Teburin Abubuwan Ciki
- 1.1 Gabatarwa & Bayyani
- 1.2 Bayanin Matsala ta Asali
- 2. Dandamalin CHOP
- 3. Hanyoyin Bincike & Kimantawa
- 4. Sakamako & Binciken
- 5. Cikakkun Bayanai na Fasaha & Tsari
- 6. Tattaunawa & Abubuwan da ke Tattare
- 7. Aikace-aikace na Gaba & Jagorori
- 8. Nassoshi
1.1 Gabatarwa & Bayyani
Wannan takarda tana ba da cikakken bincike kan takardar bincike "CHOP: Haɗa ChatGPT cikin Aikin Gabatar da Baki na EFL." Binciken ya magance wani gibi mai mahimmanci a cikin ilimin Turanci a matsayin Harshen Waje (EFL): rashin samun ra'ayi na musamman mai yawa don ƙwarewar gabatar da baki. Ya gabatar da CHOP (Dandamali mai mu'amala da ChatGPT don aikin gabatar da baki), wani sabon tsari da aka ƙera don ba da ra'ayi na lokaci-lokaci, taimakon AI ga ɗalibai.
1.2 Bayanin Matsala ta Asali
Daliban EFL suna fuskantar ƙalubale masu mahimmanci wajen haɓaka ƙwarewar gabatar da baki, gami da damuwa na magana, ƙarancin ƙamus/ɗabi'a, da kuskuren lafazi. Hanyoyin gargajiya masu dogaro da malami sau da yawa ba su isa ba saboda ƙarancin albarkatu da rashin iya ba da ra'ayi na gaggawa, na mutum ɗaya. Wannan yana haifar da buƙatar hanyoyin magancewa na fasaha masu mu'amala, masu dogaro da ɗalibi.
2. Dandamalin CHOP
2.1 Ƙira na Tsari & Tsarin Aiki
An gina CHOP a matsayin dandamali na yanar gizo inda ɗalibai ke yin aikin gabatar da baki. Tsarin aiki na asali ya ƙunshi: 1) ɗalibi yana rikodin wasan kwaikwayonsa na gabatarwa, yana iya zagayawa cikin nunin faifai idan ya so. 2) Ana rubuta sautin. 3) ɗalibi yana neman ra'ayi daga ChatGPT bisa ga ƙa'idodin da aka ƙayyade (misali, abun ciki, harshe, isarwa). 4) ChatGPT yana samar da ra'ayi na musamman, wanda ɗalibi zai iya kimantawa da amfani da shi don yin tambayoyin biyo baya don bita.
2.2 Muhimman Siffofi & Fuskar Mai Amfani
Kamar yadda aka nuna a Hoto na 1 na PDF, fuskar ta ƙunshi: (A) Kewayawa cikin nunin faifai don aikin sashi, (B) Maimaita sautin wasan kwaikwayo, (C) Nuna ra'ayin ChatGPT bisa ga kowane ma'auni tare da rubutun, (D) Ma'auni mai maki 7 na Likert don kimanta kowane abu na ra'ayi, (E) Sashen bayanin kula don bita, da (F) Fuskar hira don tambayoyin biyo baya ga ChatGPT.
3. Hanyoyin Bincike & Kimantawa
3.1 Bayanin Mahalarta & Ƙirar Bincike
Binciken ya yi amfani da hanyar haɗaɗɗun hanyoyi. An gudanar da hira ta farko tare da ɗaliban EFL 5 don fahimtar buƙatu. Babban kimantawar dandamali ya ƙunshi ɗaliban EFL 13. Ƙirar binciken ta mayar da hankali kan tattara ɗimbin bayanai na inganci da ƙididdiga game da mu'amala tsakanin ɗalibi da AI.
3.2 Tattara Bayanai & Tsarin Bincike
An yi amfani da tushen bayanai guda uku na farko: 1) Rajistan Mu'amala: Duk mu'amalar ɗalibi-ChatGPT, gami da buƙatun ra'ayi, kimantawa, da tambayoyin biyo baya. 2) Binciken Bayan: Ra'ayin ɗalibai game da amfani, gamsuwa, da ƙalubale. 3) Kimantawar Ƙwararru: Ƙwararrun malaman harshe sun kimanta ingancin samfurin ra'ayin da ChatGPT ya samar bisa ga ƙa'idodin da aka kafa.
4. Sakamako & Binciken
4.1 Kimanta Ingancin Ra'ayi
Kimantawar ƙwararru ta nuna cewa ra'ayin da ChatGPT ya samar gabaɗaya yana da alaƙa kuma ana iya aiwatar da shi don abubuwan da suka shafi babban mataki kamar tsarin abun ciki da bayyanawa. Duk da haka, ya nuna iyakoki wajen ba da shawara ta musamman, ta takamaiman yanayi game da lafazi, sautin murya, da amfani da harshe mai zurfi. Daidaiton ya dogara ne da ingancin umarnin ɗalibi na farko da rubutun sauti.
4.2 Ra'ayin ɗalibai & Tsarin Mu'amala
Dalibai sun ba da rahoton raguwar damuwa saboda yanayin malamin AI wanda ba ya yin hukunci, kuma yana samuwa koyaushe. Tsarin kimantawa mai maki 7 ya ba da bayanai masu mahimmanci game da amfanin ra'ayin da ake ganin yana da amfani. Rajistan mu'amala ta nuna cewa ɗaliban da suka shiga cikin zagayowar buƙatar ra'ayi → bita → tambayar biyo baya sun nuna ci gaba mai mahimmanci da suka ruwaito kansu. Wani muhimmin binciken shine mahimmancin abubuwan ƙira kamar bayyananniyar ƙa'idodin ra'ayi da sauƙin fuskar tambayar biyo baya wajen tsara ƙwarewar koyo.
5. Cikakkun Bayanai na Fasaha & Tsari
5.1 Injiniyan Umarni & Samar da Ra'ayi
Ingancin tsarin ya dogara da ingantaccen injiniyan umarni. Babban umarni da aka aika zuwa API na ChatGPT ana iya wakilta shi a matsayin aiki: $F_{ra'ayi} = P(Rubutun, Ma'auni, Yanayi)$, inda $P$ shine samfurin umarni, $Rubutun$ shine fitowar ASR, $Ma'auni$ su ne ma'auni na kimantawa (misali, "kimanta santsi da haɗin kai"), kuma $Yanayi$ ya haɗa da matakin ɗalibi da manufar gabatarwa. Samar da ra'ayi ba rarrabuwa mai sauƙi ba ne amma aikin samar da rubutu mai sharadi wanda aka inganta don amfanin koyarwa.
5.2 Misalin Tsarin Bincike
Hali: Bincika Tasirin Ra'ayi
Yanayi: ɗalibi ya karɓi ra'ayi: "Bayyani na hanyar da kuka yi ya bayyana a sarari, amma gwada amfani da ƙarin kalmomin haɗi kamar 'bugu da ƙari' ko 'sabanin haka'."
Aiwatar da Tsari:
1. Ƙarfi: Shin ra'ayin yana da takamaiman (yana nufin "kalmomin haɗi") ko maras tabbas?
2. Aiwatarwa: Shin yana ba da misali na zahiri ("bugu da ƙari")?
3. Ƙarfafawa Mai Kyau: Shin ya fara da ƙarfi ("bayyani mai haske")?
4. Yuwuwar Biyo Baya: Shin ɗalibi zai iya tambaya a zahiri: "Za ka iya ba ni ƙarin misalai biyu na kalmomin haɗi don kwatanta ra'ayoyi?"
Wannan tsarin, da aka yi amfani da shi a cikin rajistan mu'amala, yana taimakawa gano waɗanne tsarin umarni ke haifar da mafi inganci $F_{ra'ayi}$.
6. Tattaunawa & Abubuwan da ke Tattare
6.1 Ƙarfafawa, Iyakoki & Abubuwan Ƙira
Ƙarfafawa: CHOP yana nuna yuwuwar haɓakawa, samuwa 24/7, da keɓancewa a matakin da ya yi wahala ga malaman ɗan adam su yi daidai akai-akai. Yana haɓaka yanayin aiki mara nauyi.
Iyakoki & Kurakurai: Yanayin "akwatin baƙi" na samar da ra'ayi na iya haifar da kurakurai, musamman a cikin ilimin sauti. Ba shi da jagorar tausayi da ƙwarewar al'adu na ƙwararren ɗan adam. Dogaro da yawa na iya hana haɓaka ƙwarewar tantance kai.
Muhimman Abubuwan Ƙira: Binciken ya nuna cewa UI dole ne ya jagoranci ɗalibi don yin tambayoyi mafi kyau (misali, shawarwarin umarni na biyo baya), kuma dole ne a raba ra'ayi zuwa guntu masu sauƙin fahimta, na takamaiman ma'auni don guje wa ɗalibi.
6.2 Bincike na Asali: Fahimta ta Asali, Tsarin Ma'ana, Ƙarfafawa & Kurakurai, Fahimta masu Aiki
Fahimta ta Asali: Binciken CHOP ba kawai game da gina wani malamin AI ba ne; yana da alaƙa da binciken farko na haɗa kai na haɗin gwiwar ɗan adam-AI don ƙwarewa mai rikitarwa, mai dogaro da aiki. Ainihin sabon abu shine tsarin aiki mai tsari wanda ya sanya ChatGPT ba a matsayin maye gurbin malami ba, amma a matsayin abokin wasan kwaikwayo mara gajiyawa wanda ke shirya ɗalibi don ƙarshe, babban aji na ɗan adam. Wannan ya yi daidai da hangen nesa na haɗin gwiwar ɗan Adam-AI a cikin ilimi wanda masu bincike a Cibiyar HAI ta Stanford suka zayyana, inda AI ke kula da aikin maimaitawa da ra'ayi mai dogaro da bayanai, yana 'yantar da malamai don jagoranci mafi girma.
Tsarin Ma'ana: Ma'anar takardar tana da ƙarfi: gano wurin zafi mai dorewa, mai yawan albarkatu (ra'ayi na musamman na gabatarwa) → amfani da fasaha mai rushewa, mai amfani gabaɗaya (LLMs) → ƙira takamaiman yanayin aikace-aikace tare da kariya (dandamalin CHOP) → tabbatarwa ta hanyar bincike na haɗaɗɗun hanyoyi. Wannan shine tsarin bincike mai tasiri na EdTech.
Ƙarfafawa & Kurakurai: Ƙarfinsa shine mayar da hankali kan ƙira ta haɗawa da fahimtar ɗalibi, wucewa fiye da binciken yuwuwar kawai. Duk da haka, babban aibin binciken shine girmansa (n=13). Duk da cewa fahimta ta inganci tana da wadata, ba shi da ƙarfin ƙididdiga don yin iƙirari na tabbaci game da ingancin koyo, matsala ta gama gari a cikin aikin HCI na farko don ilimi. Kwatanta maki na gabatarwa kafin da bayan gwaji tare da ƙungiyar kulawa, kamar yadda aka gani a cikin ƙarin bincike mai tsauri kamar na tsarin koyarwa mai hankali don lissafi (misali, binciken Carnegie Learning), da ya ƙarfafa da'awarsa.
Fahimta masu Aiki: Ga malamai da manajoji samfur, abin da za a ɗauka ya bayyana a sarari: Tsarin nasara shine "AI don aiki, ɗan adam don hukunci." Kada ku yi ƙoƙarin gina AI wanda ke kimanta gabatarwar ƙarshe. Maimakon haka, ku gina AI wanda ke haɓaka ingancin aiki, yana tabbatar da cewa ɗalibai sun isa ga mai tantancewa na ɗan adam sun fi gogewa da kwarin gwiwa. Juzu'i na gaba na CHOP ya kamata ya haɗa bincike mai yawa (misali, amfani da samfuran hangen nesa don ra'ayi na matsayi da ishara, kama da aikace-aikace a cikin nazarin wasanni) kuma ya ɗauki ingantaccen tsarin kimantawa mai dogaro da ka'ida wanda ke auna ba kawai gamsuwa ba, amma canja wurin ƙwarewa na zahiri.
7. Aikace-aikace na Gaba & Jagorori
Tsarin CHOP yana da babban yuwuwar faɗaɗawa:
1. Ra'ayi Mai Yawa: Haɗa hangen nesa na kwamfuta (misali, OpenPose) don bincika yanayin jiki, kallon ido, da ishara, samar da cikakken ra'ayi na isarwa.
2. Daidaituwa na Musamman na Yanki: Daidaita dandamali don takamaiman fagage (misali, gabatarwar kimiyya, gabatarwar kasuwanci) ta hanyar daidaita ƙaramin LLM akan ƙungiyoyin da suka dace.
3. Nazarin Koyo na Tsawon Lokaci: Yin amfani da bayanan mu'amala don gina samfuran ɗalibai waɗanda ke hasashen wuraren gwagwarmaya da ba da shawarar ayyukan da aka yi niyya, motsawa daga tallafi na mayar da martani zuwa na gaggawa.
4. Haɗaɗɗun Haɗin Aji: Haɓaka dashboard na malami inda malamai za su iya duba taƙaitaccen ra'ayin AI na kowane ɗalibi, yana ba da damar shiga tsakani cikin aji cikin inganci da sanin abin da ake yi. Wannan "gauraye" samfurin yana wakiltar makomar ilimin da AI ya ƙarfafa.
8. Nassoshi
- Cha, J., Han, J., Yoo, H., & Oh, A. (2024). CHOP: Haɗa ChatGPT cikin Aikin Gabatar da Baki na EFL. arXiv preprint arXiv:2407.07393.
- Hwang, G.-J., Xie, H., Wah, B. W., & Gašević, D. (2020). Hangen nesa, ƙalubale, ayyuka da batutuwan bincike na Artificial Intelligence a cikin Ilimi. Kwamfuta da Ilimi: Artificial Intelligence, 1, 100001.
- Cibiyar Stanford don Artificial Intelligence Mai Maida Hankali ga Mutum (HAI). (2023). AI da Ilimi: Gaskiya da Yuwuwar. An samo daga https://hai.stanford.edu
- Zhu, J.-Y., Park, T., Isola, P., & Efros, A. A. (2017). Fassarar Hotuna zuwa Hotuna mara Haɗin gwiwa ta amfani da Cibiyoyin Adawa na Zagaye-Ma'ana. Proceedings of the IEEE International Conference on Computer Vision (ICCV). (An ambata a matsayin misali na ingantacciyar hanyar bincike mai tasiri a cikin binciken AI).
- Koedinger, K. R., & Aleven, V. (2016). Malamin Hankali mara Tsangwama don Amfani da Dabarun Metacognitive. International Conference on Intelligent Tutoring Systems. (Misali na ingantaccen kimantawa a cikin AI na ilimi).
- Majalisar Turai. (2001). Tsarin Ma'anar Turai na Turai don Harsuna: Koyo, koyarwa, tantancewa. Cambridge University Press. (Ingantaccen tsari don ƙwarewar harshe).