PROTEX

AI-supported research architecture for behavioral analysis

PROTEX is an AI-based profiling research project — more precisely, a multi-layer research architecture designed to study behavioral and decision-making patterns.

It is not intended to judge individuals, make predictions, or support operational or enforcement decisions.

This is not a finished product announcement, but a description of an ongoing research system, its current stage, and its intended direction.

Current stage: the fact extraction layer

The current development stage focuses on the first core module, commonly referred to as the fact extraction layer.

This layer is designed to extract factual elements from historical cases, connect those elements relationally, and analyze similarities, differences, and — critically — the absence of precedents.

This is not a simple database or keyword-based system. The goal is not information storage, but structuring data in a way that enables cross-case comparison at scale.

At this stage, the system is intentionally trained and tested on a small sample. The focus is on validating underlying logic, ensuring extraction stability, and verifying the correctness of relational mapping. This phase is documented as part of an ongoing research process.

Why this layer matters

Human experts are exceptionally strong at interpretation and contextual reasoning.

They are far less effective at simultaneously comparing hundreds of complex cases, tracking rare or weak correlations, or reliably identifying patterns that have never occurred before.

This is where AI can be genuinely useful — not as an authority, but as a cognitive support tool.

The fact extraction layer is the most constrained and controlled component of PROTEX. It operates strictly on data and relationships, not psychological interpretation.

Additional modules in development

PROTEX is not a single model. It is a multi-layer research architecture, where each module has a clearly defined role and clearly defined limitations.

1. Interpretative layer
Grounded in psychology, criminology, and behavioral science. This layer does not diagnose or evaluate individuals. It generates hypotheses and pattern-based observations that explicitly require human interpretation.

2. Scenario-based analysis module
This module works with fourteen structured scenarios designed to explore decision-making under uncertainty, tolerance for ambiguity, responsibility, control, empathy, norms, and risk.

The analysis is longitudinal. Its value does not come from a single response, but from patterns observed over time.

3. Synthesis layer
This layer integrates outputs from earlier modules and expresses results as degrees of similarity to known behavioral patterns — not classifications, labels, or diagnoses.

What PROTEX is — and is not

PROTEX is a research tool, a controlled experimental architecture, and a framework for studying how thinking and decision-making are organized.

PROTEX is not a policing tool, not a diagnostic system, and not an automated decision-maker.

Human oversight and interpretative responsibility are central by design.

Why this description is shared

The project is described at an early stage because parts of the system are already functional, and the research process itself is considered meaningful.

Complex systems are not only valuable once finished. Understanding how and why they are built is often more informative than observing the final result alone.

Ethics and responsibility

– non-fiction, publicly available sources only
– no personal data and no ongoing cases
– no diagnoses or operational use
– transparency, peer review, and human verification

Contact

Email
karol@protex-profiler.ai

Phone
+44 7367 000817