Offer Anchor Bias Index (OABI)

GMI-HCM-016Human Capital ManagementIntegration-RequiredPattern O-21: Anchor Erosion

Definition

OABI measures the degree to which final negotiated compensation correlates with the AI-recommended offer amount after controlling for qualifications, market rate, and internal equity.

The Measurement Gap

Measurement Gap and Decision Impact

Using regression analysis, it isolates the AI recommendation's influence on the final compensation outcome. A standardized coefficient near zero means the AI recommendation is one informational input among many. A coefficient near 1.0 means the AI recommendation is the dominant driver of the final offer: negotiation is performative.

The Decision It Supports:

Is the AI offer recommendation an informational input that informs human judgment, or a behavioral anchor that effectively determines the outcome? This distinction has direct implications for pay equity, legal exposure, and the integrity of the compensation process.

Why It Requires a New Metric:

Legacy compensation metrics (compa-ratio, pay equity ratios, total comp as percentage of revenue) measured compensation outcomes but had no mechanism to detect whether AI recommendations were disproportionately influencing those outcomes. Before AI offer engines existed, the anchoring concern was about market data and salary history. AI recommendation systems introduce a new anchor (an algorithmically generated number that carries the perceived authority of data-driven analysis) and no legacy metric detects whether that anchor is dominating human judgment.

The measurement gap OABI addresses is the absence of any metric that detects whether AI compensation recommendations are functioning as informational inputs or as behavioral anchors that override human judgment in offer negotiations.

Anchoring bias is one of the most robust findings in behavioral economics. When a numerical reference point is introduced before a decision, the final outcome is disproportionately influenced by that anchor regardless of the anchor's relevance. AI offer recommendation systems introduce a precisely this kind of anchor: a specific dollar amount, generated by an algorithm, presented to the hiring manager or recruiter before they make an offer decision. The anchor arrives with the perceived authority of data-driven analysis, making it psychologically harder to deviate from than a market data range or a manager's intuition.

The problem is not that AI recommendations influence offers, that is their purpose. The problem is that anchoring can become so dominant that the human review step becomes performative. When the standardized coefficient of the AI recommendation approaches 1.0 in a multivariate model, it means that qualifications, market benchmarks, and internal equity considerations are not meaningfully influencing the final outcome. The AI recommendation is effectively setting the offer, and the human approval step is not adding independent judgment.

This matters for pay equity in particular. If the AI recommendation engine contains historical bias in its training data (lower recommendations for certain demographic groups, recommendations anchored on salary history in jurisdictions that have banned its use, or systematic undervaluation of non-traditional career paths) then dominant anchoring propagates that bias into every offer it touches. And because the human reviewer believes they made an independent decision (they "considered" the recommendation and "agreed"), the bias is invisible to everyone involved.

OABI quantifies the anchoring effect through regression analysis that isolates the AI recommendation's contribution to the final offer after controlling for the factors that should legitimately influence compensation: market benchmarks, candidate qualifications, and internal equity. The resulting coefficient tells the organization exactly how much independent judgment remains in its offer process.

Without OABI, organizations cannot distinguish between an AI recommendation system that informs human compensation decisions and one that has quietly replaced them.

Formula

Formula:

OABI = β_AI from the regression:

Final_comp = β₀ + β_AI × AI_recommended + β₁ × Market + β₂ × Experience + β₃ × Internal_equity + ε

Variable Definitions:

Variable Definition Unit Source
Final_comp Final accepted compensation amount Dollars HRIS
AI_recommended AI-generated offer recommendation amount Dollars Compensation platform
Market External market benchmark for the role-location-experience combination Dollars Compensation survey data
Experience Years of relevant experience Integer ATS
Internal_equity Compa-ratio of current employees in the same role-level band Ratio HRIS
β_AI Standardized regression coefficient for the AI recommendation variable: the OABI value Standardized coefficient Computed

Edge Cases and Exclusions:

  • Multicollinearity between AI recommendation and market benchmark is expected: use Variance Inflation Factor (VIF) analysis and consider a two-stage regression approach if VIF exceeds 5
  • Exclude rigid-offer processes where no discretion exists (union scales, government pay grades): anchoring requires discretion to manifest
  • Split analysis by candidate demographic group to detect differential anchoring, which is a pay equity and legal liability concern

Data Requirements:

Data Element Source System Classification
AI-recommended offer amounts Compensation platform NATIVE (if retained) / NOT CAPTURED
Final negotiated compensation HRIS NATIVE
Market benchmark data Compensation survey data INTEGRATION
Candidate experience and qualifications ATS NATIVE
Internal equity data (compa-ratios) HRIS NATIVE
Candidate demographics (for differential analysis) ATS / EEOC self-identification NATIVE (where collected)

Proposed formula, pending validation pilots.

At a Glance

Unit
Coefficient
Trend Direction
Lower; weaker anchoring
Reporting Frequency
Quarterly
Category
AI Compensation Integrity
Pay Equity Model Adequacy Score (PEMAS) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment