THE AI-BASED JUDGE: A NEW THREAT TO THE CRIMINAL JUSTICE SYSTEM?

Author: Kyrill Ryabtsev

1. Introduction

Technological advancement influences all trias politica branches, including judiciary. Courts are undergoing the digitalization process to become more efficient. When it comes to adjudication matters, there have been plans in several States to incorporate AI-based algorithms replacing human judges. Such initiatives have raised interest among academics regarding their compatibility with the right to a fair trial under Article 6 ECHR. Specifically, the establishment of an AI-based tribunal is controversial when it comes to the criminal justice system as it deals with particularly serious crimes and penalties. Due to the complexity of criminal cases, it is debatable whether their adjudication can be left to automated machines.

Accordingly, this article will answer this research question: to what extent does the employment of AI-based robot judges in criminal courts contradict Article 6(1) ECHR. The scope of this work will be limited to the requirements of independence and impartiality of the tribunal. Firstly, this article will discuss how these standards are perceived by the ECtHR in its case law. Secondly, it will address how an AI-based algorithm can replace human judges and benefits it brings to criminal justice system. Thirdly, this article will analyze key drawbacks AI-based criminal tribunals have regarding their independence and impartiality. As this article supports the idea of AI-based criminal judges, it will suggest possible safeguards for their compliance with these standards.

1. AI as a Judge: The case of Independence and Impartiality

2. Setting the Scene

To uphold the right to a fair trial, Article 6(1) ECHR prescribes that the case should be heard by an independent and impartial tribunal. This requirement plays a pivotal role in ensuring that the RoL is upheld throughout CoE members and that human rights are effectively protected. Notably, these standards are intertwined and thus considered together by the ECtHR.

Regarding the tribunal’s independence, there are four criteria established by the Court, namely, (1) the manner by which the judges are appointed, (2) the duration of the term in office, (3) presence of any guarantees against outside pressure, and (4) an overall appearance of tribunal’s independence. Together, they allow a judge to freely decide on cases without external influence. The notion of independence covers not only the judiciary’s independence from other government branches, but also from parties to the proceedings and third parties. Importantly, the standard of independence is crucial for the judicial branch for upholding the RoL and ensuring separation of powers.

As to the impartiality assessment, the Court conducts subjective and objective tests. Through the subjective test, it examines if judges are free from personal interests or convictions about a case. By performing the objective test, the Court evaluates whether there are sufficient guarantees excluding legitimate concerns of the parties about a tribunal’s impartiality. The impartiality standard is seen as a prerequisite for the right to a fair trial, as the judgement should be based on the law and without prejudices from judges.

1. Robot Judge: A Panacea for Judicial Inefficiency

Courts are currently facing problems such as their members’ lack of independence, and racist-based prejudices. These issues directly threaten fair trial standards under Article 6 ECHR. The AI, however, is capable of overcoming these complications by automating the adjudication process.

Firstly, regarding the independence requirement, the AI algorithm can significantly reduce the degree of outside pressure on judges. As the adjudicative authority is vested in an algorithm rather than human judges, it is much more complex for third parties to exercise undue influence on judicial decisions. For instance, AI algorithms strengthen internal judicial independence by being free from judicial superiors’ pressure within the human judges’ hierarchy. An AI-based tribunal enhances external judicial independence by eliminating judges’ bribery. Indeed, to influence an AI-based court’s verdict, the accused must hack its underlying algorithm and access the parameters the robot judge uses while deciding upon the case. This however requires advanced knowledge of computer sciences and computing power.

Secondly, regarding the tribunal’s impartiality, a robot judge’s technological architecture makes it a neutral arbiter and ensures the tribunal’s compliance with subjective test of impartiality. Psychological research has proven that unconscious bias among judges will always be present, specifically regarding those of a criminal nature. It is indeed hard to be emotionless for human judges in cases of murder and drug trafficking. By contrast, robot judges, do not have feelings and can soberly tackle such cases. Similarly, the design of robot judges allows them to disregard irrelevant external factors, such as race of the accused. It would help to eliminate racial bias within the criminal justice system. Lastly, AI-based judges cannot have a personal interest in a case it deals with. The facts’ evaluation by automated criminal tribunals would therefore be more objective and equitable.

1. Threats to the Right to a Fair Trial and Safeguards

AI judges can bring significant benefits to ensure the tribunal’s impartiality and independence. Nonetheless, a turbinal composed from AI-based judges also creates justifiable concerns around its compliance with Article 6(1) ECHR.

Firstly, a robot judge is based on a specific algorithm, developed by the State with the help of private parties. Meanwhile, its functioning would be undisclosed to the accused to ensure that s/he cannot manipulate the adjudication process. Additionally, algorithm-driven court decisions will be opaque due to their technological and contextual complexity, thus making it sophisticated to explain robot judge verdict ex post. Private companies involved in AI-based judges creation can also request not to disclose it for trade secrecy reasons. This lack of transparency can cause doubts for the accused about the court’s independence from the State and its private developers. Accordingly, the population’s tendency to mistrust automatized machines should be taken into account. Resultantly, AI-based judges would not satisfy the criterion of overall appearance of the tribunal’s independence.

Secondly, to effectively perform its functions, an AI algorithm is trained on the basis of already existing data. As the modern criminal justice system is biased towards people of color, the newly developed AI algorithm is vulnerable to embody these prejudices. Therefore, there is a risk of fostering institutionalized racism via AI-based courts. Such “by-architecture” racist tribunal would not pass a subjective impartiality test. Additionally, the mere possibility of robot judges being racist by-design can create justified fear of lack of their impartiality in the accused’s eyes. This would cause citizens’ general mistrust in the judiciary and result in the AI-based tribunal failing the objective impartiality test.

Resultantly, despite their major benefits, AI-based judges risk to be insufficiently independent and impartial. Effective safeguards should thus be established. This work advocates for the use of AI-based judges in criminal cases as their advantages can effectively rectify weaknesses of human judges. Nonetheless, in case of justified concerns around AI-based judges' independence or impartiality, there should be a possibility for an accused to appeal in front of human judges. There should also be strict oversight on robot judges’ development to limit racist biases. Lastly, promoting technological literacy among the population is crucial, as it would establish people’s trust in automated tribunals. This also implies a certain degree of transparency about AI-based judges’ functioning, but to an extent precluding an accused from influencing the adjudication process.

1. Conclusion & Further Research

Technologies have a big impact on the judiciary and its operation. Automation of courts via AI can solve many issues, especially when it comes to independence and impartiality requirements under Article 6(1) ECHR.

This article has addressed these notions within the realm of the criminal justice system. An AI-based tribunal has the potential of being immune to undue influence. It also successfully satisfies subjective impartially analysis as the digital architecture can ensure complete neutrality of robot judges. Nonetheless, AI-based judges create justifiable concerns around their independence and impartiality. The ambiguity surrounding the nature of robot judges decision-making process imperils the overall appearance of automated tribunals’ independence. Similarly, they risk to fail both subjective and objective impartiality tests by being biased towards people of color. Therefore, procedural safeguards are necessary to ensure the AI-based judges’ compliance with Article 6(1) ECHR. They include a possibility to appeal automated tribunals' decision in human courts, a rigid control over robot judges’ development during their creation, and advancement of people’s technological literacy.

The topic of AI-based judiciary is highly complex and needs a broader analysis of its compatibility with the right to a fair trial. The Convention was drafted to control courts composed of human judges. It remains to be identified whether a body consisting of AI-based robot judges can be considered ‘a tribunal’ as established by the ECtHR. Additionally, AI-based judges have no room for reasonableness in their verdicts as this is a human exercise based on a human ratio rather than an empirical exercise by an algorithm. Hence, judgements might become very entrenched, rigid and arguably too strict. How is reasonableness determined by AI algorithms and how would a non liquet be avoided by AI judges remains to be solved. Further research is therefore needed to provide the reader with a more complete picture of AI’s current place within the criminal justice system.

Bibliography

Primary Sources:

Primary Law:

Convention for the Protection of Human Rights and Fundamental Freedoms (European Convention on Human Rights, as amended) [1950]

Case Law:

Henryk Urban and Ryszard Urban v Poland App no 23614/08 (ECtHR, 30 November 2010)

Findlay v the United Kingdom App no 22107 (ECtHR, 25 February 1997)

Campbell and Fell v the United Kingdom App no 7819/77; 7878/77 (ECtHR, 28 June 1984)

Moiseyev v Russia App no 62936/00 (ECtHR, 9 October 2008)

Şahiner v Turkey App no 47533/99 (ECtHR, 4 May 2006)

Micallef v Malta App no 17056/06 (ECtHR, 15 October 2009)

Kyprianou v Cyprus App no 73797/01 (ECtHR, 15 December 2005)

Castillo Algar v Spain App no 863/1074 (ECtHR, 28 October 1998)

Stafford v The United Kingdom App no 46295/99 (ECtHR, 28 May 2002)

Soft Law:

European Commission for the Efficiency of Justice, European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and their environment (Council of Europe 2018)

European Court of Human Rights, Guide on Article 6 of the European Convention on Human Rights: Right to a Fair Trial (Criminal Limb) (Council of Europe 2020)

Secondary Sources:

Books:

Cordella A and Contini F, Digital Technologies for Better Justice: A Toolkit for Action (Inter-American Development Bank 2020)

International Commission of Jurists, International Principles on the Independence and Accountability of Judges, Lawyers and Prosecutors: Practitioners Guide No. 1 (International Commission of Jurists 2007)

Trechsel S, Human Rights in Criminal Proceedings (OUP 2006)

The Sentencing Project, Racial disparity in the criminal Justice System: A Manual for Practitioners and Policymakers (The Sentencing Project: Research, Advocacy and Reforms 2016)

Cui Y, Artificial Intelligence and Judicial Modernization (Springer 2020)

Vetzo M and others, Algoritmes en Grondrechten (Boom Juridisch 2018)

Lessig L, Code: Version 2.0 (Basic Books 2006)

Edited Volumes:

Roca JG and Zapatero JMV, ‘The Right to an Independent and Impartial Tribunal (Art. 6.1 ECHR): A Specific Minimum Standards Guarantee Rather than a Rule of Justice’ in Roca JG and Santolaya P (eds), Europe of Rights: A Compendium on the European Convention of Human Rights (Martinus Nijhoff Publishers 2012)

Themeli E and Philipsen S, ‘AI as the Court: Assessing AI deployment in Civil Cases’ in Benyekhlef K (ed), AI and Law. A Critical Overview (Éditions Thémis 2021)

Kramer X and others, ‘e-Justice in the Netherlands: the Rocky Road to Digitised Justice’ in Weller M and Wendland M (eds), Digital Single Market: Bausteine eines Rechts in der Digitalen Welt (Mohr Siebeck 2018)

Journal Articles:

Huang J and Karduck A, ‘A Methodology for Digital Government Transformation’ (2017) 5 Journal of Economics, Business and Management 246

Velicogna M, ‘Justice Systems and ICT: What can be learned from Europe?’ (2007) 3 Utrecht Law Review 129

Jasper U, ‘The Impact of Artificial Intelligence on the Right to a Fair Trial: Towards a Robot Judge?’ (2020) 11 Asian Journal of Law and Economics 1

Dymitruk M, ‘The Right ti a Fair Trial in Automated Civil Proceedings’ (2019) Massaryk University Journal of Lawn and Technology 27

Završnik A, ‘Criminal justice, artificial intelligence systems, and human rights’ (2020) 20 ERA Forum 567

Meron T, ‘Judicial Independence and Impartiality in International Criminal Tribunals’ (2005) 99 The American Journal of International Law 359

Voeten E, ‘The Impartiality of International Judges: Evidence from the European Court of Human Rights’ (2008) 102 American Political Science Review 417

Dijkstra S, ‘The Freedom of the Judge to Express his Personal Opinions and Convictions under the ECHR’ (2017) 13 Utrecht Law Review 1

Garcia-Bolivar OE, ‘Lack of Judicial Independence and Its Impact on Transnational and International Litigation’ (2012) 18 Law and Business Review of the Americas 29

Coglianese C and Dor LMB, ‘AI in Adjudication and Administration’ (2020) 1 Penn Law: Legal Scholarship Repository 1

Levinson JD and others, ‘Judging Implicit Bias: A National Empirical Study of Judicial Stereotypes’ (2017) 69 Florida Law Review 63

Re RM and Solow-Niederman A, ‘Developing Artificially Intelligent Justice’ (2019) 22 Stanford Technology Law Review 242

Buocz TJ, ‘Artificial Inteligence in Courts: Legitimacy Problems of AI Assistance in the Judiciary’ (2018) 2 Court Management Articles 41

Glikson E and Woolley A, ‘Human Trust in Artificial Intelligence: Review of Empirical Research’ (2020) 14 Academy of Management Annals 627

Sourdin T, ‘Judge v Robot? Artificial Intelligence and Judicial Decision-Making’ (2018) 1 UNSW Law Journal 1114

Rachlinski JJ and others, ‘Does Unconscious Racial Bias Affect Trial Judges?’ (2009) Notre Dame Law Review 1195

Huq AZ, ‘Racial Equity in Algorithmic Criminal Justice’ (2019) 68 Duke Law Journal 1043

Meyerson D, ‘Why Should Justice Be Seen to Be Done’ (2015) 34 Criminal Justice Ethics 64

Reiling DA, ‘Courts and Artificial Intelligence’ (2020) 11 International Journal for Court Administration 1

Blogs

Levi DF, ‘What Does Fair and Impartial Judiciary Mean and Why is it Important?’ (Duke Law Bolch Judicial Institute, 26 October 2019) https://judicialstudies.duke.edu/2019/11/what-does-fair-and-impartial-judiciary-mean-and-why-is-it-important/ accessed 17 April 2021

Rikleen LS, ‘When it comes to unconscious bias, are judges at risk?’ (ABA Journal, 31 October 2019) https://www.abajournal.com/voice/article/are-judges-at-risk-for-unconscious-bias accessed 17 April 2021

Park J, ‘Your Honour, AI’ (Harvard International Review, 3 April 2020) https://hir.harvard.edu/your-honor-ai/ accessed 17 April 2021

Donald B, ‘Judges On Race: Reducing Implicit Bias In Courtrooms’ (Law 360, 6 December 2020) https://www.law360.com/articles/1309550/judges-on-race-reducing-implicit-bias-in-courtrooms accessed 17 April 2021

Eidenmüller H, ‘Series: Machine Performance and Human Failure’ (Oxford Business Law Blog, 2 November 2018) https://www.law.ox.ac.uk/business-law-blog/blog/2018/11/law-and-autonomous-systems-series-machine-performance-and-human accessed 17 April 2021

Websites:

Engstrom DF, ‘Can AI Be a Fair Judge in Court? Estonia Thinks So’ (Wired, 25 March 2019) https://www.wired.com/story/can-ai-be-fair-judge-court-estonia-thinks-so/ accessed 17 April 2021

Dai S, ‘Shanghai judicial courts start to replace clerks with AI assistants’ (South China Morning Post, 1 April 2020) <www.scmp.com/tech/innovation/article/3077979/shanghai-judicial-courts-start-replace-clerks-ai-assistants> accessed 17 April 2021

Bayer L, ‘Court of Justice says Polish judicial amendments may violate EU law’ (Politico, 2 March 2021) https://www.politico.eu/article/court-of-justice-says-polish-judicial-amendments-may-violate-eu-law/ accessed 17 April 2021

Greenstein MN, ‘AI and a Judge’s Ethical Obligations’ (American Bar Association, 3 February 2020) https://www.americanbar.org/groups/judicial/publications/judges_journal/2020/winter/ai-and-judges-ethical-obligations/ accessed 17 April 2021

Hao K and Stray J, ‘Can you make AI fairer than a judge? Play our courtroom algorithm game’ (MIT Technology Review, 17 October 2019) https://www.technologyreview.com/2019/10/17/75285/ai-fairer-than-judge-criminal-risk-assessment-algorithm/ accessed 17 April 2021

Buranyi S, ‘Rise of the racist robots – how AI is learning all our worst impulses’ (The Guardian, 8 August 2017) https://www.theguardian.com/inequality/2017/aug/08/rise-of-the-racist-robots-how-ai-is-learning-all-our-worst-impulses accessed 17 April 2021

Reports:

OECD, ‘Strengthening Digital Government’ (2019) OECD

European Court of Human Rights, ‘Judicial Seminar 2019 Strengthening the confidence in the judiciary: Background Document’ (2019) European Court of Human Rights

European Court of Human Rights, ‘Judicial Seminar 2018 The Authority of the Judiciary: Background Document’ (2018) European Court of Human Rights

Other Materials:

Finucan L, ‘Smart Courts: Roadmap for Digital Transformation of Justice in Africa’ (2018) Access Partnership

Min B and Ferris G, ‘Regulating Artificial Intelligence for Use in Criminal Justice Systems in the EU: Policy Paper’ (2020) Fair Trials

Sicilianos LA, ‘The Rule of Law and the European Court of Human Rights: the independence of the judiciary’ (2020) European Court of Human Rights

Power A, 'Judicial Independence and the Democratic Process: Some Case Law of the European Court of Human Rights’ (2012) International Bar Association Conference

Lohn AJ, ‘Hacking AI: A Primer For Policymakers on Machine Learning Cybersecurity’ (2020) Center for Security and Emerging Technology

Branting K and others, ‘Introduction to AI and Courts’ (2020) Joint Technology Committee