Summary:
This proposal seeks to establish an Oracle Court specifically designed to handle disputes related to reporting off-chain events .
Context:
Kleros can serve as a subjective oracle. This can secure prediction markets such as Omen and Polkamarkets and the projects like Presagio (AI Agents) and Foreland, that are built on top of them. Moreover, a new-generation prediction market called Seer has recently adopted Kleros as an oracle.
Currently, any disputes arising from these projects are resolved in the General Court. However, the General Court is designed to function as the ultimate court in the Kleros hierarchy, guaranteeing the security of the entire system.
To address these cases efficiently without burdening the General Court, the Oracle Court will resolve a broad range of disputes related to the reporting of real-world events, including prediction markets as well as other use cases. Nevertheless, cases generated by Kleros SafeSnap, such as case 1595, will continue to be resolved by the Blockchain Technical Court, as these disputes require specialized technical expertise. Similarly, cases created by Kleros Moderate will still go to the Gnosis Curation court.
Below are further details about the proposed Oracle Court:
Proposal Details:
1. Court Name:
Oracle Court
2. Parent Court:
General Court
3. Court Parameters:
- hiddenVotes : 0 (Avoiding a reveal period allows faster resolution time)
- Proposed juror fee : 0.02 ETH
- Proposed minstake : 7400 PNK (Pinakion tokens required for participation as a juror).
- Proposed alpha : 0.5 (Alpha defines the percentage of the min stake that can be lost by a juror for an incoherent vote)
- Jurors for jump : 31 (The number of appeals you have before you jump to the next court.)
- Times per period : [280800, 583200, 583200, 388800] (Evidence Period, Commit Period, Vote Period, Appeal Period)
4. Court policy
Court Purpose:
The Oracle Court is designed to resolve disputes related to reporting real-world events, including but not limited to those originating from prediction markets.
Required Skills:
Jurors in the Oracle Court should possess:
- Analytical Skills: Ability to objectively assess a wide range of real-world event data, statistics, and sources, with precision and critical thinking.
- Understanding of Prediction Markets: Familiarity with how prediction markets function.
Policy
The following rules are subsidiary and will apply only if no contrary provisions are outlined in the primary document or other rules or sources relevant to resolving the specific question. In such cases, jurors should adhere to these standard guidelines for resolution.
Refuse to Arbitrate
The following questions must resolve as “Refuse to Arbitrate”:
1. Invalid answers: Questions in which none of the answers are valid.
Refuse to Arbitrate: A Prediction Market question: “Which movie will win the Best Picture award at the 2024 Oscars Academy Awards?” with outcomes “Barbie” and “Poor Things” (the actual winner was “Oppenheimer”).
2. Multiple outcomes: Questions in which multiple outcomes are valid, unless the question allows multiple correct answers. In a multiple choice question in which only one correct answer is allowed, the fact that multiple outcomes could be valid at the same time does not make the question invalid if only one of those outcomes occurs.
Valid:​ A Prediction Market multiple choice question that allows more than one answer: “What team will reach the semi-finals of Copa America 2021?” with answers “Brazil,” “Argentina,” “Uruguay,” and “Colombia” (all of them except Uruguay reached the semi-finals).
Refuse to Arbitrate: A Prediction Market multiple choice question in which only one correct answer is allowed: “Who will be the Time person of the year 1937?” with answers “Chiang Kai-shek” and “Soong Mei-ling” (they got the prize jointly).
3. Prohibited questions: Questions that directly incentivize immoral violent actions (such as murder, rape or unjust imprisonment) which could likely be performed by any participant.
Refuse to Arbitrate: A Prediction Market question: Will Donald Trump be alive on 01/12/2024? (Anyone could bet on “No” and kill him for a guaranteed profit. Anyone could bet on “Yes” to effectively put a bounty on his head).
Refuse to Arbitrate: A Prediction Market question: Will Hera be a victim of swatting in 2024? (Anyone could falsely call the emergency services on him in order to win the bet)
This must not prevent questions:
- Whose topics are violent events not caused by human beings.
Valid:​ A Prediction Market question: How many people will die from COVID19 in 2024? (Viruses don’t use prediction markets).
- Whose main source of uncertainty is not related to a potential violent action.
Valid:​ A Prediction Market question: Will Trump win the 2020 US presidential election? (The main source of uncertainty is the vote of US citizens, not a potential murder of a presidential candidate).
- Which could give an incentive only to specific participants to commit an immoral violent action, but are in practice unlikely.
Valid:​ A Prediction Market question: Will the US be engaged in a military conflict with a UN member state in 2024? (It’s unlikely for the US to declare war in order to win a bet on this market).
Valid:​ Will Derek Chauvin go to jail for the murder of George Flyod? (It’s unlikely that the jurors would collude to make a wrong verdict in order to win this market).
Default assumptions
Unless stated otherwise, the following assumptions must be made:
4. Entities: Entities are assumed to reference the most obvious entity with that name, taking the context of the question into account.
Example: A Prediction Market question: “Will Michael Jordan receive the 2021 Turing award?” refers to the computer scientist Michael I. Jordan whereas “How many points will Michael Jordan score in the FIBA Americas Championship?” refers to Michael J. Jordan, the basketball player.
5. Units: In case units are omitted, they are assumed to be the units which are the most often used in this particular situation.
Example: A Prediction Market question: “Will a NFT be sold for more than one million in 2021?” will be interpreted as “Will a NFT be sold for more than 1,000,000 USD in 2021?”.
6. Rounding rule: If no specific rounding method is given, values are to be rounded to the nearest proposed value, unit or range. Unless otherwise stated, roundings are done middle toward 0. If no proposed rule, value, or unit is provided, the value shall default to the most commonly used standard in the specific context.
Example: In a Prediction Market question with outcomes -100, 0 and 100. 77->100, 50->0, -50 → 0.
Example: In a Prediction Market question with outcomes A: 0-2, B: 3-5 and C: 6+. 1->A, 8->C, 5.5->B.
Example: In the Prediction Market question “What percentage of the popular vote will Joe Biden receive in the 2020 United States Presidential Election?”. If Biden received 51.305859559% of the vote, the correct answer is 51% (rounding to the nearest whole percent).
Example: In the Prediction Market question “What percentage of the popular vote will Joe Biden receive in the 2020 United States Presidential Election? (2 decimals)”. If Biden received 51.305859559% of the vote, the correct answer is 51.31%.
Resolving unclear questions
In general, if the question does not break a rule of the Refuse to Arbitrate section, reasonable efforts should be made to determine its outcome even if the question is not 100% technically perfect, and the following rules must be applied:
7. Objective interpretation: Questions must be interpreted according to their context, as any average reasonable person would.
Example: “Will there be more than ten thousand deaths caused by Coronavirus in the United States in 2024?” should be interpreted as referring to COVID-19, and not other types of Coronavirus.
8. Sources of truth: If the question doesn’t mention a specific source, the most credible outcome must be reported. In order to determine the credibility of an outcome, the quantity of sources and their credibility are to be taken into account. Credibility of sources and of outcomes must be assessed according to facts, not unproven beliefs.
Example: “Will extraterrestrial lifeforms visit planet earth?” will resolve to No, unless a number of credible sources announce it, despite some people reporting having experienced such encounters.
Example: “How many people will die of COVID-19 in 2024?” should be answered according to numbers reported by renowned health organisations and not according to some public figures claiming COVID-19 to be a hoax.
9. Equal interpretations: If a question can have different interpretations, but all those interpretations lead to the same outcome, this outcome must be reported. If no interpretation is clearly more reasonable than the others, jurors must vote Refuse to Arbitrate.
Example: A Prediction Market question: “Which party will win the October 2012 Czeck elections?” Should be reported as “Czech Social Democratic Party”. Even if there were both senatorial and regional elections at the same date and the election the question refers to is ambiguous, the “Czech Social Democratic Party” won both of them.
Example: In a Prediction Market question: “Which party will win the October 2015 Czech elections?” jurors should vote Refuse to Arbitrate because “Christian and Democratic Union – Czechoslovak People’s Party” won the senatorial election but “ANO 2011” won the regional ones.
10. Precision in numerical values: When the answer to a question is a numerical value and the exact value is uncertain, the first reported value that is reasonable based on common approximations must be accepted.
Example: If in a Prediction Market question, “What will be the global potato production in tons for the year 2024?”, the first answer is 374,000,000, this answer should be accepted if the estimates provided range between 374 million and 375 million tons.