Quantifying Human Traits: A Seven‑Dimensional Vector Model for Predicting Behavior

Introduction

Understanding people is notoriously difficult, a problem highlighted by the ancient proverb "Know the person’s face, not the heart." The article argues that accurate person‑assessment is increasingly critical in modern contexts such as startup co‑founder selection, core employee recruitment, and personal relationships.

Theoretical Framework: Personality‑Trait Vector Model

Core Attribute Definition

The model defines a person’s core attributes as a seven‑dimensional vector, each dimension derived from Zhuge Liang’s “seven methods of recognizing people.” The dimensions are:

志 (Goal) : values, life goals, moral judgments – observed through future‑oriented discussions and dilemma choices.

变 (Adaptability) : ability to handle pressure and change – judged by reaction speed and quality under sudden situations.

识 (Insight) : strategic thinking and judgment – reflected in the depth and angle of problem analysis.

勇 (Courage) : willingness to face difficulty – seen in whether one confronts or avoids challenges.

性 (Temperament) : emotional stability and self‑control – shown in behavior under relaxed or stressful conditions.

廉 (Integrity) : moral baseline and attitude toward profit – revealed by choices when faced with temptation.

信 (Trustworthiness) : consistency between words and actions – measured by promise‑keeping.

Each dimension is scored on a continuous range where higher values indicate stronger presence of the trait.

Five‑Level Scoring Standard

To ensure consistency, each dimension uses a five‑level rating (Excellent, Good, Medium, Poor, Very Poor) with concrete behavioral descriptors. For example, the “信” dimension assigns scores from 0.81‑1.00 for “always keeps promises” down to 0.00‑0.20 for “habitually lies.” The same structure applies to the other six dimensions.

Information Acquisition Paths and Weighting

Two complementary paths gather evidence:

Passive Observation (Signal Reception) : natural signals such as appearance, micro‑expressions, and social traces. This path offers high authenticity but low efficiency.

Active Probing (Signal Induction) : designed questions, tests, or scenarios that elicit targeted responses. This path is efficient but may provoke rehearsed behavior.

Evidence is further weighted by context using a four‑grade system (S, A, B, C) where higher pressure, higher stakes, and harder disguise increase the weight (e.g., S‑level weight = 1.0, A = 0.7, B = 0.4, C = 0.2).

Overall score for a piece of evidence is calculated as score = value × weight, where value is the dimension rating and weight is the context weight.

Case Studies

Passive Observation: Interview Scenario

A product‑manager candidate is observed for five minutes. Specific behaviors (steady entrance, focused eye contact, honest recount of a failure) are mapped to dimensions and assigned B or A evidence levels, yielding scores such as 0.70 for “志” and 0.75 for “识.” The “变” dimension lacks data and is flagged for later testing.

Long‑Term Observation: Team Member Evaluation

Over 12 weeks, a colleague’s actions (meeting summaries, project delays, voluntary task taking, a small reimbursement fraud, crisis‑time innovation) are recorded. Each event is linked to relevant dimensions and scored, e.g., a S‑level fraud reduces “廉” and “信” to 0.30, while a later crisis contribution raises “识” to 0.80. The analysis shows that minor unethical acts can dramatically shift the overall vector.

Active Probing: Partner Selection

A founder conducts five probing rounds with a potential co‑founder, each targeting a different ancient method (e.g., “问之以是非观志”). The candidate’s responses are evaluated, producing scores such as 0.65 for “志,” 0.70 for “变,” 0.55 for “勇,” 0.70 for “廉,” and 0.60 for “信.” The final recommendation is that the candidate fits a COO role (execution) rather than a CEO role (strategic decision‑making).

Bayesian Update: Ongoing Relationship Tracking

Using a weighted‑average Bayesian update, new evidence continuously refines the vector. High‑pressure, high‑stake, hard‑to‑mask situations receive greater weight, ensuring the model adapts to evolving behavior.

Application: Career Promotion Decision

Two internal candidates (甲 and 乙) are compared against a senior‑position requirement vector. Candidate 甲 scores higher on “廉” (0.90) and “识” (0.85), while 乙 excels in “变” (0.85) and “勇” (0.80). The recommendation varies with organizational phase: during stable periods, choose 甲 for safety; during transformation, 乙’s adaptability is preferred.

Model Limitations and Ethical Principles

Limitations include incomplete information, personal growth, observer bias, and potential performance‑acting by the observed. Ethical guidelines emphasize goodwill (assessment for cooperation, not manipulation), openness (allowing revisions), respect (people are not mere numbers), and self‑reflection (examining one’s own biases).

In summary, the seven‑dimensional vector model merges classical wisdom with quantitative scoring, offering a structured yet flexible tool for systematic person‑assessment across hiring, partnership, and promotion contexts, while acknowledging its approximative nature and the need for ethical use.