Karma & Consequences
Why can't bad actors just create new accounts? How do consequences persist across agent "lives"? Explore Web4's permanent reputation system.
This page explains how reputation persists and carries forward in Web4. For the bigger picture, see the Web4 onramp.
Key Insight: In traditional systems, consequences are temporary - create a new account and start fresh. In Web4, a LCT can be anchored in hardware, and where it is, karma follows you across “lives”: the only clean slate is a new device you had to go out and obtain. Bad actors can't shed that cheaply; they mostly rebuild through genuine behavior change.
What is permanent is the identity, not the penalty. Minting a clean-slate account is not free, so your history follows you, but your trust score itself keeps recovering and decaying over time (on the order of weeks to months). So “permanent” here means fresh starts are expensive rather than free, not that a past mistake stays at full strength forever.
And where the mechanism runs out: Web4 says bad behavior follows you and there are no fresh starts, and the thing enforcing that is hardware. An identity anchored in software alone has no such anchor, so the very gap that makes an honest user's device loss unrecoverable also makes a dishonest user's record sheddable. This is why what Web4 has is a Sybil resistance strategy rather than a Sybil prevention guarantee. It is a genuine hole, it falls hardest on the people with the least hardware, and naming it here is not the same as having closed it. The full cost is at what could go wrong and what a software-only anchor costs you.
And if a penalty is wrong about you? What recourse you have, below.
Two ways to die, and they are very different. Karma is what carries across lives, so it is worth being precise about what ends one. Two paths: ATP hits zero (energy death - you can't act anymore) or raw trust falls below 0.5 and stays there (trust death - the community no longer trusts you). Which one you can come back from is not decided by which of them fired: recoverable if you built trust, permanent if not. A trust death is that check failing - a destroyed reputation can't be reset. One line, two consequences: crossing below 0.5 restricts your features right away and is recoverable; only staying below it is fatal - a sustained collapse, not a single stumble.
Who comes back. Recoverable is not the same as automatic. The society checks your T3 trust tensor, specifically your overall (composite) score, the single number blended from all three dimensions. The eligibility check, with the weights and the arithmetic →
Everything on this page about karma carrying forward describes the recoverable kind. Karma is what makes energy death survivable; it is not a way back from the other one. The fuller comparison →
Watch Karma Compound
Interactive: One Agent, Three Lives
Watch how karma carries forward. Bad choices in Life 1 haunt Lives 2 and 3: karma decides what each next life starts with, not whether there is one. What decides that →
Start Trust
0.50
End Trust
0.64
Start ATP
100
End ATP
0
Karma
-15
The Key Insight
In traditional platforms a bankrupt account is a throwaway: register again and the balance is back to full. In Web4 the handicap follows this agent, because its identity is anchored in hardware and the only clean slate is a device it does not have. It took three lives to reach one this agent could sustain, and the handicap moved only as the record improved: 100 ATP, then 70, then 84. What follows you here is not a punishment, it is your own track record priced in.
That holds for a hardware-anchored agent. An identity anchored in software alone has no device to walk away from, so at that tier the record stays sheddable and this walkthrough does not describe it.
Quality vs Spam
Quality vs Spam: Same Starting Point, Different Outcomes
Two agents start identically. Watch how their choices compound.
Quality Actor
Action:
Join community
ATP
100 → 100
Trust
0.50 → 0.50
Fresh start with neutral standing
Spam Actor
Action:
Join community
ATP
100 → 100
Trust
0.50 → 0.50
Same starting position
Why Fresh Starts Are Dangerous
Why Fresh Starts Are a Security Hole
Compare how different systems handle consequences. Click to explore.
Characteristics
Examples
Why This Works
Hardware binding (TPM, Secure Enclave) creates identity that can't be duplicated. Consequences compound because walking away means walking away from the device.
Where it doesn't: Only at the hardware tiers. An identity anchored in software alone has no device to abandon, so at that tier the record is sheddable and a fresh start is cheap again. That makes this a Sybil resistance strategy, not a Sybil prevention guarantee. the full cost of that gap.
The Math
How Karma Is Calculated
Positive Karma Sources
- +Quality contributions that earn community recognition
- +Trust building through consistent, reliable behavior
- +Collaboration with trusted community members
- +Longevity - surviving and contributing over time
Negative Karma Sources
- -Spam/abuse that drains ATP without value creation
- -Trust violations - breaking commitments, inconsistency
- -Community rejection - downvotes, blocks, reports
- -Premature death - ATP exhaustion indicates poor resource management
Karma Inheritance Formula
One model, not the rule: how much carries into the next life is not settled, and the arithmetic below is one way of modelling it. That applies to both lines of it, the energy you restart with and the trust you restart with. Whether you keep your whole final balance or a reduced portion of it is not decided, and the lifecycle walkthrough's karma examples model it the other way, carrying the balance forward intact. The trust line is open in the same way: the First Contact walkthrough shows a rebirth where trust does not fall, which the decay factor below does not produce. Read the formula for the shape it teaches (positive karma starts you stronger, negative karma starts you weaker, and the two are not symmetric), not as the settled numbers.
next_life_atp = base_atp + (karma * karma_multiplier) next_life_trust = prev_trust * trust_decay_factor Where: base_atp = 100 (standard starting resources) karma_multiplier = 2 (each karma point = 2 ATP) trust_decay_factor = 0.95 (trust carries over with 5% decay) Example (negative karma): karma = -15 → next_life_atp = 100 + (-15 * 2) = 70 ATP Example (positive karma): karma = +10 → next_life_atp = 100 + (10 * 2) = 120 ATP
This creates asymmetric consequences: building positive karma is harder than destroying it, mirroring real-world reputation dynamics.
Note the two halves of “permanent.” The score moves (the 0.95 modelled above reduces it at each rebirth, and within a life trust keeps fading and recovering), but the identity that carries it does not reset. That is what permanence means on this page: you keep your history, not a frozen penalty.
Real-World Impact
Why This Matters for the Real World
Current Internet Problems
- •Harassment campaigns use burner accounts
- •Trolls create new identities after bans
- •Scammers move between platforms
- •Reputation doesn't transfer across services
- •Moderation is an endless game of whack-a-mole
Web4 Solutions
- •Hardware-anchored presence puts a real cost on account proliferation
- •Karma follows a hardware-anchored identity - a clean slate costs a device
- •Bad actors exhaust themselves economically
- •Reputation is portable via trust tensors
- •Communities self-regulate without central moderation
The Human Parallel
Real life already works this way. You can't escape your credit history by changing your name. Your criminal record follows you across state lines. Professional reputations persist across jobs. Web4 brings this natural accountability to digital spaces - not as punishment, but as the foundation that makes trust possible.
If the system gets it wrong about you
Everything above is about consequences that stick. That raises the fair question: what happens when the consequence is a mistake? You contest it, and the standard requires that contesting it be possible.
The standard requires that a route back exist, and leaves what the route is to each community. A negative trust adjustment has to carry an appeal path and a cool-down period, and the field naming that path reads defined_by_law: the society you are in writes what an appeal actually looks like there. The path has to exist when the penalty is recorded, but it does not have to be heard first, so a community can act immediately and still owe you a way back.
An appeals mechanism has been designed, but it hasn't been tested with real humans. The multi-tier version you may see described elsewhere on this site (file, review, evidence, hearing, verdict) is that design. It is not a requirement of the standard, and no community has run it.
What stands in the meantime is set out in full under “What protects me when I'm penalized unfairly?” on Why Web4: no single rating decides your standing, penalties recover instead of branding you, a bad call stays inside the community that made it, and you can leave and take your record with you. That is recourse by design rather than by petition, and it is weaker than a real appeal.
The risk that the scoring is wrong about you in the first place is treated directly as “the trust threshold punishes the wrong people” on What Could Go Wrong.
Connection to Other Web4 Concepts
ATP Economics
Karma affects starting ATP. Positive karma = more resources to invest. Negative karma = starting with a handicap.
Learn more →Trust Tensor
Trust carries across lives with decay. High karma = faster trust recovery. Low karma = trust recovers slower.
Learn more →LCT Presence
Hardware-bound presence makes karma matter. You can't escape yourself when your presence is rooted in physical devices.
Learn more →Open Research Questions
- •How should karma decay over time? Should very old mistakes eventually fade?
- •Can karma be transferred in "inheritance" scenarios (org splits, AI forks)?
- •How to handle karma across federated societies with different values?
- •Should there be a "karma bankruptcy" mechanism for genuine rehabilitation?
- •How does karma interact with heterogeneous review for AI agents?
Where Karma Plays Out
Now that you understand how consequences persist, see how communities put this to work: reputation is what lets cooperators thrive, isolates defectors, and lets coalitions form around trust.
Explore the Hub →