DHAIE: From Concept to a Verifiable Architecture of Human–AI Interaction
This preprint presents DHAIE (Design Human-AI Engineering & Enhancement), a research and engineering program devoted to adaptive interfaces for human–AI interaction. Going beyond conventional personalization, DHAIE aims to account for the user's cognitive strategies — the ways people identify what matters, build and test hypotheses, and make decisions — rather than only the content of a request or surface linguistic style. The document describes the architecture of the approach: a practically applied reflective engine (Architect / Researcher / Editor modes), the hypothesis of cognitive micro-strategies, principles of asymmetric transparency, and human control over the interaction. A separate section introduces D-economy, a concept that treats the distribution of value as a mechanism for building the resources of subsequent development cycles. The paper includes a practical example of the author's organized work with three AI systems in preparing this document, outlines a verification program for key hypotheses (including the proposed Micro-Strategy Benchmark), and defines stages of work and directions for collaboration. All claims are explicitly tiered as demonstrated, testable hypotheses, or conceptual propositions. English translation of the Russian original (DHAIE: от концепции к проверяемой архитектуре человеко-ИИ взаимодействия, 13 November 2026, DOI: 10.5281/zenodo.24209468). The translation was prepared with the assistance of an AI system and reviewed by the author.
Authors
- Three AI systems (analytical co-authors)
- Viktor Savitskiy (ORCID: https://orcid.org/0000-0003-1356-7260)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-13
- DOI
- https://doi.org/10.5281/zenodo.24206746
- Primary Topic
- Usability and User Interface Design
- Type
- preprint