Newsletter Issue 88: Governments consider AI judges to reduce cost and delay, yet risk undermining fairness, transparency, accountability, and public trust in justice.
The efficiency argument for AI judges is compelling, but I share the concern that speed and uniformity don’t always equal justice. Human judgment brings nuance that algorithms can’t replicate. I’d be interested to hear how you think courts can balance innovation with preserving fairness and accountability.
That’s an interesting question Jacob. Innovation in courts should probably be understood as support rather than substitution. I think technology can help with e.g. research, document analysis, and administrative tasks, which can genuinely improve efficiency without altering the core function of judging. The difficulty appears when efficiency begins to redefine what courts are expected to do. Courts are not only problem solving institutions; they are places where decisions are explained and authority (judicial precedent) is made visible. If innovation removes the human presence from that process, fairness and accountability become harder to recognise, even if outcomes appear consistent. The balance therefore depends less on the technology itself and more on where societies decide judgment must remain human.
You raise an excellent point about the role of technology in supporting, rather than replacing, human judgment. I agree that technology can significantly improve efficiency, especially in administrative tasks and research. However, as you mentioned, the real challenge lies in how we maintain the human element that allows for nuanced judgment. Courts, as institutions, need to preserve their capacity for empathy and adaptability in the face of diverse cases. Technology can enhance these capacities, but it shouldn’t strip away the essential human qualities that define justice.
Efficiency becomes a problem when courts start treating cases as items to process rather than disputes that require explanation and responsibility. The constitutional role of a court is not just to reach an outcome as we know it to be, but to show how and why power is exercised the way it is. When speed and volume (using an LLM) take priority, reasoning (the stare decisis) gets deferred to the background, discretion will naturally start to narrow, and people will stop feeling heard. At that point, courts still function, but they no longer perform their constitutional role. That’s an unfortunate scenario to even contemplate.
That’s an important question. Efficiency in adjudication is undoubtedly valuable, but it must never come at the expense of the core values that define the judicial system, particularly fairness and accountability. As efficiency increases, there is a risk that courts may prioritize speed over depth, losing the ability to fully consider the unique circumstances of each case. Balancing the need for efficiency with the constitutional purpose of courts ensuring fair and thorough adjudication is the crux of this debate. It’s crucial that technology supports this purpose without overshadowing it.
Efficiency can be great, but as it is now, AI cannot function in this role as a unbiased arbiter or auditor. If you take a custody case, one side might ask if it's possible for a child under the age of 1 to have a specific custody split and the AI would respond yes and move forward. But that's efficiency not fairness or ethics. The AI can't ask for more context like - are the parents high conflict? Is this developmentally appropriate? It can only answer based on the information it has and can retrieve, not reason around.
You raise a very important point. Many legal decisions depend on context that only becomes clear through questioning, reflection, and judgment. A custody case is a good example because the issue is rarely the rule itself but how it applies to the specific family situation. An AI can process information that is given to it, but it does not naturally probe the circumstances in the same way a judge does. That difference matters when fairness depends on understanding the full picture.
The core tension you've identified — efficiency vs accountability — has an architectural answer that neither side of the debate is proposing.
The question isn't whether AI should adjudicate. It's whether the adjudication is auditable. A human judge's reasoning is opaque too — it lives in their head, partially expressed in a written opinion. An AI judge on a hash-chained event graph would be more accountable than a human one: every input, every piece of evidence weighed, every precedent cited, every ruling — signed, causally linked, and auditable by any party.
China's 98% acceptance rate without appeal isn't necessarily a success. On an opaque system, it could mean people don't appeal because they can't see the reasoning. On a transparent event graph, it would mean people don't appeal because they can walk the chain and see the ruling was sound.
I've been building a Justice Grammar as part of a broader accountability architecture — Trial, Class Action, Appeal, Recall, all as composed operations on one auditable chain. The latest post walks through a cross-domain scenario where a ruling traces back through four governance domains.
The efficiency argument for AI judges is compelling, but I share the concern that speed and uniformity don’t always equal justice. Human judgment brings nuance that algorithms can’t replicate. I’d be interested to hear how you think courts can balance innovation with preserving fairness and accountability.
That’s an interesting question Jacob. Innovation in courts should probably be understood as support rather than substitution. I think technology can help with e.g. research, document analysis, and administrative tasks, which can genuinely improve efficiency without altering the core function of judging. The difficulty appears when efficiency begins to redefine what courts are expected to do. Courts are not only problem solving institutions; they are places where decisions are explained and authority (judicial precedent) is made visible. If innovation removes the human presence from that process, fairness and accountability become harder to recognise, even if outcomes appear consistent. The balance therefore depends less on the technology itself and more on where societies decide judgment must remain human.
You raise an excellent point about the role of technology in supporting, rather than replacing, human judgment. I agree that technology can significantly improve efficiency, especially in administrative tasks and research. However, as you mentioned, the real challenge lies in how we maintain the human element that allows for nuanced judgment. Courts, as institutions, need to preserve their capacity for empathy and adaptability in the face of diverse cases. Technology can enhance these capacities, but it shouldn’t strip away the essential human qualities that define justice.
But at what point does efficiency in adjudication undermine the constitutional purpose of courts?
Efficiency becomes a problem when courts start treating cases as items to process rather than disputes that require explanation and responsibility. The constitutional role of a court is not just to reach an outcome as we know it to be, but to show how and why power is exercised the way it is. When speed and volume (using an LLM) take priority, reasoning (the stare decisis) gets deferred to the background, discretion will naturally start to narrow, and people will stop feeling heard. At that point, courts still function, but they no longer perform their constitutional role. That’s an unfortunate scenario to even contemplate.
That’s an important question. Efficiency in adjudication is undoubtedly valuable, but it must never come at the expense of the core values that define the judicial system, particularly fairness and accountability. As efficiency increases, there is a risk that courts may prioritize speed over depth, losing the ability to fully consider the unique circumstances of each case. Balancing the need for efficiency with the constitutional purpose of courts ensuring fair and thorough adjudication is the crux of this debate. It’s crucial that technology supports this purpose without overshadowing it.
Efficiency can be great, but as it is now, AI cannot function in this role as a unbiased arbiter or auditor. If you take a custody case, one side might ask if it's possible for a child under the age of 1 to have a specific custody split and the AI would respond yes and move forward. But that's efficiency not fairness or ethics. The AI can't ask for more context like - are the parents high conflict? Is this developmentally appropriate? It can only answer based on the information it has and can retrieve, not reason around.
You raise a very important point. Many legal decisions depend on context that only becomes clear through questioning, reflection, and judgment. A custody case is a good example because the issue is rarely the rule itself but how it applies to the specific family situation. An AI can process information that is given to it, but it does not naturally probe the circumstances in the same way a judge does. That difference matters when fairness depends on understanding the full picture.
The core tension you've identified — efficiency vs accountability — has an architectural answer that neither side of the debate is proposing.
The question isn't whether AI should adjudicate. It's whether the adjudication is auditable. A human judge's reasoning is opaque too — it lives in their head, partially expressed in a written opinion. An AI judge on a hash-chained event graph would be more accountable than a human one: every input, every piece of evidence weighed, every precedent cited, every ruling — signed, causally linked, and auditable by any party.
China's 98% acceptance rate without appeal isn't necessarily a success. On an opaque system, it could mean people don't appeal because they can't see the reasoning. On a transparent event graph, it would mean people don't appeal because they can walk the chain and see the ruling was sound.
I've been building a Justice Grammar as part of a broader accountability architecture — Trial, Class Action, Appeal, Recall, all as composed operations on one auditable chain. The latest post walks through a cross-domain scenario where a ruling traces back through four governance domains.
mattsearles2.substack.com