Artificial intelligence could serve as a built-in, blockchain-based judge in prediction markets, according to Andrew Hall, professor of political economy at Stanford Graduate School of Business.
He illustrated the problem of “fair” dispute resolution using the example of Venezuela’s presidential election.
Last year, contracts worth more than $6 million were placed on the election outcome. However, after the vote, the market fell into uncertainty:
-
the government declared Nicolás Maduro the winner;
-
the opposition and international observers reported widespread fraud.
“Should prediction market contracts be settled based on the ‘official’ outcome (Maduro’s victory) or the ‘consensus of credible reports’ (an opposition win)?” Hall asked.
This is not an isolated case, he noted. In another instance, someone allegedly manipulated maps of Ukraine related to a territorial dispute.
Hall argues that building a trusted, fair dispute-resolution system is critical. Only then can market prices become meaningful signals for society.
The problem goes beyond prediction markets
Similar challenges affect financial markets. For years, the International Swaps and Derivatives Association (ISDA) has struggled with settlement issues in credit default swaps — contracts that pay out if a company or country defaults.
Decision committees vote on whether a credit event has occurred, but the process is frequently criticized for its lack of transparency, potential conflicts of interest, and inconsistent outcomes.
“The core problem remains the same: when large sums depend on defining what happened in ambiguous situations, any settlement mechanism becomes a target for manipulation, and ambiguity becomes a source of controversy,” Hall said.
Key properties of an effective solution
Hall outlined several essential characteristics any viable system should possess:
-
Resistance to manipulation — if verdicts can be influenced by editing Wikipedia, spreading fake news, bribing oracles, or exploiting loopholes, the market becomes a game won by the best manipulator.
-
Reasonable accuracy — while perfect accuracy is impossible, the system must be correct in most cases and avoid systematic errors.
-
Transparency — traders must clearly understand how decisions are made.
-
Neutrality — participants must trust that the system does not favor any particular user or outcome.
Human-based committees can satisfy some of these conditions, but they remain vulnerable to manipulation and cannot ensure true neutrality.
AI as a solution
Hall proposes using large language models (LLMs) as automated judges, with each model and its prompt permanently recorded on the blockchain at the time of contract creation.
The basic architecture works as follows:
-
When creating a contract, the market maker specifies not only the settlement criteria in natural language, but also the chosen LLM and the exact prompt used to determine the outcome.
-
This specification is cryptographically recorded on the blockchain.
-
Once trading begins, participants can review the full settlement mechanism, knowing exactly how the model accesses information sources and reaches its verdict.
This approach addresses several major challenges:
-
Manipulation resistance — LLM outputs are difficult to tamper with. To alter results, attackers would need to manipulate the underlying information sources.
-
Accuracy — AI systems can rapidly scan large volumes of online information.
-
Transparency — the entire dispute-resolution process is open to inspection, eliminating rule changes and subjective decisions.
-
Neutrality — LLMs have no financial interest in outcomes and cannot be bribed.
However, drawbacks remain. AI systems can make mistakes, misinterpret news articles, or generate incorrect facts.
Manipulation is not impossible — only harder. Bad actors could attempt to plant misleading information in major media outlets, which is costly but feasible.
There is also the risk of poisoning training data, though such attacks would need to occur long before contract creation.
Conclusion
AI-based settlement replaces one set of problems with another — but a more manageable one. Hall believes platforms should actively experiment with different LLMs to gain operational experience.
As best practices emerge, the community should work toward standardizing AI model configurations, which would help concentrate liquidity and improve overall market efficiency.
ES 
EN