透過您的圖書館登入
IP:3.21.76.0
  • 學位論文

論人工智慧演算法時代的解釋權─歐盟GDPR與我國個人資料保護法之比較研究

The Right to Explanation in the Artificial Intelligence’s Algorithmic Era: A Comparative Study of EU’s GDPR and Taiwan’s Data Protection Act

指導教授 : 楊岳平

摘要


人工智慧發展如日中天,在運用自動化決策機制處理個人資訊的過程中,對於透明性、演算法可歸責性的要求,逐漸為人重視。2018年正式施行的歐盟一般個人資料處理規則中,即對於前述情形均加以規範,並延伸而出一個重要觀念:「解釋權」,亦即在運用自動化技術對個人資料進行決策之後,資料當事人是否具有要求資料控制方提供對機制本身、或是處理結果之說明與解釋的權利。此權利之建立,對資料當事人據以聲明後續相關權利,如要求人工介入、表達意見、對決策結果異議等,或對於不公正之決策結果要求補救措施或司法訴訟,均有莫大助益,然而,由於現有GDPR條文並未將解釋權明文化,因而也引起眾多學者討論與批評。本論文將探討GDPR中解釋權的定義、範圍、爭議與適用上的困境,並討論是否得以運用其核心概念於我國個資相關法律。 本論文發現,我國現行個資相關法律,較著重於資料取得之合法性,而忽略資料取得之後、處理過程中的正確性與公平性,對於資料控制方之告知義務與資料當事人之知情權內容,亦缺乏演算法相關資訊。對此,本論文分別就不成為演算法自動決策資料客體的權利、通知義務與資訊近用權、特定決策解釋、以演算法為中心重新設計安全維護措施、強化第三方機構協助之角色五項重點分別討論之,並提出調整建議,以期在面臨人工智慧時代來臨之際,能夠加強對個人資料之保護。

並列摘要


The future development of artificial intelligence seems prosperous, however, the request of transparency and algorithmic accountability of automated decision-making on personal data attracts attention at the same time. EU’s general data protection regulation (GDPR) implemented in 2018 legislates rules concerning above-mentioned situations, and entails an important idea: right to explanation, which central arguments circumscribing on whether a data subject has a right to ask the data controller to provide explanations to the automated processing system itself, or to the decisions made by the mechanism. To build the right to explanation will be a great help to data subject to invoke other relevant rights, such as requesting human intervention, express opinions, contesting the decision, or to ask for remedies or to file a lawsuit when an unfair decision is made. Nevertheless, the right to explanation is not legislated in GDPR, and it occurs wildly discussions and criticism. The definition, range, disputes and practical predicaments about the right are explored in this article, and sequentially discussed if its core concepts could be applied to Taiwan’s current data protection laws. The article discovered, Taiwan’s current data protection laws emphasis on the legality on obtaining the personal data, but ignore the accuracy and fairness after the date is obtained and analyzed with an automated processing system. In the meantime, the notification duty of data controller and right to be informed of data subject lack of providing related algorithmic information as well. Accordingly, this article discusses separately on 5 possible approaches which can be applied to Taiwan’s data protection law, including the right not to be a subject to automated decision-making system, notification duty and right to access to data, explanation on specific decisions, algorithm-centered safety protection measures in design, enhancing the third parties’ capacities to support, and suggests some adjustments, to strengthen the protection of personal data when facing the upcoming artificial intelligent era.

參考文獻


【中文文獻】
1、李世德(2018)。〈GDPR與我國個人資料保護法之比較分析〉,《臺灣經濟論衡》,第16卷第3期,頁69-93。
2、李國偉(03/27/2018)。〈人工智慧的名稱政治學〉,《科學月刊》,580期,載於:http://scimonth.blogspot.com/2018/03/blog-post_35.html。
3、林勤富、劉漢威(2018)。〈人工智慧法律議題初探〉,《月旦法學雜誌》,第274期,頁195-215。
4、徐彪豪(2016)。〈物聯網時代的資料保護防線-以歐盟GDPR為中心〉,《科技法律透析》,第28卷第10期,頁56-71。

延伸閱讀