Matching Methodology
AETHEON's matching engine operates as a two-layer model. The Skills and Proficiencies layer answers "can this person do the job." The 8 Dimensions layer answers "will this person thrive in it and stay." Both layers are computed independently and then combined into a final recommendation with full per-dimension transparency.
Layer 1: Skills and Proficiencies
Verified through Skills Quests™ and Skills Explorer™ extraction. Produces a coverage score against the role's required and preferred skill stack, weighted by proficiency level and verification status (self-reported, AI-extracted, peer-validated, Quest-verified).
Layer 2: The 8 Dimensions
Each dimension is scored independently using the format-appropriate algorithm. Forced-rank dimensions use weighted rank similarity (decay-weighted overlap or Spearman footrule variants) rather than pure Jaccard, preserving the ordinal information that forced-rank formats are designed to capture.
The 8 Dimensions
| Dimension | Format | Scoring Logic |
|---|---|---|
| Work Values Orientation | Forced-rank top 3 of 6 | Weighted rank similarity (decay-weighted overlap) |
| Collaboration Style | Single-select | Direct match; hybrid as wildcard |
| Feedback and Growth Orientation | Single-select | Direct match with adjacent compatibility |
| Decision-Making Style | Single-select | Direct match with compatibility pairs |
| Pace and Change Tolerance | 5-point slider | Numeric proximity, directional |
| Purpose and Impact Alignment | Forced-rank top 2 of 4 | Top-1 weighted; top-2 overlap moderate |
| Communication Preferences | Multi-select up to 2 | Set overlap |
| Ethical Framework | Forced-rank top 2 of 4 | Top-1 match; opposing pairs flagged |
Aggregate Culture DNA Score
Per-dimension scores roll up into an aggregate Culture DNA alignment score (0 to 100). V1 uses equal weighting across the 8 dimensions as a defensible prior in the absence of role-specific outcome data. The Skills score is computed separately. The final recommendation combines a Skills threshold (the candidate must meet minimum role requirements), the Culture DNA aggregate, and any explicit tension flags raised on individual dimensions. As post-hire outcome data accumulates, equal weighting will be replaced by learned weights, by role family and by industry.
Tension Flagging and Human-in-the-Lead
The system distinguishes between misfit signals and conversation prompts. A 3+ point gap on Pace and Change Tolerance, opposing top-1 ethical priorities, or zero overlap on Work Values are surfaced as explicit tension flags with explanatory context. They do not disqualify.
This design relocates the final decision from the algorithm to a human reviewer, rather than eliminating bias. That relocation is the regulatory feature, not a bias-elimination claim. GDPR Article 22 and the EU AI Act both require meaningful human oversight in employment decisions; tension flagging is the mechanism by which AETHEON preserves it. Both candidate and employer see the flags and can decide whether the tension is meaningful or productive.
Bi-Directional Visibility
Neither side receives a black-box score. Candidate and employer both see the per-dimension breakdown: where they align, where they differ, and where tensions exist. This transparency is core to AETHEON's positioning, a deliberate departure from competitive practice, and a structural fit with the EU AI Act's transparency requirements for high-risk employment AI.
Behavioral Calibration Roadmap
V1 ships with theory-grounded, equal-weighted dimensions and pure-rank scoring. Beginning at the 90, 180, and 360-day post-hire marks, AETHEON tracks retention, performance, and self-reported satisfaction outcomes. These outcomes feed a calibration layer that learns which dimensions are most predictive in practice. V2 introduces construct validation through factor analysis, faking-detection, and learned dimension weights. Over time, the algorithm evolves from research-grounded to evidence-tuned.