People trust unfavourable AI-assisted court decisions – if it’s explained how the AI works, new research finds

Explaining how AI reaches decisions can significantly increase people’s trust in AI-assisted court rulings and reduce their intention to appeal an unfavourable outcome, finds new research from UCD Michael Smurfit Graduate Business School and Trinity College Dublin.

In a study involving 859 UK adults, participants were asked to imagine receiving an unfavourable court decision made with the assistance of AI. Some were given an explanation of how the AI worked, others were told how it reached the recommendation, while the third were given no explanation.

Both types of explanations significantly increased cognitive and affective trust in AI-assisted decisions, even though the result didn’t go their way.

Cognitive trust, measuring how confident participants felt that the process was reliable, increased from 2.82 without an explanation on a seven-point scale to 4.22 when an explanation was provided. Affective trust, which reflects participants’ emotional confidence in the process, also increased from 2.72 to 3.60.

The research also found that explanations reduced participants’ intention to appeal the decision. Again, on a seven-point scale, intention to appeal fell from 5.26 among those given no explanation to 4.46 for those given a process explanation and 4.33 for those given an outcome explanation.

The findings suggest that explaining how AI is used in judicial decision-making can help strengthen trust in the process and its perceived legitimacy. The researchers found that explanations build trust by strengthening perceptions of procedural justice – whether people feel the process is fair. Respect was the strongest driver, accounting for 61 to 72% of the relationship between explanations and trust.

"These explanations matter because they help people understand how AI works and signal that the court respects people’s need to understand the process and the decision affecting them. By making the reasoning behind the decision clearer, courts can help people feel that the process is fair, and their perspectives and concerns are being taken seriously,” says Marius Claudy, researcher and Associate Professor of Marketing at UCD Smurfit School.

The findings highlight the importance of how artificial intelligence is communicated to people involved in court proceedings, as courts and policymakers consider the growing use of AI-assisted judicial decision-making.

ENDS/

For more information, to receive a copy of the paper or speak to researcher Professor Marius Claudy, please email Lucy Whytock at BlueSky Education at [email protected]