关键词: 算法错误侵害, 算法解释权, 社会化救济, 救济基金

Abstract: The development of artificial intelligence technology not only creates infinite possibilities for the advancement of human society, but also makes algorithm decision-making a new type of technical power. As a result, it exerts an important influence on the “power-right” pattern in the digital field. In this regard, in the face of the violation of the rights and interests of data subjects caused by algorithm errors, the mechanism of algorithm interpretation right has been widely mentioned as an ideal way to make algorithm decision-making transparent. However, this mechanism still encounters dual dilemma in the process of correcting the alienation of algorithm power: the difficulty of explaining the internal logic and the high cost of power exercise. Therefore, it is imperative to establish an appropriate harm risk distribution model of algorithm errors. The traditional “perpetrator-victim” bipolar model cannot alleviate the negative external effects of algorithmic decision-making due to the change of the attribution basis, the blurred and broken chain of causality and the imbalance between the scale of damage and the property capacity of the perpetrator. In view of this, it is essential to explore how to allocate risks from the perspective of socialized relief to the damage. Although liability insurance strengthens the compensatory ability of the person responsible for the property, it cannot provide sufficient relief to the victims because of its natural parasitic nature of tort liability. In contrast, being not limited to the determination of tort liability, the relief fund can expand the channels of risk dispersion through the socialization of capital composition and the weakening of causal elements, thus achieving sufficient, efficient and timely relief effects.

Key words: algorithm error harm, algorithm interpretation right, socialized relief, relief fund