Scoring MIRR without a dark-trait index

Research context

Measurement critiques warn that composite scores and model choices can change the meaning of dark-trait results. MIRR does not attempt to reproduce those scores. Its proposed summaries concern the six named practices only and remain provisional pending empirical evaluation.

Machiavelli development analysis

The initial scoring rule is intentionally inspectable. Responses use a one-to-five agreement scale. The current implementation requires all sixty answers. Within each domain, two oppositely keyed items are transformed as six minus the selected response, all ten keyed responses are averaged, and round((mean − 1) × 25) produces the displayed zero-to-one-hundred index. The direction is consistent: larger values indicate stronger endorsement of the named practice. The score is not an estimate of hidden intent. In particular, the system should never subtract this number from a maximum and label the result narcissism, psychopathy, manipulation, or interpersonal dangerousness.

The software currently produces no report from a partial questionnaire and does not offer item skipping. A future experiment might compare complete responses with an optional-omission format that requires at least eight items per domain. That is an unimplemented research proposal, not a validated rule. In MIRR, omission could be especially informative when someone has had no opportunity to repair a strained exchange. Researchers would need to distinguish that absence from reluctance to answer and avoid guessing from other domains or demographic information. Refusing participation is not evidence of hidden intent.

Interpretation should remain anchored to the response scale. A mean near the upper end means that the participant endorsed most of the proposed practice statements; it does not mean that their behavior has been independently verified. The zero-to-one-hundred display used in the current report is a response-scale transformation rather than a percentage of goodness or a percentile rank. The development version has no representative norms. Its reports should therefore avoid comparisons such as more considerate than most leaders or lower risk than ninety percent of adults.

There should be no overall interpersonal risk score. A composite would hide meaningful differences and invite uses the framework cannot support. For instance, someone might describe transparent persuasion but difficulty repairing a strained exchange. Averaging those answers into one reassuring number would lose the practical question. Domain contrasts can be shown as prompts, with a reminder that small differences may be noise. The interpretation should be conditional and should ask for examples before recommending an optional practice.

Version control is essential because even a minor wording change can alter what a score means. Saved reports should retain the exact item version, key, completion date, and missingness rule. Software tests can verify the arithmetic and ensure that no diagnostic labels appear in outputs. They cannot establish that the scale measures a coherent construct. A future technical report must separately evaluate structure, precision, stability, and usefulness. Until then, the transparent scoring specification is a documented design choice, not evidence that MIRR is ready for consequential decisions.

A participant-facing example should show the key directly: agreeing with a statement about diminishing another person’s achievement reduces recognition-balance endorsement. It does not reveal a diagnosis. This concrete explanation makes the meaning of reverse scoring easier to inspect than an unexplained composite.