IntDEx
An AI-native delivery experience framework that defines a seven-stage AI-native delivery process. Its core principles are evidence grading and treating self-reports as unverified, emphasizing the independence, paired verification and continuous verification of behavioral validation.
Timeline (last 90 days)
IntDEx is an AI-native delivery framework that defines a seven-stage process, in which the fourth stage is behavioral verification.
IntDEx's ranking of evidence sources is: execution evidence above human evidence, human evidence above cross-model evidence, and cross-model evidence above comparison with prior artifacts.
The IntDEx principle stipulates that a creating model's claim that a check passed carries no evidentiary weight and must be recorded as unverified rather than passed.
IntDEx employs paired behavior verification, for example pairing an end user with a tester, or a user with an analyst.
IntDEx's Automated Continuous Behavior Verification (ACBV) re-runs behavior evaluations when the engine configuration changes and uses the pass rate as the threshold for changes.
Applications of the IntDEx framework to date cover only a single project and a single primary evaluator, and are regarded as preliminary evidence of feasibility rather than universal proof.
The IntDEx framework was developed using a design science research methodology and views the framework as an artifact continually refined through application on a public health data experimentation platform.
IntDEx treats a checkpoint's zero rejection rate as a finding requiring investigation rather than a quality indicator.
All Facts (8)
IntDEx is an AI-native delivery framework that defines a seven-stage process, in which the fourth stage is behavioral verification.
50%UnverifiedIntDEx's ranking of evidence sources is: execution evidence above human evidence, human evidence above cross-model evidence, and cross-model evidence above comparison with prior artifacts.
50%UnverifiedThe IntDEx principle stipulates that a creating model's claim that a check passed carries no evidentiary weight and must be recorded as unverified rather than passed.
50%UnverifiedIntDEx employs paired behavior verification, for example pairing an end user with a tester, or a user with an analyst.
50%UnverifiedIntDEx's Automated Continuous Behavior Verification (ACBV) re-runs behavior evaluations when the engine configuration changes and uses the pass rate as the threshold for changes.
50%UnverifiedApplications of the IntDEx framework to date cover only a single project and a single primary evaluator, and are regarded as preliminary evidence of feasibility rather than universal proof.
50%UnverifiedThe IntDEx framework was developed using a design science research methodology and views the framework as an artifact continually refined through application on a public health data experimentation platform.
50%UnverifiedIntDEx treats a checkpoint's zero rejection rate as a finding requiring investigation rather than a quality indicator.
50%