← All Protocols / Methodology

Methodology

How scores are assigned, what the criteria mean, and how to challenge them. Every score on this site is an analytical judgment, not a fact — we welcome corrections and disagreements via GitHub issues.

Theoretical Foundation

This analysis applies Nadia Asparouhova’s framework from Dangerous Protocols (Summer of Protocols, 2023) to AI agent infrastructure. Asparouhova argues that protocols are “procedural systems of social control” — and that the most powerful ones are invisible. Participants follow them not because they were commanded to, but because they believe compliance reflects their authentic selves.

The emergence of AI agent protocols in 2024–25 is a live instance of Asparouhova’s “Protocolization 2.0”: rather than managing data, these protocols manage ideas, decisions, and autonomous action at machine speed. Understanding their control dynamics now — before they become invisible infrastructure — is the goal.

“Protocols are dangerous precisely because they control us so well. Though it may seem contradictory, the more powerful a protocol is, the harder it is to understand or explain it to others.”
— Nadia Asparouhova

◈ Kafka Index

Six criteria for identifying bad protocol design — protocols that create complexity instead of abstracting it away. Each dimension is scored Low / Medium / High / Critical.

Feedback Loop
No (or hidden) feedback loop
Edge Case Sprawl
Too many edge cases addressed at once
Outcome Ambiguity
Success outcomes randomized or ambiguously defined
Protocol Redundancy
Multiple protocols solving the same problem
Recursive Nesting
Recursive, nested protocols
Exit Cost
No market or alternatives exist

Exit Cost — Decision Rules

Low
No lock-in. Standard open protocol. Trivial to replace with any alternative.
Medium
Some ecosystem effects (e.g. tooling, community) but technical switching is straightforward.
High
50+ enterprise partners OR dominant market share (>40%) with no credible alternative at comparable scale.
Critical
Financial lock-in (tokens, sunk costs) OR protocol controls access to a market participants cannot afford to leave (merchants must adopt or become invisible).

⚠ Dangerous Protocols Analysis

Four dimensions evaluating the social control dynamics of each protocol, drawn directly from Asparouhova’s framework.

Identity Penetration
How deeply has the protocol entered participants' identity layer?
Agency Preservation
How much decision-making power does the participant retain?
Control Invisibility
How invisible is the protocol's control over participants?
Crisis Mindset
Is adoption driven by urgency/fear rather than genuine utility?

Control Invisibility — Decision Rules

Low
Protocol mechanisms are explicit, well-documented, and visible in normal operation.
Medium
Some mechanisms are implicit — protocol shapes behavior through defaults or constraints without requiring explicit compliance.
High
Protocol control operates through financial incentives, social norms, or identity without participants recognizing it as external control.
Critical
Protocol control is fully internalized as identity. Participants cannot distinguish protocol compliance from self-expression. Exit feels like self-betrayal.

Archetypes

◎WhiteheadDesirable

Balanced power between protocol and participant

“Civilization advances by extending the number of important operations which we can perform without thinking about them.”
◉BartlebyUndesirable

Participant holds too much power; high agency limits ability to manage complexity

“I would prefer not to.”
◈KafkaUndesirable

Protocol holds too much power; participant trapped in maze they can't understand or escape

“I can't find my way round in this darkness.”

Adoption Stages

From Asparouhova’s framework. Protocols move through these stages as they mature — and become progressively harder to exit as they advance.

1
Explicit Rules
Participants know the protocol and willingly enter
2
Social Expectation
Widely understood but not written down; peer-enforced
3
Identity Layer
Internalized; participants believe compliance is self-expression

Overall Risk Score

The overall risk rating is an editorial judgment, not a mechanical average. It weighs:

Low
Protocol design is generally sound. Explicit, debuggable, preserves human agency.
Medium
Notable concerns in 1–2 dimensions. Worth monitoring, especially as adoption grows.
High
Significant control risks. High exit cost or invisible control mechanisms present.
Critical
Severe, compounding risks. Financial identity fusion, near-zero exit, or designed opacity. Proceed with eyes open.

How to Contribute or Challenge a Score

Every score on this site can be challenged. If you believe a dimension is wrong, open a GitHub issue with:

  1. Which protocol and dimension you’re challenging
  2. The score you think it should be
  3. Evidence (links, not assertions)

To add a new protocol, submit a PR with a new JSON file in /data/protocols/ following the schema in /data/schema/protocol.schema.json.

Framework by Nadia Asparouhova · Summer of Protocols 2023