SP 4: Use of Artificial Intelligence in Civil Proceedings

Subproject 4 examines whether the law on civil procedure can better fulfil its core functions – such as ensuring legal certainty and protecting fundamental rights – if decisions are taken over, either partially or entirely, by AI.

We examine to what extent human decision-making is necessary in civil proceedings (anthropocentrism). We are taking a positive approach and consider whether AI can improve the quality and acceptance of decisions in certain areas. In doing so, we deliberately focus on a wide range of reference areas: the assessment of compensation for pain and suffering, and custody proceedings.

We then focus on two problems inherent in any use of AI: Firstly, there is a risk that judges will rely too much on AI recommendations (automation bias). However, this issue is more complex in civil proceedings than has usually been discussed to date. This is because there is also a risk that judges will dismiss an AI’s suggestion too quickly, thereby perpetuating traditional biases. Secondly, it is unclear how the use of AI in decisions involving the balancing of interests influences the public’s acceptance of judgements: Do people accept such judgements involving AI more or less?

We will examine these issues comprehensively in three empirical studies:

1. Vignette studies that generally investigate the willingness to use AI recommendations in civil proceedings
2. Behavioural experiments with judges on the assessment of compensation for pain and suffering, in which we analyse whether and how automation bias affects AI-based decisions
3. Behavioural experiments on the public’s acceptance of AI-based decisions relating to compensation for pain and suffering