Returning to the Human Judge in the Age of Artificial Intelligence

Gabrielle Appleby

This is the first in a special series of posts reflecting on 25 years of the Gilbert + Tobin Centre of Public Law. To see other posts in this series, click here.

Gabrielle Appleby

12.10.2026

I first began studying the judiciary, and how we regulate and support judges as key constitutional actors, in 2012. Back then, I was not confronting questions that many of my University of New South Wales (UNSW) colleagues are now deeply engaged with, about what judging would look like in the age of generative artificial intelligence (AI), and the increasingly real possibility of AI judging, or at least AI-assisted judging. But I was concerned about the importance of having human judges, engaging with human parties, and exercising their discretion and judgment on human problems, as well as the concomitant dangers of this. I have now spent more than a decade researching the nature of these vulnerabilities, and how best to design the systems through which judges are appointed, educated, supported, and held to account, to ameliorate them.

While there is now justified attention being paid to what AI might mean for judging, the focus on the human being as judge, for me, remains the right one. In fact, the possibility of AI judges – or, more likely in the nearer future, human judges relying on AI in decision-making – has brought renewed attention to the question of what it is about the human judge that we value in the first place, and how to support that human in performing the judicial role.

That question has been the fulcrum of much of the work of The Judiciary Project at the Gilbert + Tobin Centre of Public Law (the Centre), both since I joined Professor Andrew Lynch as Co-Director of the Project in 2015, and previously under his leadership. We have worked with colleagues from within academia, the courts, the profession and the government on questions around judicial appointment, misconduct, ethical support, capacity, education, impartiality and recusal, judicial reasoning and dissent, diversity, leadership, collegiality and post-retirement activity.

These areas of inquiry raise a number of different perspectives on the same question: how do we build judicial systems, institutions, and cultures that support the possibilities, and accommodate the vulnerabilities, of entrusting judicial power to human beings across the course of their judicial life?

The judge, the judiciary and the court

As Professor Lynch and I have argued, judging has individual, collective and institutional dimensions. A judge decides cases as an individual, exercising discretion, interpreting law, and responding to particular circumstances of any individual case. But judges also work collectively, as members of the judiciary, sitting on multi-member panels, and within broader judicial and legal communities. And finally, they operate within the courts as institutions whose structures, cultures and practices shape what judicial values such as independence, impartiality, accountability, transparency, efficiency, and diversity, mean in practice.

The intersection between the individual, collective, and institutional was recognised in the Australian Law Reform Commission’s (ALRC) landmark 2021 report, Without Fear or Favour: Judicial Impartiality and the Law on Bias (Without Fear or Favour). The Commission emphasised that the foundational value of judicial impartiality must not be understood solely as the responsibility of the individual judge; it is an institutional value that requires collective and institutional support.

Better support for judicial education is one part of that institutional response. Judicial education is now relatively uncontroversial at an abstract level across the country, and is increasingly recognised as pivotal to ensure not only technically and jurisprudentially competent judges, but also judges who are able to conduct the courtroom and navigate the situation of parties and others in ways that draw out the strengths of having a human judge involved. Work undertaken at the Judiciary Project in 2021 with myself, Dr Jess Kerr (University of Western Australia), Professor Suzanne Le Mire (now at the University of Queensland), Professor Lynch (UNSW), and Professor Brian Opeskin (University of Technology Sydney), highlighted the extent and nature of judicial education offerings in Australia. Our conclusions included that the opportunities and institutional support for judicial education differed dramatically across the country, and levels of courts.

The ALRC subsequently reinforced these conclusions and recommended (Recommendation (9)) a more structured approach to continuing judicial education to combat concerns particularly around cultural competencies, unconscious biases, and stereotypical thinking. It specifically recommended (Recommendation (10)) an ongoing program of Aboriginal and Torres Strait Islander cross-cultural education developed and delivered under Aboriginal and Torres Strait Islander leadership. Of course, as Dr Kerr has pointed out, unless individual judges are willing to engage in self-reflective practice, institutional support will only go so far.

Encouragingly, the recently revised fourth edition of the Guide to Judicial Conduct (Guide), released by the Council of Chief Justices of Australia and New Zealand, takes up the ALRC’s recommendation. It expressly recognises continuing education as part of judicial competence, and that it is an individual responsibility but one that requires institutional support. The Guide refers to substantive and procedural courses, as well as those that provide social context training, recognising the importance of legal knowledge, but also the capacity of one human to know about the people and communities upon whom judicial power is exercised, and, to draw upon the words of UNSW Founding Dean Hal Wootten, upon whom it is felt most often and most harshly.

This increasing institutional acknowledgment of the importance of judicial education takes on a new significance as AI is poised to enter judicial work. Future judicial education will need to equip judges to understand the capacities and limitations of the technologies they use, but the challenge is more than technological competence. If AI is increasingly relied upon by judges to assist with research, synthesis, drafting or other aspects of decision-making, the qualities that this modern iteration of judicial education has increasingly sought to cultivate – self-reflection, awareness of bias, cultural competence, and attentiveness to the experiences of those before the court – become more rather than less important.

Human judgment

Human judgment is central to what judging is. However, the characteristics that human beings bring to judging create vulnerabilities – including bias, partiality and inconsistency – that judicial institutions must seek to ameliorate. But eliminating human variability is not the same thing as improving judging. Experience, perspective and discretion also make possible contextual judgment, responsiveness to a more diverse society, and the development of the law itself.

The Judiciary Project, together with the Gender and Public Law Project at the Centre, has had a longstanding commitment to the feminist judgments projects, and broader critical judgment projects, including in hosting the international website for these projects. These projects represent an increasingly rich vein of global critical scholarship and practice that makes visible the choices and discretion involved in judging. Choice is revealed to operate at a substantive level: which facts and experiences are understood as legally significant, how doctrine is characterised, and how competing principles are weighed. It operates at a procedural level: whose voices are heard and the regulation of participation. And it operates in judicial style: how reasons are written, whose perspectives are acknowledged, and how judicial authority is expressed. These projects remind us of the many ways in which the human matters in judging, and that so often the legal output is only one part of delivering justice within a human dispute. The potential into the future of significant AI use by judges threatens the progress that has been made in centring the need for critical reflection by judges. Generative AI using large language models may reproduce the language of critique, but they do not themselves bring human experience or situated perspective to the exercise of judgment. This potential underscores, once again, the need for the critical role of the human in the judicial role.

I have also considered the importance of diversity of perspectives within judicial leadership roles. Having previously written on the role of Chief Justice with Associate Professor Heather Roberts (Australian National University), arguing that this office brings with it particular responsibilities to protect and promote judicial institutional values at time of conflict and crisis, we have more recently turned to consider what this means in the context of feminist judicial leadership. We argued that diversity of leadership, drawing on case studies of first-generation Australian female judicial leaders in the Australian states, can affect what problems are noticed, how traditional judicial values such as judicial independence are interpreted and applied to modern problems, and which values are emphasised, with greater emphasis being placed, for instance, on transparency and diversity. The significance of diversity in judicial leadership cannot be reduced to the production of potentially different answers to institutional problems. It relies, rather, on different experiences and perspectives to positions of institutional authority, affecting what problems are noticed and prioritised, how institutional values are understood based on lived experience, and how judicial culture should develop. This is another important element of human judging in the collective sense that must be remembered if attention is to turn to AI-assisted judging.

Judicial vulnerabilities

Another substantial strand of the Centre’s work has concerned what institutions should do when human judges fail. An argument can and has been mounted that such failings would not affect an AI judge. But that misses the point. The point is that a human judge comes with a wide list of benefits that cannot be replicated with AI, and so we also need to make institutional avenues for addressing potential failings. Further, the possibility of extensive use of AI by judges raises the spectre of abuse of AI by judges, and new questions of judicial responsibility and misconduct. For instance, when does inappropriate reliance on AI, failure to properly scrutinise output or failure to disclose cross the threshold of judicial misconduct?  

My early work with Professor Le Mire asked how Australia should design systems for dealing with judicial misconduct while respecting, and even supporting, judicial independence. Following the administrative findings concerning the sexual harassment of associates by former Justice of the High Court Dyson Heydon, I organised a letter signed by more than 500 women in law calling for institutional reform, including reform to appointments and complaints handling. Work I subsequently undertook with Rosalind Dixon and Prabha Nandagopal on behavioural misconduct in constitutionally significant workplaces exposed the challenges of designing institutional responses to sexual misconduct in these unique settings.

Courts may be constitutionally significant, but they remain workplaces. Workplace hierarchies, cultures, reporting structures and constitutional arrangements can make misconduct easier to perpetrate and harder to report. More recent concerns about the rehabilitation of Heydon within parts of the Australian judiciary and legal profession have raised difficult questions about institutional memory, delivering accountability, and the difficulties of changing cultures where certain voices still dominate within the institution.

The revised fourth edition of the Guide includes a substantial new section on ‘Workplace conduct, including harassment, sexual harassment and bullying’. That recognises that judicial conduct cannot be understood solely through the traditional image of the judge sitting in court deciding cases. Judges are also colleagues, supervisors, and institutional leaders. In these roles, as the Guide makes clear, they exercise considerable power, particularly in relation to others in the judicial workplace.

The revised Guide represents important progress and leadership in articulating those institutional responsibilities. But guidance is only one part of the necessary institutional architecture. Professor Le Mire and I continue to argue for an independent system for dealing with complaints about federal judges. The ALRC took up that issue in its Without Fear or Favour report, recommending the establishment of a federal judicial commission with complaints-handling and educative functions. The Commonwealth government agreed in principle in 2022 and committed to a public consultation, but four years later a federal judicial commission has still not been established. Complaints about the conduct of federal judges continue to be handled within courts themselves, ultimately against the background of the Commonwealth Parliament’s constitutional removal power. The absence of an independent, transparent, and appropriately designed complaints body remains one of the most significant unfinished pieces of Australia’s judicial integrity framework.

Supporting human judges in a time of AI

Reflecting on the work of the Judiciary Project over the last decade and more, the human judge emerges as both valuable and fallible. Independence, impartiality, competence, and integrity depend on the individual selected to be the judge, the collective and cultural environment of the judiciary, and the institutional design within which these judges are supported to exercise their judgment.

These insights are important as the legal profession increasingly turns our attention to AI and the courts. The question is much more complicated than what judicial tasks a machine can perform ‘better’ than humans, or whether all concerns can be allayed by keeping a ‘human in the loop’. We need to also be asking what we will lose in terms of the human connection, perspective, and discretion if AI is relied upon in different parts of judicial decision-making, and how the introduction of AI might change the conditions under which human judges exercise their role, including what they see, the choices they make, how they reason with colleagues, and how responsibility for decisions is understood. AI, then, brings with it its own complex questions of judicial institutional design that engage with the individual, collective and institutional. The answers to these questions should be informed by the concern that animates the Judiciary Project’s work, which is how to design institutions and collective cultures that enable individual human judges to exercise the responsibility of judging well.

Gabrielle Appleby is a Professor at the UNSW Faculty of Law & Justice and the Director of the Judiciary Project at the Gilbert + Tobin Centre of Public Law.

Suggested citation: Gabrielle Appleby, ‘Returning to the Human Judge in the Age of Artificial Intelligence’ (12 October 2026) <https://www.auspublaw.org/blog/2026/10/returning-to-the-human-judge-in-the-age-of-artificial-intelligence>