인공지능의 자동화된 의사결정과 기본권 보장에 관한 연구 = A Study on Ensuring Constitutional Rights in Automated Decision-Making by Artificial Intelligence

다국어 초록 (Multilingual Abstract)

Automated decision-making by artificial intelligence significantly improves efficiency in industrial and public sectors but raises serious concerns regarding infringements of constitutional rights such as the right to informational self-determination, equality rights, and the right to access information. To effectively ensure constitutional rights, the following legal and institutional measures are necessary.
First, securing transparency and explainability of algorithms is essential. Although Personal Information Protection Act of Korea stipulates the right to demand explanations for automated decisions, it is necessary to expand the scope of such explanations. Clear obligations should be established for documenting, storing, and disclosing algorithm operations and decision logs. Furthermore, an independent impact assessment or audit and supervisory system for algorithms involved in automated decisions need to be developed. Moreover, under Personal Information Protection Act of Korea, personal information controllers are required to disclose the fact of automated decision-making, its purpose, and the scope of data subjects involved. However, it is not easy to verify whether data subjects have accessed this information, and they cannot exercise their rights to refuse automated decisions or request explanations or reviews unless they have actually reviewed it, necessitating improvements in this regard. Second, data subjects must be guaranteed substantial rights to object to and seek remedies against automated decisions, including rights of rebuttal and redress. Current laws only guarantee the right to request explanations or reviews but lack provisions for human intervention and objection rights. A structured objection and reassessment procedure, multiple channels for relief, deadlines, mandatory reporting, and disclosure of decision outcomes should be systematically established. Legal supplements are also needed to facilitate easy access and procedural clarity for objections and reassessment requests regarding human reprocessing results by personal information controllers. Third, legal accountability for automated decision-making must be strengthened, clarifying responsible entities. While existing laws impose basic duties on data processors, they lack clear definitions of responsible parties at each stage, division of roles, and effective enforcement mechanisms. Korea should pursue enhanced accountability through explicit definitions of responsible parties, obligations and sanctions per party, introduction of strong burden-shifting principles, mandatory technical documentation, and logging requirements. Fourth, separate legislation and institutional measures are urgently needed in fields directly related to social opportunity distribution—such as hiring, promotion, credit evaluation, and welfare—to prevent bias and discrimination. This includes prior notice and consent concerning AI usage, explanations of evaluation criteria and data types, mandatory fairness and bias assessments by independent bodies with public disclosure of results, guarantees of objection and alternative evaluation requests, and administrative and criminal sanctions along with supervision for legal violations.