{"id":64321,"date":"2025-09-12T12:20:24","date_gmt":"2025-09-12T11:20:24","guid":{"rendered":"https:\/\/www.cambridge.org\/core\/blog\/?p=64321"},"modified":"2025-09-12T12:20:24","modified_gmt":"2025-09-12T11:20:24","slug":"when-justice-becomes-code-rebuilding-trust-when-ai-enters-courts","status":"publish","type":"post","link":"https:\/\/www.cambridge.org\/core\/blog\/2025\/09\/12\/when-justice-becomes-code-rebuilding-trust-when-ai-enters-courts\/","title":{"rendered":"When Justice Becomes Code: Rebuilding Trust when AI enters Courts"},"content":{"rendered":"
<\/div>\n

Artificial intelligence (AI) is making its way into courtrooms worldwide, promising efficiency and consistency in judicial decision-making. From predicting case outcomes to drafting legal documents, AI tools are reshaping the way courts operate. But as these technologies become increasingly embedded in judicial systems, an urgent question arises: Can we trust a justice system that is governed, in part, by code?<\/em><\/strong><\/p>\n\n\n\n

<\/div>\n\n\n\n

From Rule of Law to Rule of Code?<\/strong><\/p>\n\n\n

\n
\"\"
Unsplash\/Markus Spiske.<\/sup><\/sub><\/figcaption><\/figure><\/div>\n\n\n

This shift is part of a larger phenomenon known as algocracy, or governance by algorithms\u2014a concept explored by John Danaher<\/a>, who warns that when public decision-making is delegated to opaque computational systems, democratic legitimacy may erode. The concern is not just theoretical. Real-world cases from the Netherlands (SyRI<\/a>) and Denmark (Udbetaling Danmark<\/a>) illustrate how algorithmic systems can automate fraud detection and welfare decisions with limited human oversight, often at the cost of fairness, privacy, and transparency.<\/p>\n\n\n\n

In judicial contexts, where decisions carry personal consequences, the implications of algorithmic governance are especially critical. Courts are more than service providers\u2014they are symbols of justice, expected to exercise discernment, impartiality, and empathy. These human qualities are difficult to replicate, let alone guarantee, in algorithmic systems.<\/p>\n\n\n\n

<\/div>\n\n\n\n

The Crisis of Trust in Algorithmic Justice<\/strong><\/p>\n\n\n\n

Trust is subjective<\/a>. It directly relates to the AI user\u2019s will to be vulnerable to its decision-making within the judicial system. Trust in the judiciary is further built on principles of fairness, transparency, and human judgment. According to Popelier et al. (2022),<\/a> trust in multilevel judicial systems is not just about competence, it also hinges on perceptions of integrity and benevolence. As such, trust is directly related to what individuals anticipate from a tribunal\u2019s carriage of justice. And most times, when individuals enter a courtroom, they expect a fair trial where their case is carefully considered by a human judge who understands the nuances of their situation.<\/p>\n\n\n\n

This is breached with the interference of AI in judicial decisions. Furthermore, beyond individual trust,<\/strong> there is also societal trust<\/strong> that represents a collective faith in the judicial system as a symbol of justice and equality<\/a>, as well as an organizational trust<\/strong> that is tied with a perception of procedural fairness<\/a>. Without public trust, AI systems breach the principles of transparency, since they alienate<\/a> the citizens from judicial decision-making and open algorithmic justice to different kinds of societal interpretations that vary amongst each other.<\/p>\n\n\n\n

AI tools often operate as \u201cblack boxes\u201d, they provide outputs without clear explanations. This explainability gap undermines procedural justice: if people cannot understand or contest a decision, they may feel that their legal rights have been bypassed. Scholars like Kinchin (2024)<\/a> argue, fully automated systems risk creating a procedural void where individuals are unable to meaningfully participate in the adjudication of their own cases.<\/p>\n\n\n\n

<\/div>\n\n\n\n

Bias by Design<\/strong><\/p>\n\n\n\n

Far from neutral, AI systems are shaped by the data they are trained on. When historical legal data reflects systemic bias, based on race, gender, or socioeconomic status, algorithms trained on this data can replicate and reinforce injustice. Studies of AI risk assessment tools in criminal justice systems, particularly in the U.S., have shown how these tools can disproportionately penalize marginalized communities. As Amnesty International (2021)<\/a> has noted, the use of AI in public administration without safeguards may result in discriminatory outcomes that violate fundamental rights.<\/p>\n\n\n\n

<\/div>\n\n\n\n

Towards Trustworthy Judicial AI<\/strong><\/p>\n\n\n\n

What, then, can be done? <\/em>Rather than rejecting AI outright, scholars and practitioners are advocating for a trust-centered approach. Any attempt of algorithm governance in the judiciary should call for:<\/p>\n\n\n\n