Scoring without trust scores, pass marks, or false precision

Research context

The honesty literature leaves important measurement questions open, while method-bias research explains why shared response procedures can distort relationships. Neither supports a universal trust score. MIO’s proposed scoring is deliberately descriptive, with the limits of self-report stated alongside any numbers.

Machiavelli development analysis

The development scoring specification starts with a simple rule. For a five-point response from one to five, an oppositely keyed item is transformed using six minus the response. The current software requires all sixty responses, averages all ten keyed items in each domain, and reports round((mean − 1) × 25). This produces an integer response index from zero to one hundred. This arithmetic is easy to inspect, but arithmetic correctness is not psychological validity. The report should explain that a larger mean indicates stronger endorsement of the domain’s proposed practices within the chosen reference period. It should not translate the number into a probability that the participant is honest.

The current completion policy withholds the entire report until all sixty statements have answers; it does not estimate missing responses or offer item skipping. A different policy could be tested in a future research-only version: allow omission and calculate a domain only when eight of ten items are available. That alternative is not implemented and its threshold is not validated. Research would need to examine whether omitted disclosure or boundary items systematically alter the meaning of the remaining integrity-practice summary. Declining participation or stopping before submission must never be interpreted as a suspicion signal.

A display can be precise mathematically and misleading psychologically. Reporting several decimal places suggests a level of certainty the instrument has not earned. A rounded response summary, the scale anchors, and a plain explanation are preferable during development. The implemented zero-to-one-hundred transformation rescales the same information; it does not create a percentile. A percentile requires a suitable reference distribution and an explanation of whom it represents. MIO has no validated norms, so comparisons with an imagined average professional are not justified.

The proposal also rejects an overall pass mark. A threshold would turn a reflection tool into a classification procedure, requiring evidence about error rates, consequences, and the intended decision. There is no defensible cutoff for trustworthy or untrustworthy in this version. Domain contrasts may be discussed cautiously, but their uncertainty should remain visible. If a participant’s disclosure mean is slightly above their commitment mean, the report should not assert that the difference is meaningful. Measurement error and ordinary response variability could explain it.

Finally, scoring and interpretation must be versioned together. A saved report should identify the exact question set, key, date, and method used. Correcting a scoring defect should lead to transparent regeneration, while changing a construct definition should produce a new version. Automated checks can verify reversed items, missingness rules, and output boundaries, but they cannot certify the framework as scientific. A responsible report therefore pairs transparent computation with modest interpretation and an invitation to compare its suggestions against lived examples before acting on them.