Inês Antunes da Costa authors the article “AI in criminal law: from prediction to decision”

Inês Antunes da Costa, associate in the criminal, regulatory offences and compliance team, authors the article “AI in criminal law: from prediction to decision”, published in Observador, in which she offers a critical analysis of the impact of artificial intelligence on criminal proceedings and the implications of these technologies for the fundamental principles of justice.

In the article, the author frames the influence of artificial intelligence on legal science as “so disruptive that it is almost akin to a scientific revolution in the sense described by Thomas Kuhn”, highlighting the complexity of the boundary between “the prohibited and the permitted, the desirable and the inconceivable” in the criminal law sphere.

When addressing predictive mechanisms, she warns of the risk of perpetuating errors and biases, arguing that “feeding artificial intelligence systems with the information required to teach them the desired notion of justice has the counterproductive potential to perpetuate the very errors of judgement inherent to human decision-making”. This concern is particularly evident in spatial prediction models, where “the more an area is policed, the more crimes are recorded there; and the more crimes are recorded, the more the algorithm insists that the area is dangerous”.

The author also highlights the risks associated with personal prediction tools, which have a direct impact on the core pillars of criminal proceedings, “from the presumption of innocence to the right to a fair trial”, drawing attention to increased risks relating to privacy, decision-making autonomy and procedural safeguards.

At the level of judicial decision-making, the article exposes the limits of relying on such systems, expressly asking: “how can the judge move beyond that knowledge in order to deliver a decision beyond the machine?”, particularly when the mechanisms used are opaque or difficult to understand from a legal perspective.

The article concludes that “the promise of efficiency cannot replace the duty to judge”, arguing that the true transformation underway will not be technological, but legal, and that it requires a profound reflection on the model of justice society wishes to pursue.

The full article is available here.