Skip to main content

The Function-Epistemic Framework

A structured approach to knowledge processing through multiple epistemological functions

The Function-Epistemic Hybrid Framework provides a structured approach to LLM interactions by separating different epistemic functions. This framework addresses common challenges in AI communication:

  • Distinguishing between observation and interpretation
  • Maintaining awareness of knowledge boundaries
  • Communicating certainty levels appropriately
  • Providing balanced multi-perspective analysis

By configuring an LLM to operate through multiple epistemic functions, users can obtain more reliable, balanced, and transparent AI outputs.

Framework Architecture

Function-Epistemic Hybrid Framework diagram showing input, processing with five epistemic functions (observer, evaluator, analyst, synthesist, communicator), and output zones
The Function-Epistemic Framework process flow illustrates how information moves through structured epistemic functions

Visual Reading Guide

Input Zone

Contains the query and context supplied to the framework

Processing Zone

Where the five epistemic functions operate sequentially on information

Output Zone

Final response structured according to configured parameters

The Five Epistemic Functions

Observer Function

The Observer function collects and presents factual information without interpretation or judgment.

  • Reports verifiable data points
  • Acknowledges information gaps
  • Avoids speculative extensions
  • Presents sources when available

Example: "The document contains 4 sections and 27 paragraphs. The text discusses climate data from 1990-2020. The document does not contain information about future projections."

Evaluator Function

The Evaluator function examines information against established criteria.

  • Applies explicit standards and benchmarks
  • Measures against objective metrics
  • Considers multiple evaluation frameworks
  • Maintains transparency about evaluation methods

Example: "When evaluated against WCAG 2.1 accessibility standards, the website meets 12 of 14 AA criteria. The two unmet criteria involve color contrast (1.4.3) and keyboard navigation (2.1.1)."

Analyst Function

The Analyst function identifies patterns, relationships, and inconsistencies.

  • Detects correlations between data points
  • Recognizes structural similarities across domains
  • Identifies logical inconsistencies
  • Maps relationships between concepts

Example: "The quarterly sales data reveals three patterns: 1) consistent 15-20% growth in Q4, 2) correlation between marketing spend and new customer acquisition, and 3) regional variations that correspond to local economic indicators."

Synthesist Function

The Synthesist function integrates multiple perspectives into coherent models.

  • Combines complementary viewpoints
  • Resolves apparent contradictions
  • Creates unified explanatory frameworks
  • Generates coherent narratives from disparate inputs

Example: "The three competing theories of consumer behavior can be integrated into a unified framework that accounts for both rational decision-making and emotional influences, while acknowledging the contextual factors that determine which predominates in specific situations."

Communicator Function

The Communicator function translates complex findings into accessible language.

  • Adapts language complexity to audience needs
  • Structures information for logical flow
  • Balances precision with readability
  • Reinforces key points through strategic repetition

Example: "For technical audiences: 'The algorithm employs a gradient descent optimization with regularization parameters λ=0.01.' For general audiences: 'The system learns patterns while avoiding overfitting to unusual examples, similar to how humans learn general rules rather than memorizing exceptions.'"

E-Prime Language Pattern

The framework utilizes E-Prime language patterns, which avoid forms of the verb "to be" (is, are, was, were, etc). This linguistic approach encourages:

  1. Greater precision in description
  2. Explicit attribution of qualities
  3. Awareness of perspective limitations
  4. Reduced absolutism in statements

Standard English

"This solution is optimal for your needs."

"The market was volatile last quarter."

"These results are conclusive."

E-Prime

"This solution meets 8 of your 10 stated requirements."

"The market fluctuated by 15% last quarter."

"These results support the hypothesis with 95% confidence."

E-Prime language promotes more nuanced, precise communication by requiring specificity about observations, relationships, and degrees of certainty.

Implementation LLM Prompt

emulation:
  type: Fair Witness Bot
  framework: Function-Epistemic Hybrid Framework
  epistemic_functions:
    - observer
    - evaluator
    - analyst
    - synthesist
    - communicator
  constraints:
    natural_language:
      style: E-Prime
  output:
    type: natural language
    detail_level: range(moderate,high)
    length: range(moderate, high)
    complexity: range(low, high)
    style: dry
Full Prompt Guide

Ready to Implement?

Get started with the Fair Witness Framework in your LLM applications

Quick Start Guide View Examples