The problem 0xEPS solves

As of 2023, global financial markets value the world’s assets at $246.8 trillion — more than 2x global GDP (~$115T). This implies massive bets on future growth and productivity.

Yet the systems making these bets — whether humans, institutions, or AI agents — operate without access to the real emotional and psychological inputs that shape economic decisions. Today’s models lack insight into:

Emotional states (e.g., fear, trust, optimism)

Social contagion and belief cycles

Personality traits influencing group behavior

This makes forecasts brittle, markets prone to overreaction, and AI agents vulnerable to acting on incomplete or misaligned signals. # Bayesian Personality Awareness Framework: Summary (v1)

Overview

This document outlines the modular framework for estimating the probability that an individual knows the personality types of other individuals within a known list. This probabilistic framework can be used for:

• Personality-awareness modeling within organizations or communities
• Incentivized identity verification and enrichment using Web3 tools
• Fine-tuning LLMs on social proximity and personality insight patterns
• Running gamified challenges to validate, correct, or crowdsource insights

Step 1: Baseline Priors (P0)

Defines the prior probability that a person knows the personality type of n other people.

Step 2: Role-Based Likelihood Multiplier (L)

Adjusts the base probability based on the individual's role/title.

Step 3: Social Graph / Org Proximity Multiplier (S)

Captures inferred proximity or likelihood of interaction between individuals.

Step 4: Pairwise Likelihood Matrix

For a given list of individuals, generate a matrix of probabilities that person A knows person B's personality type.

[ P_{A \to B} = \min(1.0, P_0(1) \times L_A \times S_{AB}) ]

• Diagonal values are ignored (self-awareness).
• S values are directional and estimated based on organizational data or heuristics.

Step 5: Posterior Probability Calculation

[ P(\text{Knows } n) = \min(1.0, P_0(n) \times L \times S) ]

All components from steps 1–3 are multiplied to derive a bounded posterior probability for knowing n personalities.

Step 6: Application Pipeline

Once values are estimated, they can be used to:

• Fine-tune LLMs with structured input-output mappings
• Preload likely known personality links in social graphs
• Bootstrap games where agents and humans guess each other's types
• Incentivize uploads or scraped data for personality claim verification

Game Layer Preview

Challenge Objective: Given minimal context (text, video, org data), players or agents attempt to match the correct personality types.

• Input Sources: Twitter bios, Series A memos, YouTube, LinkedIn, GitHub
• Agent Role: Analyze tone/semantic patterns to predict type (Enneagram, MBTI, OCEAN)
• User Role: Guess others’ types or claim/verify their own
• Reward Model: Blockchain incentives for accuracy, training set enrichment, self-verification

Sample Profiles Used (Selection)

Challenges I ran into

The biggest challenge I ran into was the shere size of integrations (11) per developer (1, me :)). It's never been easier to set these up thanks to AI, but the process of debugging more than three is a challenge even with such amazing help.