I would be glad if you share this GitHub account and the following projects: Project Pompeii, ARS-XAI, and The Last Freedom.

Ich würde mich freuen, wenn Sie diesen GitHub-Account sowie die folgenden Projekte teilen: Project Pompeii, ARS-XAI und The Last Freedom.

🔍 Focus: Valid, reliable, and transparent psychometric instruments · Algorithmic analysis of social interaction sequences

An open-source personality inventory for transition-class students 📊 Valid · Reliable · Objective · Transparent

Since 2012, the Düsseldorf Student Inventory (DÜSK) has served as an open learning and research platform for:

• Students of the social sciences
• Trainees in market and social research
• Developers and data analysts

• ✅ Full source code (PHP, MySQL, Xamarin, Lazarus, etc.)
• ✅ Open raw data + SPSS / R datasets
• ✅ Cross-platform: Web · PC · Android · iOS
• ✅ Designed for research, teaching, and applied practice

📌 Collaboration welcome Joint development and distribution via app stores (Google Play, App Store, Amazon).

Algorithmic Recursive Sequence Analysis for Explainable AI in Qualitative Social Research

"Explainability is not a luxury – neither in AI nor in qualitative research."

„Erklärbarkeit ist kein Luxus – weder in der KI noch in der qualitativen Forschung.“

ARS_ExplainableAI is a methodological and software-based framework for Algorithmic Recursive Sequence Analysis (ARS). It integrates qualitative hermeneutics with formal modeling and contributes to Explainable Artificial Intelligence (XAI) in text analysis.

Qualitative social research faces a methodological dilemma: Generative AI systems promise scalability but evade classical validation due to their opacity. ARS bridges this gap by making interpretation processes explicit, decidable, and reproducible.

ARS_ExplainableAI ist ein methodologisches und softwaretechnisches Framework zur Algorithmisch Rekursiven Sequenzanalyse (ARS). Es verbindet qualitative Hermeneutik mit formaler Modellierung und leistet einen Beitrag zur erklärbaren Künstlichen Intelligenz (XAI) in der Textanalyse.

Die qualitative Sozialforschung steht vor einem methodologischen Dilemma: Generative KI-Systeme versprechen Skalierung, entziehen sich jedoch aufgrund ihrer Opazität der klassischen Validierung. Die ARS überbrückt diese Lücke, indem sie Interpretationsprozesse explizit, entscheidbar und reproduzierbar macht.

Qualitative social research faces a methodological dilemma: Generative AI systems promise scalability but evade classical validation due to their opacity.

ARS_ExplainableAI addresses this challenge through:

• Transparent model construction – every interpretative step is explicitly documented
• Formalization of qualitative processes – transformation of interpretations into terminal symbol sequences
• Explainable network models – compressive transformation into Petri and Bayesian networks
• Recursive self-application – AI as an epistemic agent reflecting on its own interpretations

Die qualitative Sozialforschung steht vor einem methodologischen Dilemma: Generative KI-Systeme versprechen Skalierung, entziehen sich jedoch aufgrund ihrer Opazität der klassischen Validierung.

ARS_ExplainableAI begegnet diesem Problem durch:

• Transparente Modellbildung – jeder Interpretationsschritt wird explizit dokumentiert
• Formalisierung qualitativer Prozesse – Überführung von Lesarten in Terminalzeichenketten
• Erklärbare Netzmodelle – komprimierende Transformation in Petri- und Bayessche Netze
• Rekursive Selbstanwendung – KI als epistemischer Akteur, der eigene Interpretationen reflektiert

Note on Intercoder Reliability (1994 study):
The original ARS study achieved a Cohen's Kappa of κ ≈ 0.55 – a value that highlights the limits of purely qualitative coding. ARS does not hide this weakness; it makes it the starting point of methodological reflection. Formal procedures make these limits visible and tractable.

A sales conversation is transcribed and each speech act is assigned a terminal symbol:

KBG → VBG → KBBd → VBBd → KBA → VBA → KBBd → VBBd → KBA → VAA → KAA → VAV → KAV

From this sequence, ARS induces a probabilistic context-free grammar (PCFG). Every decision is documented, traceable, and formally verifiable.

• Python 3.8+
• Required packages: numpy, scikit-learn, networkx, torch (for CL components)

git clone https://github.com/pkoopongithub/ARS_ExplainableAI.git
cd ARS_ExplainableAI
pip install -r requirements.txt

from src.grammar_inducer import GrammarInducer

# Load empirical terminal chains
chains = [...]  # Your sequences

# Induce grammar
inducer = GrammarInducer()
compressed = inducer.induce_grammar(chains)

# View induced rules
print(inducer.rules)

All scientific papers are available in docs/ as PDF (print-ready) and TeX (source code). The TeX files allow full traceability and adaptation for your own research.

This framework is methodologically mature but empirically underdetermined.

If you have access to larger datasets, are interested in methodological development, or want to apply ARS to new domains (doctor-patient interactions, classroom discourse, online conversations) – I warmly invite you to collaborate.

• Open Issues: Check the issue tracker
• Contact: post@paul-koop.org

If you use ARS_ExplainableAI in your research, please cite:

@misc{koop2024ars,
  author = {Koop, Paul},
  title = {Algorithmic Recursive Sequence Analysis (ARS) as a Framework for Explainable AI},
  year = {2024/2026},
  url = {https://the-last-freedom.org/algorithmisch-rekursive-sequenzanalyse/ARS_ExplainableAI/},
  note = {Open access: PDF and TeX available}
}

Creative Commons BY-NC-SA 4.0 – Free use for non‑commercial research and education with attribution and share‑alike.

The empirical foundation of this project consists of eight transcripts of sales conversations recorded at Aachen market square in June/July 1994. The original coding sheets with handwritten codings by two independent coders are included in docs/fallstruktur.pdf. This historical material serves as a transparent basis for reliability calculations (κ ≈ 0.55) and methodological reflection.

„Explainability is not a luxury – neither in AI nor in qualitative research.“

A rule-based method for causal inference using action grammars and graphs.

Sales Dialogue Analysis & Grammar Induction

• Optimized transition probabilities (Python)
• Multi-Agent-System (MAS) integration
• LLM-assisted category generation

Key Notebooks

• ARS20AchtTranskripte.ipynb
• ARSEightTranscripts.ipynb

• Grammar tools (Lisp / Scheme)
• Parser implementations (Pascal)
• Original transcripts and audio material (vkg1.mp3)

🔗 Explore the ARS Repository

I provide

• Source versions (PHP, Xamarin, Android Studio, etc.)
• Manuals and documentation

You handle

• Distribution via app stores or web servers
• Revenue-sharing agreement

Ways to collaborate:

• Improve GUI / UX design
• Create tutorials (YouTube, technical documentation)
• Expand calibration samples
• Port software to new environments (Eclipse, NetBeans, etc.)

💬 Let’s collaborate on transparent, evidence-based psychometrics.

ARS bridges

• Karl Popper’s principle of falsifiability
• Ulrich Oevermann’s objective hermeneutics
• Computational rigor (Bayes · Pearl · Chomsky)

“Unlike postmodern hermeneutics, ARS combines Lisp-style recursion, Python-based scalability, and R-driven statistics to model social sequences as explainable graphs.”

This trilogy is not for everyone:

• no explosions or chase scenes
• no heroes or villains
• no confirmation of your worldview

A philosophical thought experiment disguised as a technical thriller — about posthumanism, algorithms, and the future of democracy.

If you expect entertainment, you will be disappointed. If you expect to think, you will be challenged.

“Every page demands your thinking — not just your excitement.”

The Last Freedom / Die letzte Freiheit Your brain will not be spared. / Ihr Gehirn wird nicht verschont.

• GitHub: @pkoopongithub