Right to explanation
A tsari na lissafi mai tsauri, musamman na wucin gadin da kuma ta subfield na na'ura koyo, mai dama don bayani (ko dama don wani bayani) ne a dama da za a bai wa wani bayani ga wani fitarwa na algorithm. Waɗannan haƙƙoƙin suna da farko suna nufin haƙƙin mutum don a ba shi bayani game da yanke shawara da ke shafar mutum da gaske, musamman ta fuskar doka ko ta kuɗi. Misali, mutumin da ya nemi lamuni kuma aka hana shi na iya neman bayani, wanda zai iya zama " Rahoton hukumar bayar da bashi ta X da ta sanar da fatarar ku a bara; ba zai ba ku rancen da kuka nema ba. "
 Right to explanation | |
|---|---|
| Bayanai | |
| Ƙaramin ɓangare na |
subjective right (en)  |
| Facet of (en) | Kirkirar Basira (Artificial Intelligence) |
Wasu daga cikin irin waɗannan haƙƙoƙin na haƙƙin doka sun wanzu, yayin da ikon gama gari game da '' haƙƙin bayani '' lamari ne da ke ci gaba da tattaunawa.
Biyan kuɗi a Amurka
gyara sashe
A ƙarƙashin Dokar Bayar Da Lamunin Daidaita (Dokar B ta Dokar Dokokin Tarayya ), Take na 12, Fasali na X, Kashi na 1002, §1002.9, ana buƙatar masu ba da bashi don sanar da masu neman waɗanda aka hana su daraja tare da takamaiman dalilan dalla-dalla. Kamar yadda yake a cikin §1002.9 (b) (2): [1]
Fassarar hukuma ta wannan ɓangaren tana bayyana irin maganganun da za a yarda da su. Masu ba da bashi suna bin wannan ƙa'idar ta hanyar samar da jerin dalilai (galibi aƙalla 4, ta fassarar ƙa'idodi), wanda ya ƙunshi adadin reason code (azaman mai ganowa) da bayanin haɗin gwiwa, gano manyan abubuwan da suka shafi lambar daraja. [2] Misali na iya kasancewa kamar haka:
- 32: Adadin ma'auni akan katin banki ko asusun juyawa yayi yawa idan aka kwatanta da iyakokin bashi.
Tarayyar Turai
gyara sasheDokar Kare Bayanai ta Ƙungiyar Tarayyar Turai (wanda aka kafa a shekarar 2016, wanda zai fara aiki a shekara ta 2018) ya shimfida ikon yanke hukunci kai tsaye a cikin Dokar Kare Bayanai ta shekara ta 1995 don samar da wata hanyar da aka yi ta takaddama a kanta na hakkin bayani, wanda aka bayyana kamar haka a Recital 71 : "[ ya kamata batun bayanan ya kasance] da ‘yancin ... don samun bayanin hukuncin da aka yanke”. A cikakke:
Koyaya, yadda ƙa'idojin da kansu suke ba da '' haƙƙin bayani '' ana muhawara sosai. [3] [4] Akwai manyan layuka guda biyu na suka. Akwai mahimman batutuwan doka tare da haƙƙi kamar yadda aka samo a cikin Mataki na 22 - kamar yadda tunatarwa ba ta ɗaure, kuma ba a ambaci haƙƙin bayani a cikin abubuwan da ke ɗauke da rubutun ba, kasancewar an cire su yayin aiwatar da doka. Bugu da kari, akwai wasu hane-hane masu mahimmanci kan nau'ikan yanke shawara na atomatik wadanda aka rufe - wanda dole ne ya zama duka biyu "kawai" bisa aikin sarrafa kansa, kuma suna da halaye na doka ko makamancin haka - wanda ke iyakance kewayon tsarin sarrafa kansa da yanke shawara wanda dama zai yi amfani. [3] Musamman, haƙƙin ba zai yiwu a yi amfani da shi ba a yawancin shari'o'in rikice-rikice na algorithmic waɗanda aka ɗauka a cikin kafofin watsa labarai.
An nuna tushen tushe na biyu na irin wannan haƙƙin a cikin Mataki na 15, "'yancin samun dama ta hanyar bayanan". Wannan ya sake maimaita irin wannan tanadi daga Dokar Kare Bayanai na shekara ta 1995, yana bawa mai bayanan damar samun "bayanai mai ma'ana game da dabaru da ke ciki" a cikin mahimmin abu guda daya, yanke shawara kai tsaye, wanda aka samu a Mataki na 22. Amma duk da haka wannan ma yana fama da ƙalubalen da ake zargi da suka shafi lokacin da za a ɗora wannan haƙƙin, da kuma ƙalubale na aiki waɗanda ke nufin bazai yuwu ba a lokuta da yawa na damuwar jama'a.
Faransa
gyara sasheA Faransa Loi zuba une République numérique (Digital Republic Act ko loi numérique ) sun gyara lambar gudanarwar kasar don gabatar da sabon tanadi don bayanin hukuncin da hukumomin ma'aikatun gwamnati suka yanke game da mutane. Ya lura cewa inda akwai "shawarar da aka yanke bisa tsarin maganin algorithmic", dole ne a sanar da dokokin da ke ayyana wannan magani da kuma "manyan halayensa" ga dan kasa idan aka nema, inda babu wariyar (misali don ƙasa tsaro ko kariya). Wadannan ya kamata sun haɗa da kamar haka:
- digiri da yanayin gudummawar aikin algorithmic zuwa yanke shawara;
- bayanan da aka sarrafa da kuma tushen sa;
- sigogin jiyya, da kuma inda ya dace, nauyin su, ya shafi yanayin wanda abin ya shafa;
- ayyukan da aka gudanar ta hanyar magani.
Masana sun lura cewa wannan haƙƙin, yayin da yake iyakance ga yanke shawara na gudanarwa, ya wuce haƙƙin GDPR don a bayyane ya shafi goyan bayan yanke shawara maimakon yanke shawara "kawai" dangane da sarrafa kansa, da kuma samar da tsari don bayyana takamaiman yanke shawara. Haƙiƙa, GDPR haƙƙin yanke shawara na atomatik a Tarayyar Turai, ɗayan wuraren da ake neman "haƙƙin bayani" a ciki, gano asalinsu a dokar ƙasarFaransa a ƙarshen shekara ta 1970s.
Sukar
gyara sasheWadansu suna jayayya cewa "'yancin bayani" ba shi da kyau, a mafi munin cutarwa, kuma yana barazanar dakile kirkire-kirkire. Musamman takaddama sun haɗa da: fifita yanke shawara na ɗan adam akan shawarar na'ura; kasancewa mara aiki tare da dokokin da ke akwai da kuma mai da hankali kan tsari kan sakamako. [5]
Mafi mahimmanci, yawancin algorithms da ake amfani dasu a cikin ilmantarwa na inji ba saukin bayyanawa. Misali, fitowar cibiyar sadarwar zurfin jiki ya dogara da yawa na lissafin lissafi, an haɗa su a cikin hanya mai rikitarwa, kuma babu wanda ya shigar da shi ko lissafin da zai iya zama babban rinjaye. Fannin Bayyana AI yana neman samar da ingantaccen bayani daga algorithms da ake dasu, da kuma algorithms waɗanda suka fi saurin bayyana, amma filin saurayi ne kuma mai aiki.
Hakanan, ba za a iya bayanin yanke shawara na mutum sau da yawa a sauƙaƙe ba: suna iya dogara ne da azanci ko wani ji na ciki wanda ke da wahalar bayyanawa. Wasu zai yi jayayya cewa bai kamata a buƙaci injina su cika ƙa'idar da ta fi ta mutane ba.
Duba kuma
gyara sashe- Tabbatar da algorithmic
- Bayanin hankali na wucin gadi
- Dokar algorithms
Manazarta
gyara sashe- ↑ Consumer Financial Protection Bureau, §1002.9(b)(2)
- ↑ US FICO credit risk score reason codes: Fundamental document from FICO listing all of the FICO credit score reasons that a score is not higher, March 31, 2010, by Greg Fisher
- ↑ 3.0 3.1 Cite error: Invalid
<ref>tag; no text was provided for refs named:0 - ↑ Is there a 'right to explanation' for machine learning in the GDPR?, Jun 1, 2017, Andrew Burt
- ↑ EU's Right to Explanation: A Harmful Restriction on Artificial Intelligence, Nick Wallace, Center for Data Innovation, January 25, 2017
Hanyoyin haɗin waje
gyara sashe- Lantarki na Artificial Yana Baku Bayani, na John Frank Weaver, Slate, Mayu 8, 2017