Bincike: Samfurin yana da ƙaramin ƙuntatawa, mai yiwuwa yana haifar da bayani mai faɗi, irin na encyclopedia.
Tambaya Mai Ƙarfi (Ƙaramin Ruɗani): "Ka yi aiki a matsayin mai ba da shawara kan muhalli. Ka rubuta sakin layi mai gamsarwa na kalmomi 80 da aka yi niyya ga matasa, ka gamsar da su su sake yin amfani da kwalabe na filastik. Ka yi amfani da sautin kai tsaye da gaggawa, kuma ka haɗa da ƙididdiga ɗaya."
Bincike: Wannan tambaya ta ƙayyade rawa (mai ba da shawara), masu sauraro (matasa), manufa (gamsarwa), mai da hankali kan abun ciki (kwalaben filastik), tsawon (kalmomi 80), sauti (kai tsaye, gaggawa), da abu (ƙididdiga). Yana rage rarraba sakamakon samfurin sosai.

7. Ayyuka na Gaba & Hanyoyin Bincike

Binciken ya buɗe hanyoyi da yawa don aikin gaba:

Malamai na Tambayoyi Masu Daidaitawa: Haɓaka malamai masu amfani da AI waɗanda ke bincika tambayar ɗalibi kuma suna ba da ra'ayi na ainihi kan yadda ake inganta shi (misali, "Ka gwada ƙayyadaddun masu sauraronka").
Nazarin Tsawon Lokaci: Bin diddigin yadda ƙwarewar injiniyanci na tambayoyi na ɗalibai ke tasowa akan lokaci tare da ko ba tare da koyarwa na yau da kullun ba.
Kwatancen Al'adu & Harshe: Binciken ko dabarun injiniyanci na tambayoyi sun bambanta a cikin harsuna da mahallin ilimi na al'adu.
Haɗawa da Ilimin Rubutu: Bincike kan yadda za a iya saɓa tsare-tsaren injiniyanci na tambayoyi cikin samfuran tsarin rubutu na yanzu (kafin rubutu, daftarin aiki, bita).
Girma na Da'a & Mahimmanci: Faɗaɗa ilimin AI fiye da inganci don haɗa da kimanta sakamakon AI, gano son zuciya, da amfani mai da'a.

8. Nassoshi

Woo, D. J., Guo, K., & Susanto, H. (2023). Cases of EFL Secondary Students’ Prompt Engineering Pathways to Complete a Writing Task with ChatGPT. Manuscript in preparation.
Caldarini, G., Jaf, S., & McGarry, K. (2022). A Literature Survey of Recent Advances in Chatbots. Information, 13(1), 41.
Long, D., & Magerko, B. (2020). What is AI Literacy? Competencies and Design Considerations. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1–16.
OpenAI. (2023). GPT-4 Technical Report. arXiv preprint arXiv:2303.08774.
Zhao, W. X., et al. (2023). A Survey of Large Language Models. arXiv preprint arXiv:2303.18223.
The Stanford Center for AI Safety. (n.d.). AI Literacy. Retrieved from https://aisafety.stanford.edu/ai-literacy

Ra'ayin Mai Bincike: Rarraba Muhimmancin Injiniyanci na Tambayoyi

Babban Fahimta: Wannan binciken ba game da ɗalibai da ChatGPT kawai ba ne; yana da alaƙa da ƙalubalen mu'amalar ɗan adam da AI na asali a zamanin bayan ChatGPT. Babban fahimtar shi ne cewa "tambaya (prompting)" shine sabon shirye-shirye (programming). Hanyoyi huɗu (Mai Ragewa, Mai Gyara A Jere, Mai Tsara Tsari, Mai Bincike da Gwaji) ba kawai salon koyo ba ne; su ne samfuran nau'ikan masu amfani waɗanda za su ayyana gibin samar da aiki da ƙirƙira a cikin ma'aikatan da AI ta ƙarfafa. Takardar ta gano daidai cewa ba tare da ilimi mai tsari ba, yawancin masu amfani za su koma ga hanyoyin ragewa marasa inganci ko gwaji da kuskure a jere, suna barin babban yuwuwar kayan aiki kamar GPT-4, kamar yadda aka yi cikakken bayani a cikin rahotonsa na fasaha, ba a yi amfani da su ba.

Kwararar Hankali & Ƙarfafawa: Ƙarfin takardar yana cikin hanyarta ta gaskiya, ta zahiri. Ta yin amfani da rikodin allo, ta kama gwagwarmayar ɗan fari, wanda ba a tace shi ba. Wannan yana motsa tattaunawar fiye da tsare-tsaren ka'idojin ilimin AI (kamar na Long & Magerko) zuwa aikin lura. Gano Mai Tsara Tsari a matsayin hanya mai inganci yana da mahimmanci. Yana tabbatar da hasashen masana'antu cewa ingantacciyar tambaya tana kama da takardar ƙayyadewa—bayyananne, ƙuntatacce, kuma an daidaita shi da mahallin. Wannan ya yi daidai da bincike kan yadda manyan samfuran harshe (LLMs) ke aiki a matsayin "aku masu yiwuwar bazuwa (stochastic parrots)" waɗanda ke jagorantar ta hanyar rarraba yiwuwar yanayi; takamaiman tambaya ta lissafi tana rage sararin samfurin, kamar yadda aka tattauna a cikin cikakkun bincike kamar na Zhao et al.

Kurakurai & Makaho: Babban aibin binciken shine iyakokin sa—aiki ɗaya tare da masu amfani na farko. Ba ya nuna ko Mai Bincike da Gwaji, wanda ake iya cewa yana nuna mafi girman sha'awar ciki da binciken tsarin, zai iya zama mafi ƙwarewar mai amfani akan lokaci. Bugu da ƙari, ya kauce wa mahimmin girma na da'a da ilimin mahimmanci. Dalibi na iya zama Mai Tsara Tsari mai hazaka, yana samar da makala mai gamsarwa cikakkiya tare da ChatGPT, amma ya kasance ba shi da wani ra'ayi game da son zuciya, kuskuren gaskiya, ko rashin tunani na asali a cikin sakamakon. Kamar yadda cibiyoyi kamar Cibiyar Stanford don Tsaron AI suka jaddada, ainihin ilimin AI dole ne ya ƙunshi kimantawa, ba kawai samarwa ba.

Fahimta Mai Aiki: Ga malamai da masu tsara manufofi, abin da za a ɗauka ba shi da sasantawa: Injiniyanci na tambayoyi dole ne ya zama ainihin abin da aka tantance na manhajojin ilimin dijital, farawa yanzu. Wannan ba zaɓi ba ne. Binciken ya ba da tsari: motsa ɗalibai daga zama masu cin sakamakon AI marasa aiki (Mai Ragewa) zuwa masu gudanarwa masu dabarun aiki (Mai Tsara Tsari). Tsare-tsaren darasi yakamata su koyar da tsare-tsaren tambayoyi a fili—rawa, masu sauraro, tsari, sauti, misalai (RAFTE). Ga masu haɓaka fasaha, fahimtar ita ce gina "tsarin ginin tambaya (prompt scaffolding)" kai tsaye cikin mu'amalar ilimi—samfuran mu'amala, injunan ba da shawara, da tambayoyin tunani (metacognitive prompts) waɗanda ke tambayar masu amfani, "Shin kun yi la'akari da ƙayyadaddun...?" Gaba ba na waɗanda za su iya amfani da AI ba ne, amma ga waɗanda za su iya ba da umarninta da daidaito da mahimmanci.