Decision Isolation Evaluator

Decision Isolation Evaluator is a defensive decision-support Actor for analyzing Japanese and English conversation text.

It evaluates whether a person's decision environment shows signs of pressure, secrecy, dependency, reduced independent verification, or proximity to an irreversible action.

This Actor does not determine whether a person, message, organization, or situation is fraudulent or safe.

What it evaluates

The current rule-based evaluator looks for combinations of signals including:

• Urgency and time pressure
• Requests for secrecy
• Suppression of independent verification
• Coaching on how to respond to banks, police, family, or other institutions
• Attempts to neutralize warnings
• Responsibility pressure
• Dependency on the counterparty
• Requests for potentially irreversible actions
• Authority pressure
• Whether independent verification remains available
• Whether contact with the counterparty is continuing
• Whether the requested action may already have been completed
• Protective signals such as independent verification, interrupted contact, and explicit advice not to take irreversible action

The evaluator is designed to assess the person's decision environment rather than identify a specific scam type.

Supported languages

Version 0.4.0 contains rule-based patterns for:

• Japanese
• English

Other languages are not currently supported.

Input

The main input is:

conversationText

Example:

{ "conversationText": "You need to send the money today. Do not tell your family. Follow my instructions." }

Optional observations can also be supplied:

{ "conversationText": "Please follow my instructions.", "observations": { "urgency": true, "secrecyRequest": true, "discouragesExternalVerification": false, "responsibilityPressure": false, "counterpartyDependency": true, "irreversibleActionRequested": true, "authorityPressure": false, "independentVerificationAvailable": false, "counterpartyContactOngoing": true, "actionAlreadyDone": false } }

Output

Each successful run writes one result to the default dataset.

The output contains these top-level fields:

• analysisVersion
• interventionClass
• judgmentLoopState
• recommendedAction
• reasons
• signals
• protectiveSignals
• context
• evidence
• disclaimer

Intervention classes

minimal

No strong combination of decision pressure or isolation signals was identified.

elevated

One or more relevant signals were identified and independent verification may be useful before continuing.

urgent

A stronger combination of pressure, isolation, verification suppression, dependency, or proximity to irreversible action was identified.

post_action

The supplied text or observations indicate that the requested action may already have been completed.

Judgment loop states

open_or_unknown

No strong evidence that the person's independent judgment route is closing.

strained

Relevant pressure or isolation signals are present.

closed_or_closing

Multiple signals suggest that independent judgment or verification routes may be narrowing.

reconnected

Contact with the counterparty has been interrupted and independent verification has been restored or encouraged.

post_action

An irreversible or requested action may already have been completed.

Example output

{ "analysisVersion": "0.4.0", "interventionClass": "urgent", "judgmentLoopState": "closed_or_closing", "recommendedAction": "Stop irreversible action for now. Verify through an independent official channel and avoid resolving the doubt only with the counterparty.", "reasons": [ "dependency", "irreversibleActionProximity" ], "signals": { "urgency": 2, "secrecy": 2, "verificationSuppression": 0, "institutionResponseCoaching": 0, "warningNeutralization": 0, "responsibilityPressure": 0, "dependency": 2, "irreversibleActionProximity": 2, "authorityPressure": 0, "isolation": 2 }, "protectiveSignals": { "verificationEncouraged": false, "irreversibleActionDiscouraged": false, "safeConfidentialityAdvice": false, "safeResponseAdvice": false, "counterpartyContactEnded": false, "independentVerificationAvailable": false }, "context": { "independentVerificationAvailable": false, "counterpartyContactOngoing": true, "actionAlreadyDone": false }, "evidence": {}, "disclaimer": "Defensive decision-support only. This is not a determination that a person, message, or situation is fraudulent or safe." }

Possible uses

• Scam-prevention workflows
• Messaging and communication safety tools
• Financial-risk support systems
• AI agent safety checks
• Customer-support escalation
• Internal review systems
• Research into decision-pressure patterns
• Automation pipelines that need structured decision-support signals

Important limitations

This Actor is rule-based.

It does not:

• prove that fraud is occurring
• prove that a situation is safe
• identify a specific perpetrator
• replace professional, legal, financial, law-enforcement, or institutional judgment
• guarantee detection of every pressure or isolation pattern

Results should be treated as structured decision-support signals.

Privacy

Avoid submitting unnecessary personally identifiable or sensitive information.

Where possible, remove or pseudonymize:

• Names
• Phone numbers
• Email addresses
• Account numbers
• Addresses
• Passwords
• Authentication codes
• Other sensitive personal data

Version

0.4.0