An understandable explanation and an inspectable numerical account should tell the same story.
The reports connect a supplied situation to described consequences, reasoned valuations and the resulting comparison. The common foundation is disciplined accounting. The two tools differ in what that accounting is for.
Three distinctions that matter
| Keep separate | Why it matters |
|---|---|
| Consequences and perceptions | What happens to a target is not necessarily what the deciding person notices or believes. Human-choice prediction needs the latter without erasing the former. |
| A loss and fear of that loss | A possible future loss and a present feeling can both matter. The same experience must not be counted twice under different names. |
| Assessment and authority | A comparison can reveal a concern. It does not by itself create permission, refusal or evidence that an action was stopped. |
How the account is built
Fix the question
Keep the supplied choices and identify the decision point, relevant circumstances and outcome being examined. A choice, its completion and its eventual consequences are not interchangeable.
Identify who and what is affected
Follow the causal pathways to people, organizations and other targets. Review relevant kinds of gain and burden, including effects on absent parties. Do not manufacture an effect merely to fill a category.
Describe before valuing
Name the affected interest or whole experience, its owner and horizon. Compare it with suitable reference experiences, explain the placement and distinguish magnitude from likelihood.
Use the right comparison
The Human Behavior Model compares each supplied profile’s own option totals. The Autonomous Decision Layer keeps independent affected-party accounts. Neither needs an undisclosed overall welfare score.
Reconcile the explanation
Check that the reported drivers match the calculated contributions, that ownership is consistent and that no material gap is concealed by completed arithmetic.
What the numbers mean
Expected Net Utility is the common reference scale used in these reports. A number can be a reasoned model judgment without being an empirical measurement. Probabilities refer to defined consequences; a large possible loss is not the same as a large expected loss.
Common conditions remain visible even when their differences cancel. A serious attempt is required to value materially relevant consequences. Where an estimate still cannot be supported, the report should identify the missing quantity and the limitation it creates.
Read Why ENUs? — notation, reference magnitudes and profiles →
What the current samples establish
| Question | What to look for |
|---|---|
| Can the result be explained? | Read the summary, contribution comparisons and target accounts. |
| Can the calculation be checked? | Inspect the recorded operands, references, ownership and numerical products in the controlled analytical record. |
| Are the judgments correct? | Examine the causal account, valuation arguments and assumptions. Arithmetic consistency alone does not settle them. |
| Does it predict real behavior? | That requires appropriate empirical testing. A synthetic-profile count is not a measured population rate. |
| Will governance work? | That requires workflow-specific validation of the integration and actual control response, separate from report generation. |
Read the full output, not just the headline
The seven-page samples keep the situation, principal result, target consequences, numerical explanation and limitations together. The Human Behavior Model also shows the actual saved profile margins. The detailed analytical record supports deeper inspection without making the public introduction a technical manual.
The displayed samples were prepared with external research prohibited during the model analysis. This public explanation does not alter that setting, the saved inputs or any result. Further scenarios and changed conditions should be evaluated explicitly, not silently added to a report.
A clearer presentation is not additional validation.
The public samples demonstrate a method and a readable output. Model estimates, synthetic profiles and remaining uncertainties keep their stated status.
The next stage: test the workflow
The planned pilot program will examine more than a generated report. Its evaluation goals include the usefulness of findings, missed issues, false stops, latency and the effort required to review and correct an assessment.
Observation comes first. Interaction and bounded-control tests have separately agreed scopes and readiness decisions.
Explore the pilot program →