The Subclinical Radar: Mapping the Elements of Executive Risk

We do not assign clinical diagnoses. Instead, we anchor our technology in the Elemental Psychopathy Assessment — Short Form (EPA-SF), a peer-reviewed framework that breaks subclinical psychopathy into 18 behavioral facets nested under four broad factors: Antagonism, Narcissism, Disinhibition, and Emotional Stability. To build our system, we paired these established psychological metrics with spoken language from individuals answering targeted, cognitive load-inducing questions. This approach allowed us to train and validate an expansive suite of machine learning models. By tracking these traits—from overarching factors down to highly specific facets—we give boards and investors an objective view of leadership integrity long before a vulnerability impacts the balance sheet.

The Flaw in Traditional Measurement

Historically, organizations struggled to measure these traits in the C-suite. Traditional assessments rely on self-report surveys. Sophisticated leaders easily manipulate these tests—a phenomenon known as social desirability bias—rendering the results useless.

The Breakthrough: Acoustic Affective Computing

Arche AI solves this problem by analyzing acoustic vocal biomarkers instead of semantic text. We measure how leaders speak, rather than what they say.

When a leader masks their true intent, the effort requires measurable cognitive load. This load triggers the autonomic nervous system, causing involuntary micro-tensions in the vocal folds and unnaturally flat pitch percentiles. Our machine learning pipeline extracts 88 of these acoustic features from unstructured audio, such as quarterly earnings calls. Because these markers link directly to involuntary physiological responses, executives cannot fake them. To make the data immediately useful, our system translates the raw probabilistic outputs into a standard 1.00 to 5.00 Likert scale.

The Blind Spot of Prestige Bias

Understanding an executive's underlying personality is only half the equation; detecting when they actively deceive the market is the other. In a 2023 study published in the Strategic Management Journal, researchers (Hyde et al.) demonstrated that financial analysts frequently fail to detect CEO deception in real-time. Due to “prestige bias,” analysts often inadvertently reward deceptive executives with higher recommendations before the deceit becomes public knowledge.

To solve this, Arche AI uses an advanced iteration of the acoustic machine learning deception model originally developed for that study. We keep active deception detection and underlying personality trait extraction as two independent signals, providing a two-dimensional view of an executive's behavior.

Case in Point: The CarMax Earnings Call

Consider our analysis of the CarMax (KMX) Q3 2025 earnings call. Using our multi-model approach, we first established the overarching baseline traits of the executive team.

The “Subclinical Radar” immediately flagged a high-risk profile in the CFO, who exhibited a constant baseline of defensive anxiety (scoring an abnormally low 2.18 in Emotional Stability). While his words projected calm, the acoustic model detected the involuntary vocal micro-tensions associated with chronic stress and a defensive posture. Later, during an unscripted segment of the Q&A regarding SG&A cost leverage, the CFO retreated into vague corporate jargon—stating the company was “pivoting from building capabilities to leveraging and enhancing them.”

While human analysts accepted the platitude, our independent deception model registered the massive cognitive load of the evasion, logging a maximum Deception Score.

When an executive operating from a constant baseline of defensive anxiety actively attempts to deceive the market, the context of the call changes. Arche AI provides the objective, data-driven warning that allows boards to quantify risk before it materializes.

Conclusion: The Subclinical Radar

Executive risk management can no longer rely on intuition or one-dimensional financial models. Boards and investors cannot afford to mistake boldness for leadership, or confidence for competence. By pairing the EPA-SF framework with acoustic machine learning, Arche AI provides the objective metrics required for modern governance. This technology allows organizations to look past the executive mask and verify that those in power possess the cognitive alignment necessary for positions of trust.