Behruzbek HamidovAxloq – bu xulq-atvorda nima to’g’ri va nima noto’g’ri ekanligini belgilovchi, axloqiy tanlovlarni boshqaradigan tamoyillar to’plami. Kichik miqyosdagi jamoalar boshqa shaxslarning axloqini baholash uchun shaxsiy tajriba va suhbatga tayanishi mumkin, ammo jamiyatlar o’sib borishi bilan axloqni baholashning an’anaviy usullari keng miqyosda amalga oshirish uchun etarli bo’lmaydi.
https://googleads.g.doubleclick.net/pagead/ads?client=ca-pub-9585941727679583&output=html&h=250&slotname=5494213682&adk=315283653&adf=995531428&pi=t.ma~as.5494213682&w=300&lmt=1696938823&format=300×250&url=https%3A%2F%2Fwww.psypost.org%2F2023%2F10%2Fpeople-believe-their-moral-traits-are-too-distinct-for-ai-to-judge-study-finds-213968&host=ca-host-pub-2644536267352236&wgl=1&uach=WyJXaW5kb3dzIiwiMTAuMC4wIiwieDg2IiwiIiwiMTE3LjAuNTkzOC4xNTAiLFtdLDAsbnVsbCwiNjQiLFtbIkdvb2dsZSBDaHJvbWUiLCIxMTcuMC41OTM4LjE1MCJdLFsiTm90O0E9QnJhbmQiLCI4LjAuMC4wIl0sWyJDaHJvbWl1bSIsIjExNy4wLjU5MzguMTUwIl1dLDBd&dt=1696957473881&bpp=2&bdt=360&idt=265&shv=r20231004&mjsv=m202310030101&ptt=9&saldr=aa&abxe=1&cookie=ID%3D2b196ed822638318-228f449b45df0062%3AT%3D1696957015%3ART%3D1696957015%3AS%3DALNI_MZhsJb4E4109alSlPl2S7-kOf8cRg&gpic=UID%3D00000d9869619fc4%3AT%3D1696957015%3ART%3D1696957015%3AS%3DALNI_MYdoqiWzzDM6kkKYz11EKq0D6313A&prev_fmts=750×280&correlator=4425782298123&frm=20&pv=1&ga_vid=120765468.1696957474&ga_sid=1696957474&ga_hid=772420914&ga_fc=0&u_tz=300&u_his=4&u_h=864&u_w=1536&u_ah=824&u_aw=1536&u_cd=24&u_sd=1.25&dmc=8&adx=452&ady=1409&biw=1505&bih=790&scr_x=0&scr_y=0&eid=44759876%2C44759927%2C44759837%2C31078593%2C31078657%2C44804782%2C44805099%2C31078673%2C31078663%2C31078665%2C31078668%2C31078670&oid=2&pvsid=1471702984887259&tmod=435951133&uas=0&nvt=1&ref=https%3A%2F%2Fwww.psypost.org%2F&fc=896&brdim=30%2C26%2C30%2C26%2C1536%2C0%2C1536%2C882%2C1522%2C790&vis=1&rsz=%7C%7CeEbr%7C&abl=CS&pfx=0&fu=0&bc=31&td=1&nt=1&ifi=2&uci=a!2&btvi=1&fsb=1&xpc=edt9j7EI0O&p=https%3A//www.psypost.org&dtd=273
Aniq axloqiy va ijtimoiy muammolar tufayli, inson axloqini kuzatish, baholash va baholash uchun sun’iy intellektdan foydalanishga umumiy munosabatimizni tushunishga ehtiyoj bor.
Frantsiyadagi Tuluza universitetidan tadqiqot mualliflari Zoe Pursel va Jan-Fransua Bonnefon: “Odamlarning mashinalarning axloqini baholashi bo’yicha ko’p ish olib borilgan bo’lsa-da, biz odamlarning axloqini hukm qiladigan mashinalar haqidagi yangi savolga e’tibor qaratamiz”. .
Buni tekshirish uchun to’rtta kichik tadqiqot o’tkazildi.
1-tadqiqotda Buyuk Britaniyadagi 446 ishtirokchi ishtirok etdi, ular sodiqlik, halollik, tajovuzkorlik yoki irqchilik kabi turli axloqiy xususiyatlar bo’yicha AI ballining maqbulligi va kutilayotgan sifatini baholadilar.
https://googleads.g.doubleclick.net/pagead/ads?client=ca-pub-9585941727679583&output=html&h=250&slotname=5494213682&adk=315283653&adf=2022854937&pi=t.ma~as.5494213682&w=300&lmt=1696938823&format=300×250&url=https%3A%2F%2Fwww.psypost.org%2F2023%2F10%2Fpeople-believe-their-moral-traits-are-too-distinct-for-ai-to-judge-study-finds-213968&host=ca-host-pub-2644536267352236&wgl=1&uach=WyJXaW5kb3dzIiwiMTAuMC4wIiwieDg2IiwiIiwiMTE3LjAuNTkzOC4xNTAiLFtdLDAsbnVsbCwiNjQiLFtbIkdvb2dsZSBDaHJvbWUiLCIxMTcuMC41OTM4LjE1MCJdLFsiTm90O0E9QnJhbmQiLCI4LjAuMC4wIl0sWyJDaHJvbWl1bSIsIjExNy4wLjU5MzguMTUwIl1dLDBd&dt=1696957473883&bpp=1&bdt=361&idt=336&shv=r20231004&mjsv=m202310030101&ptt=9&saldr=aa&abxe=1&cookie=ID%3D2b196ed822638318-228f449b45df0062%3AT%3D1696957015%3ART%3D1696957015%3AS%3DALNI_MZhsJb4E4109alSlPl2S7-kOf8cRg&gpic=UID%3D00000d9869619fc4%3AT%3D1696957015%3ART%3D1696957015%3AS%3DALNI_MYdoqiWzzDM6kkKYz11EKq0D6313A&prev_fmts=750×280%2C300x250&correlator=4425782298123&frm=20&pv=1&ga_vid=120765468.1696957474&ga_sid=1696957474&ga_hid=772420914&ga_fc=0&u_tz=300&u_his=4&u_h=864&u_w=1536&u_ah=824&u_aw=1536&u_cd=24&u_sd=1.25&dmc=8&adx=452&ady=1975&biw=1505&bih=790&scr_x=0&scr_y=0&eid=44759876%2C44759927%2C44759837%2C31078593%2C31078657%2C44804782%2C44805099%2C31078673%2C31078663%2C31078665%2C31078668%2C31078670&oid=2&pvsid=1471702984887259&tmod=435951133&uas=0&nvt=1&ref=https%3A%2F%2Fwww.psypost.org%2F&fc=896&brdim=30%2C26%2C30%2C26%2C1536%2C0%2C1536%2C882%2C1522%2C790&vis=1&rsz=%7C%7CeEbr%7C&abl=CS&pfx=0&fu=0&bc=31&td=1&nt=1&ifi=3&uci=a!3&btvi=2&fsb=1&xpc=x8xh8pi5OK&p=https%3A//www.psypost.org&dtd=347
Tahlillar shuni ko’rsatdiki, ishtirokchilarning 40 foizi AI axloqiy bahosini nomaqbul deb bilishgan, ishtirokchilarning atigi 26 foizi esa AI axloqiy bahosini maqbul deb bilishgan. Bundan tashqari, qabul qilish idrok etilgan aniqlik bilan bog’liq edi. Boshqacha qilib aytganda, odamlar sun’iy intellektdan foydalanishni qabul qilish ehtimoli kamroq bo’lgan, chunki ular bu ularning axloqini aniqlashda to’g’ri emasligiga ishonishgan.
2a tadqiqotida odamlar o’zlarining axloqiy profillarining o’ziga xosligini ortiqcha baholaydilarmi yoki yo’qmi, ishtirokchilarning o’z profilining tarqalishi haqidagi e’tiqodlarini haqiqiy tarqalish bilan solishtirish orqali tekshirildi.
https://googleads.g.doubleclick.net/pagead/ads?client=ca-pub-9585941727679583&output=html&h=250&slotname=5494213682&adk=1044959112&adf=1718569669&pi=t.ma~as.5494213682&w=300&lmt=1696938823&format=300×250&url=https%3A%2F%2Fwww.psypost.org%2F2023%2F10%2Fpeople-believe-their-moral-traits-are-too-distinct-for-ai-to-judge-study-finds-213968&host=ca-host-pub-2644536267352236&wgl=1&uach=WyJXaW5kb3dzIiwiMTAuMC4wIiwieDg2IiwiIiwiMTE3LjAuNTkzOC4xNTAiLFtdLDAsbnVsbCwiNjQiLFtbIkdvb2dsZSBDaHJvbWUiLCIxMTcuMC41OTM4LjE1MCJdLFsiTm90O0E9QnJhbmQiLCI4LjAuMC4wIl0sWyJDaHJvbWl1bSIsIjExNy4wLjU5MzguMTUwIl1dLDBd&dt=1696957473884&bpp=1&bdt=362&idt=462&shv=r20231004&mjsv=m202310030101&ptt=9&saldr=aa&abxe=1&cookie=ID%3D2b196ed822638318-228f449b45df0062%3AT%3D1696957015%3ART%3D1696957015%3AS%3DALNI_MZhsJb4E4109alSlPl2S7-kOf8cRg&gpic=UID%3D00000d9869619fc4%3AT%3D1696957015%3ART%3D1696957015%3AS%3DALNI_MYdoqiWzzDM6kkKYz11EKq0D6313A&prev_fmts=750×280%2C300x250%2C300x250&correlator=4425782298123&frm=20&pv=1&ga_vid=120765468.1696957474&ga_sid=1696957474&ga_hid=772420914&ga_fc=0&u_tz=300&u_his=4&u_h=864&u_w=1536&u_ah=824&u_aw=1536&u_cd=24&u_sd=1.25&dmc=8&adx=452&ady=2448&biw=1505&bih=790&scr_x=0&scr_y=0&eid=44759876%2C44759927%2C44759837%2C31078593%2C31078657%2C44804782%2C44805099%2C31078673%2C31078663%2C31078665%2C31078668%2C31078670&oid=2&pvsid=1471702984887259&tmod=435951133&uas=0&nvt=1&ref=https%3A%2F%2Fwww.psypost.org%2F&fc=896&brdim=30%2C26%2C30%2C26%2C1536%2C0%2C1536%2C882%2C1522%2C790&vis=1&rsz=%7C%7CeEbr%7C&abl=CS&pfx=0&fu=0&bc=31&td=1&nt=1&ifi=4&uci=a!4&btvi=3&fsb=1&xpc=MyFOxAf5Oj&p=https%3A//www.psypost.org&dtd=469
Tadqiqotchilar “Sizning axloqingiz” loyihasidan onlayn so’rovnomalarni to’ldirgan 131 015 ishtirokchidan iborat mavjud ma’lumotlar to’plamidan ma’lumotlarni tahlil qilishdi ( yourmorals.org ). Loyihada g’amxo’rlik, adolat, sadoqat, hokimiyatning ma’naviy o’lchovlari baholandi. Ushbu ishtirokchilar to’rtta o’lchovning har biri uchun qanchalik yuqori yoki past bo’lganligiga qarab 16 ta alohida axloqiy profilga bo’lingan va tadqiqotchilar ushbu profillarning har biri populyatsiyada qanchalik tez-tez sodir bo’lishini hisoblab chiqdilar.
Shundan so’ng, Purcell va Bonnefon bir xil onlayn so’rovnomalarni to’ldirish uchun AQShda 495 ishtirokchini jalb qildi va keyin ularni 16 axloqiy profildan biriga ajratdi. Keyin ishtirokchilarga umumiy aholi orasidan har bir axloqiy profilga ega bo’lgan odamlarning foizini tavsiflovchi yorliqsiz grafik taqdim etildi, so’ngra ishtirokchilardan o’zlarining profil turiga ega deb hisoblagan odamlarning foizini tanlashni so’rashdi.
2b tadqiqoti 2a tadqiqotini takrorladi, lekin yorliqsiz grafikdan ularning axloqiy profili qanchalik keng tarqalganligini to’g’ri taxmin qilish uchun moliyaviy rag’bat berdi va AQShdan 496 ishtirokchidan iborat yangi guruhni tahlil qildi.
Birgalikda quyi tadqiqotlar shuni ko’rsatdiki, ishtirokchilarning 88 foizi o’zlarining axloqiy profillari aholi orasida qanchalik keng tarqalganligini etarlicha baholamagan, bu ularning axloqi yanada noyobroq ekanligiga ishonishlarini ko’rsatadi.
Nihoyat, 3-tadqiqot odamlar AI tomonidan yaratilgan profillarning odatiy profillarga nisbatan aniqligini qanday baholaganliklarini baholadi.
Tadqiqotchilar AQShda onlayn so’rovlarni yakunlagan 506 ishtirokchini jalb qilishdi. 1-tadqiqga o’xshab, ular maqbullik va kutilgan sifatli axloqiy xususiyatlarni baholadilar, ammo keyin AI tomonidan axloqiy baholash odatiy yoki noyob profillar uchun aniqroq va ishonchliroq bo’lishiga ishonishlari yoki AI yaxshi ishlashiga ishonishlari so’ralgan. o’zlarining axloqiy profillarini baholash.
Ushbu kichik tadqiqot shuni ko’rsatdiki, ishtirokchilar sun’iy intellekt o’ziga xos axloqiy profillarni baholashda qiyinchiliklarga duch kelishiga ishonishadi va bu xuddi shu ishtirokchilarni AI ularning axloqini noto’g’ri tavsiflaydi deb o’ylashlariga olib keldi.
Purcell va Bonnefon shunday xulosaga kelishdi: “Bizning topilmalarimiz texnologiya aniqroq bo’lganda ham, AI axloqiy ball olish uchun katta ishtaha bo’lmasligi mumkinligini ko’rsatadi. Bu odamlar sun’iy intellektning axloqiy bahosiga qarshi qat’iy qoidalarni ma’qullashi mumkinligini anglatsa-da, bu shuningdek, ushbu vositaning tijorat salohiyati cheklangan bo’lishi mumkinligini anglatadi, hech bo’lmaganda odamlar o’zlarining axloqlarini baholash uchun mashinalar uchun juda o’ziga xos his qilsalar.
Tadqiqot AI axloqiy reytingining psixologik maqbulligi haqida qimmatli ma’lumotlarni taqdim etsa-da, ba’zi cheklovlarni ta’kidlash kerak.
Misol uchun, tadqiqot faqat mavhum stsenariylarga qaratilgan, ish uchun arizalar yoki tanishish kabi aniq kontekstlarga emas. Shunday qilib, topilmalar AI axloqiy reytingi haqiqiy hayotda qanday ishlashini to’liq aniqlay olmaydi. Bundan tashqari, tadqiqot mualliflari axloq keng ma’noda tahlil qilinganligini va “umumiy axloqiy xarakter, o’ziga xos axloqiy xususiyatlar va axloqiy qadriyatlar” ni o’z ichiga olganligini ta’kidladilar. Shu sababli, kelajakdagi tadqiqotlar axloqning turli tomonlarini batafsilroq o’rganishi mumkin.
Izzattillo Islomov
