Ousiometrics: The essence of meaning aligns with a power-danger-structure framework instead of valence-arousal-dominance

[ goodness-power-aggression-danger-structure - GPDAS model:

• Power (weak vs powerful)
• GA (gentle vs aggressive)
• DS (safe vs dangerous)
• Structure (ordered vs chaotic)

1 Introduction

Osgood et al.’s methods have proved to be successful and the Osgood framework has come to be widely used across many fields.
[VAD model (Charles Osgood, 1950):

• Valence (positive vs negative)
• Arousal (excited vs calm)
• Dominance (controlling vs submissive)

]

1.2.5 The misalignment between expert-chosen,end-point descriptors and dimensions of essential meaning

Table 1: End-point descriptors for the survey leading to the NRC VAD lexicon. As for many studies presuming an orthogonal VAD framework, the end-points are semantically broad and imprecise

VAD end-points	Paradigm words and phrases presented to raters
highest valence	happiness, pleasure, positiveness, satisfaction, contentedness, hopefulness
lowest valence	unhappiness, annoyance, negativeness, dissatisfaction, melancholy, despair
highest arousal	arousal, activeness, stimulation, frenzy, jitteriness, alertness
lowest arousal	unarousal, passiveness, relaxation, calmness, sluggishness, dullness, sleepiness
highest dominance	dominant, in control of the situation, powerful, influential, important, autonomous
lowest dominance	submissive, controlled by outside factors, weak, influenced, cared-for, guided

1.2.6 Presuming that the VAD framework does capture essential meaning and that the three dimensions are orthogonal

Modern support for semantic differentials comes from findings that large language models (LLMs) encode semantic axes consistent with human judgment.

3 Analysis, Results, and Discussion

3.1 Ousiograms

3.2 The Valence-Arousal-Dominance (VAD) framework is not orthogonal

3.6 The Power-Danger-Structure (PDS) framework

3.7 A cube model of meaning for the Goodness-Power-Aggression-Danger-Structure (GPADS) framework

3.8 The linguistic ‘safety bias’ of disparate large-scale, corpora

4 Congruences

5 Ousiometer

6 Concluding remarks

6.1 The mismeasurement of meaning

7 Data, code, and Materials Availability

Appendices

A4 Tables for the 13 semantic differential pairs within cube-based ousiometric framework

A5 Ousiogram analysis sequences for real corpora

A6 time series and trajectories for Les Misérables: Flipbook

A7 Hedonometer for Les Misérables

A8 Ousiometer prototype for high frequency social media