Algorithmic decision making (ADM) takes on increasingly complex tasks in the criminal justice system. Whereas new developments in machine learning could help to improve the quality of judicial decisions, there are legal and ethical concerns that thwart the widespread use of algorithms. Against the backdrop of current efforts to promote the digitization of the German judicial system, this research investigates motivational factors (pragmatic motives, fairness concerns, and self-image-related considerations) that drive or impede the acceptance of ADM in court. We tested two hypotheses: (1) Perceived threat of inequality in legal judgments increases ADM acceptance, and (2) experts (judges) are more skeptical toward technological innovation than novices (general population). We conducted a preregistered experiment with 298 participants from the German general population and 267 judges at regional courts in Bavaria to study how inequality threat (vs. control) relates to ADM acceptance in court, usage intentions, and attitudes. In partial support of the first prediction, inequality threat increased ADM acceptance, effect size d = 0.24, 95% confidence interval (CI) [0.01, 0.47], and usage intentions (d = 0.23, 95% CI [0.00, 0.46]) of laypeople. Unexpectedly, however, this was not the case for experts. Moreover, ADM attitudes remained unaffected by the experimental manipulation in both groups. As predicted, judges held more negative attitudes toward ADM than the general population (d = -0.71, 95% CI [-0.88, -0.54]). Exploratory analysis suggested that generalized attitudes emerged as the strongest predictor of judges' intentions to use ADM in their own court proceedings. These findings elucidate the motivational forces that drive algorithm aversion and acceptance in a criminal justice context and inform the ongoing debate about perceptions of fairness in human-computer interaction. Implications for judicial praxis and the regulation of ADM in the German legal framework are discussed. (PsycInfo Database Record (c) 2024 APA, all rights reserved).
Read full abstract