Research: The Anti-Induction Problem

The Anti-Induction Problem

This is the most important design insight in the Chronomancy system. It’s also the least discussed.

The Problem, Stated Precisely

Prediction markets aggregate probabilistic beliefs from many participants and surface them as prices. This is their value: they synthesize distributed information into a single number that represents collective belief about a future outcome.

But as prediction markets grow larger — more capital, more participants, more accurate — something changes. A sufficiently large market creates a new kind of incentive for actors who appear in its markets.

The anti-induction problem:

If a prediction market shows 80% probability that I will do X, I now have financial incentive to do not-X and bet against myself. The bigger the market, the bigger the incentive to do the maximally surprising thing.

The prediction market that correctly models my behavior creates the conditions for me to behave unexpectedly.

This isn’t a hypothetical. It’s mathematically inevitable as market caps grow.

Why This Matters at Scale

Consider the progression:

Small market (today): $TIMEMACHINE market cap ~$34K. A market about whether a specific politician will sign a specific bill trades $50K notional. The politician’s legal team has never heard of Chronomancy. Anti-induction risk: essentially zero. The profit from gaming the market doesn’t justify the reputational or legal cost.

Medium market (growth phase): $TIMEMACHINE market cap $10M. Markets on major sporting events, elections, geopolitical outcomes trade $1M+ notional. Now a well-funded actor (a hedge fund, a political operation, a sports franchise) might notice. The financial incentive to position against publicly predicted outcomes begins to be real.

Large market (maturity): $TIMEMACHINE market cap $500M+. Markets trade hundreds of millions in notional. At this scale, the probability that central actors in major markets are also market participants — and are gaming the prediction — becomes significant. The market becomes a manipulation surface.

Prediction markets have a ceiling on their accuracy. That ceiling is set by the point where central actors’ anti-induction incentives exceed the cost of market manipulation.

The Broader Framing: Goodhart’s Law

This is a generalization of Goodhart’s Law: “When a measure becomes a target, it ceases to be a good measure.”

When a prediction market’s output becomes large enough to be financially interesting, it becomes a target. The moment an actor has significant incentive to manipulate a predicted probability, the probability loses some informational validity.

Traditional prediction market theory assumes that price formation is driven by participants updating on information. Anti-induction is the reverse: a sufficiently accurate price creates information-destroying incentives.

Why Pure Prediction Markets Can’t Solve This

The structure of a pure prediction market — “bet on what you believe will happen” — has no mechanism to address anti-induction. The market’s job is to aggregate beliefs. It has no mechanism to distinguish between:

A participant who believes the outcome will be X and bets accordingly
A central actor who bets against X and then engineers not-X

Both look identical from the market’s perspective. Both contribute to the same price signal. The anti-induction actor is, in fact, improving market accuracy in a narrow sense — they’re ensuring the market is calibrated against their own behavior — while destroying the market’s usefulness as an information tool.

Chronomancy’s Structural Antidote

The answer to anti-induction is binding consequences.

Prediction markets fail to deter anti-induction because they only settle financially between traders. The real-world actor (the politician, the athlete, the central banker) is not a party to the prediction market contract. They bear no financial consequence from the market’s accuracy or inaccuracy.

Chronomancy’s insurance and derivatives layer changes this:

REWIND (Position Insurance)

Users who hold prediction positions can buy insurance against those positions. The act of buying insurance creates a revealed-preference signal: this user believes the outcome is uncertain enough to pay a premium to hedge it.

More importantly, insured positions create counterparties who have financial interest in accuracy. The insurance pool profits when predictions are accurate and losses are claimed appropriately. The insurer is financially motivated to monitor market integrity.

Fast-Forward (Capital Efficiency)

FF vault mechanics require the vault to price positions based on their probability of correct resolution. A vault that consistently prices incorrectly depletes capital. The vault’s survival depends on having good probability estimates — creating institutional incentive to flag markets with anti-induction dynamics.

FORK (Conditional Markets)

Conditional markets create explicit linkages between outcomes: “P(B | A occurs)”. These nested conditionals make the relationship structure of events explicit. An actor who intends to game outcome A while outcome B is conditional on A faces more complex and expensive anti-induction strategies.

The “Binding Consequences” Principle

The deepest insight is this:

Pure prediction markets aggregate belief but don’t bind action. Insurance and derivatives markets create binding financial consequences that align market participants with real-world outcomes.

A comment on Scott Alexander’s Mantic Monday (from “Synthetic Civilization”): “Prediction markets aggregate belief, but they don’t bind action. No one is required to listen, update, or decide differently. So probabilities get sharper while incentives stay flat. Truth improved; control surfaces didn’t.”

Chronomancy is the attempt to build the control surface that pure prediction markets never were. When the market output directly triggers economic consequences — insurance payouts, vault settlements, derivative expirations — the market becomes something closer to a binding contract about the world.

This doesn’t eliminate anti-induction. It makes it more expensive. An actor who wants to game a prediction market that has $10M in insured positions riding on its accuracy needs to game the insurance pool too. The cost of anti-induction scales with the depth of binding consequences layered on top of the prediction market.

Practical Implications for Protocol Design

Market selection matters

The anti-induction risk is highest in markets where:

Central actors have large financial interest in outcomes
Those actors can plausibly influence the outcome
The market is large enough to make the manipulation financially worthwhile

These markets should be approached with caution in the REWIND insurance pool: don’t offer insurance on a market where the insured party can manufacture the insured outcome.

Chrono Score as early warning

An actor who is engineering anti-induction will have deteriorating Chrono Score over time — they’re deliberately generating incorrect predictions in markets they’re gaming. CS tracking creates an audit trail that makes systematic gaming visible.

The FREEZE dispute layer

Markets where resolution seems anomalous — where the outcome appears to contradict strong prior probability — can trigger FREEZE suspension and community review. This isn’t a solution to anti-induction, but it creates friction that makes large-scale market manipulation more difficult.

Honest Assessment

Anti-induction is a ceiling on prediction market scale and accuracy. It’s not solvable — it’s a fundamental feature of systems where predictions influence the things being predicted.

What Chronomancy can do:

Build products that add binding economic stakes to predictions (REWIND, FF)
Create reputation infrastructure that makes anti-induction actors visible (Chrono Score)
Design market selection rules that avoid the highest-risk anti-induction contexts
Be honest with investors and users that pure prediction markets have this structural limitation

The goal isn’t to defeat anti-induction. The goal is to build valuable financial infrastructure that acknowledges this limit and works within it — while making the protocol more resilient than a pure prediction market at every scale level.

Related:

Resolution Problem — the oracle failure problem (distinct from anti-induction)
Identity & Sybil Resistance — anti-manipulation at the user layer
Freeze — dispute resolution mechanics
Market Microstructure — structural dynamics that enable anti-induction