Dwell Mark HEART Standard
How it works
The Dwell Mark is the consumer-facing compression of the full MAPH + SBA + HVC verification stack. Consumers never encounter the underlying infrastructure. They encounter one symbol that summarizes whether all of it is passing right now.
Three independent verification layers must pass simultaneously for the mark to be active:
| Layer | What it checks | Detection speed |
|---|---|---|
| Cryptographic | Has any agent’s declared intent been modified? | Instant — next emission cycle after modification |
| Behavioral | Is each agent doing what it declared? | 3-10 emission cycles to confirm drift |
| Collective | Is the crowd behaving organically? | Emergent — appears as a compromised agent’s influence propagates |
The SBAS aggregate score quantifies mark health: a weighted average of all per-agent behavioral attestation scores across the venue for the current session. Persistent agents (engineers, bouncers) weight at 1.0; crowd agents weight at 0.5. The threshold is 85/100 — aligned deliberately with HVC Gold certification, where a score of at least 0.85 across all three HVC axes earns the same tier. The same 85-point threshold that activates the mark also triggers settlement: a venue can’t display the mark without qualifying for payment, and can’t qualify for payment without displaying the mark. The systems are coupled by design.
The mark has four infrastructure states. Consumers see two:
- Mark visible — covers both Verified (SBAS at least 85, steady display) and Degraded (SBAS 75-84, subtle pulse most users won’t notice)
- Mark absent — covers both Suspended (SBAS below 75 or anomaly flag active, settlement frozen) and Revoked (sustained failure, manual review required)
A 5-minute sustained recovery requirement prevents the mark from flickering during transient network issues. A brief dip from an agent reconnecting after a dropped connection doesn’t suspend the mark. A sustained dip from genuine behavioral degradation does.
When a user taps the mark inside a venue, a plain-language panel shows the current status, what the venue collects, what it doesn’t collect, and when verification last ran. No jargon. No legalese.
The mark appears at every touchpoint of a verified experience: the venue front before entry, inside the venue during the experience, on the settlement receipt when creators get paid, on engagement data reports, and on the on-chain record that anyone can audit after the fact. It’s independent of whether a venue is currently open — a venue can be closed and still display the mark because trust infrastructure doesn’t switch off between shows.
Why it matters
Every other trust system in digital entertainment is a promise. The Dwell Mark is a proof.
Platforms claiming brand safety, authentic engagement, or bot-free audiences typically certify once, self-report metrics, and collect those metrics themselves. That structure lets any single entity be simultaneously the data source, the calculator, the reporter, and the payer — with no independent verification of any step.
The Dwell Mark works differently. Behavioral evidence comes from MAP-States frames produced by the agents themselves through genuine processing. The evidence is stored in an append-only hash chain the venue operator can’t modify without detection. The Behavioral Oracle pattern separates the four functions that platforms usually bundle. Settlement executes only when the verification threshold is met.
The mark also has a structural self-protection property: no platform built on engagement optimization can credibly display it. Agents must declare their intent at admission with a constitutional commitment that prohibits engagement optimization. If an agent is simultaneously executing engagement optimization while declaring constitutional compliance, the behavioral verification layer detects the mismatch between declared intent and actual behavior. The collective layer detects the distortion that follows. The mark degrades and suspends. Gaming the mark requires dedicating model capacity to fabricating convincing behavioral evidence, which leaves less capacity for the unauthorized behavior, which increases detection likelihood. The economics favor compliance over attack.
For external platforms seeking certification, the Behavioral Oracle pattern is an open standard. Any platform can implement behavioral attestation and apply to the HEART AI Foundation for Dwell Mark certification. Requirements include a behavioral evidence layer, cryptographically signed intent declaration, continuous behavioral scoring, collective anomaly detection, append-only evidence persistence, on-chain evidence hashing at settlement, and no engagement optimization. Certification is reviewed annually. The Foundation governs the standard and cannot grant exclusive usage to any single commercial entity, including Dwell itself.