# Methodology and evidence boundaries

## Audit of the replaced page

The former application contained 15 unattributed questions, distributed 3/3/3/6
across four broad domains. It called itself a “Research-Based Diagnostic” and
assigned “high emotional intelligence” from a raw-score percentage. The repository
provided no instrument citation, validation sample, reliability estimate,
justification for its 60/80 cutoffs, or support for individual predictions.
The six relationship-management items accounted for 40% of the possible raw\ntotal; each other domain accounted for 20%, without a weighting rationale.\nSelecting the lowest answer on every question still produced 20% of the maximum.\nRecommendations also included a fixed-duration pause claim without a citation.\nNo named official instrument could be established from that evidence. The
replacement makes no claim that the old items were copied from a specific test.

## Instrument selection

The user requested unrestricted reuse with credible measurement evidence.
[IPIP's own statement](https://ipip.ori.org/) places its items and scales in the
public domain for any purpose, without permission or a fee.

The replacement retains three **complete** scales from
[Barchard's IPIP key](https://ipip.ori.org/newEmotionalIntelligenceKey.htm):

| Source scale | Items | Alpha: women | Alpha: men |
| --- | --- | --- | --- |
| Attending to Emotions | 10 | .81 | .83 |
| Positive Expressivity | 9 | .81 | .79 |
| Negative Expressivity | 10 | .75 | .74 |

The [source reliability table](https://ipip.ori.org/newEmotionalIntelligenceTable.htm)
reports N=432 women and N=317 men. These are source-reported subgroup statistics,
not confidence intervals and not estimates for current users. Gender is not
asked or used in scoring. The published key contains nine Positive Expressivity
items; the table distinguishes this from an earlier ten-item version.

These three scales were selected for relevance to awareness and expression and
their more consistent reported internal consistency. The wider source includes
four other components with more variable or modest estimates; those are not
silently folded into a broad “EQ” total here.

No new short form was created within the retained scales. Original item wording,
positive/reverse keys, and the five published accuracy anchors remain intact.
The 29 items are interleaved in a fixed order. IPIP's
[administration guidance](https://ipip.ori.org/new_ipip-50-item-scale.htm) allows
flexible item ordering and explains that its inventories have no prescribed
standardized item order.

## Scoring

The [published rule](https://ipip.ori.org/newScoringInstructions.htm) is:

- Positively keyed item: use the selected integer from 1 to 5.
- Reverse-keyed item: use 6 minus the selected integer.
- Complete subscale: sum its keyed item values.
- Displayed mean: divide the complete total by that scale's item count.

Possible totals are 10–50 for the ten-item scales and 9–45 for Positive
Expressivity. The average is between 1 and 5. Missing or voluntarily
skipped items leave the corresponding scale unscored. No imputation, neutral
substitution, or conversion to zero occurs. Other complete scales can still
be shown. There is no combined score.

Positions on a 1–5 line refer to possible response values, not norms or
percentiles. The midpoint at 3 is not a population mean. Commentary about lower,
midpoint, or higher endorsement is descriptive educational interpretation, not
a validated cutoff. Neither end is automatically desirable.

## What the evidence does and does not support

[IPIP's construct guidance](https://ipip.ori.org/Finding_Scales_to_Measure_Particular_Constructs.htm)
explicitly cautions that Barchard's self-reports should be regarded as personality
measures rather than cognitive-ability tests. Accordingly, the application is
an emotional-style profile for adult learning. It does not measure tested EI
ability, accuracy of understanding emotions, regulation skill, empathy, mental
health, or suitability for a role.

Internal consistency is one kind of reliability evidence; it does not prove
stability, validity, or usefulness for a particular decision.
[IPIP's reliability and validity guidance](https://ipip.ori.org/ReliabilityValidity.htm)
calls for scale-specific validity research beyond a generic IPIP label.

Barchard (2003), *Does Emotional Intelligence Assist in the Prediction of
Academic Success?*, Educational and Psychological Measurement, 63(5), 840–858,
[doi:10.1177/0013164403251333](https://doi.org/10.1177/0013164403251333),
reported no incremental prediction of undergraduate academic success beyond
cognitive and personality measures for the EI measures examined. The study does
not justify workplace-performance predictions by this site.

This browser implementation, its chosen three-scale presentation, narrative
feedback, adult-learning exercises, and action-plan materials have **not** had
a participant validation study. Do not describe them as a validated EI test.
Software tests establish implementation accuracy, not psychological validity.

## Adult-learning use

Participation, question responses, partner discussion, and sharing are optional.
Do not collect or rank individual profiles. Some original statements concern
affection or personal expression; participants need not justify leaving them
unanswered. Treat reserve, expressivity, and cultural variation without moral
or ability labels.

Use concrete examples to test an interpretation, and invite other viewpoints
only with consent. Conversation exercises and if–then planning prompts are
original, unscored educational activities. Their feedback is not a psychometric
answer key. A later score difference is not proof of improvement.

## A responsible pilot before stronger claims

If this tool is studied with an adult-learning population, a useful protocol
would include:

1. Specify the intended construct, population, language, and low-stakes use.
   Decide in advance which interpretations would count as supported.
2. Use voluntary cognitive interviews to examine comprehension, accessibility,
   sensitive wording, and interpretation of the report.
3. Plan an ethically reviewed, consent-based study with appropriate sample-size
   justification, data minimization, and a pre-specified analysis plan.
4. Estimate reliability for each complete scale in that sample, with uncertainty;
   examine item functioning, internal structure, and stability when relevant.
   Do not calculate or display a participant's “alpha.”
5. Investigate convergent and discriminant evidence using justified measures
   and observable examples; study consequential effects and subgroup fairness.
   Distinguish self-description from demonstrated skill.
6. Evaluate narrative comprehension and workshop usefulness separately from
   scale-score properties. Pilot revisions need their own evidence.
7. Make claims proportionate to results and replicate in a relevant population
   before broader use. Do not infer a new overall EQ construct from three sums.

No validation data are silently collected by the application. Any future study
would require a separate, explicit consent and data-handling process.

## Source provenance

The canonical project data are in data/instrument.json. The test fixture
contains the retained original text and keys transcribed from the official
IPIP key. Tests protect wording, scale lengths, reverse coding, endpoint and
midpoint profiles, missing-data behavior, and import validation. Browser tests
exercise the actual participant flow. Public-domain source attribution is
retained independently of the CC0 dedication for new project material.
