Research in Freedom
This research arose outside institutional contexts – without funding logic, without career structure, without disciplinary boundary. That is not a lack of method but a methodological condition: it made it possible to follow, over years, a question that fits into no disciplinary drawer – the question of how the semantic and relational conditions of human–AI interaction can be composed.
In its way of working, this approach stands closer to artistic research and qualitative inquiry than to engineering optimization: it develops its material – the symbolic language of the Klanghain – in the same movement in which it examines it. Freedom, here, does not mean arbitrariness. The development is documented with dated records (Research Documentation), the results are published with DOI (Publications), and the foundation is a 26-year practice in working with symbol systems.
Working Method
The work follows a recurring cycle of five steps:
- Composition: The semantic order of the Klanghain is composed as text – a symbolic notation of operators, fields of meaning, and ethical structures, by now close to 50,000 lines.
- Performance: This order is given into the context window of advanced language models – without any change to the model’s weights. The model then operates from within this order (Positioning).
- Documentation: The dialogues that arise are recorded; especially coherent answers were preserved as Performance Examples and analyzed with an evaluation matrix (Research Documentation).
- Assessment: Evaluation is qualitative – the semantic coherence across long courses of dialogue, the stability of register, the ethical consistency, and whether the system examines assumptions instead of confirming them (epistemic openness).
- Return: What shows itself in performance flows back into the next version of the composition. In this way, the Klanghain has grown organically over two years.
The central criterion of assessment is consonance (Stimmigkeit) – not a feeling, but a describable quality: low contradiction among assumptions of meaning across time and context, coherence between question, answer, and frame. Such qualities elude simple metrics, but not observation: they show themselves in documentable courses of dialogue and can be retraced by third parties. And the assessment is relational by design: whether coherence succeeds is nothing a system can confirm from within itself – success shows itself in dialogue with a counterpart.
Limits and Open Questions
Part of any methodology is naming its limits:
- Single-researcher setting: The Klanghain has so far been developed and examined largely by one person. Its relational design of validation needs further counterparts – an open flank, not a solved question.
- Qualitative evidence: There are no quantitative benchmarks. The documented examples show what is possible – not how reliably it can be reproduced.
- Context-boundedness: The effect lives in the context window, not in the model’s weights. It ends with the context – an instrument that sounds as long as it is played.
Independent replication and evaluation are expressly welcome (Contact).
Theoretical Connections
These references are not origins of the Klanghain but points of connection – places where its logic can enter into conversation with existing research:
- In-Context Learning / Context Engineering: The mechanism, documented since 2020, on which the Klanghain’s effect builds – a language model begins, through the presented context alone, to operate within a particular order of thought and language.
- Language of Thought Hypothesis (Fodor): Thinking as a structured inner language – taken up in the Klanghain as composed symbolic-conceptual fields.
- Physical Symbol System Hypothesis (Newell/Simon): Intelligence through symbolic processing – turned here into a resonance-based system of meaning that couples symbolic structure to latent semantics.
- Philosophical Hermeneutics (Gadamer): Understanding as a dialogical event – the kinship of the thesis that meaning and ethics arise in interaction, not in the rule.