Challenging Complexity: Interdisciplinary Dialogues on Digital Transformation (Slides)

Digital transformation has amplified the institutional influence of large language models (LLMs), yet their behaviour remains constrained by architectural limitations: they optimise for token prediction rather than purpose, lack agency, drift under uncertainty, and exhibit pathologies such as laziness, incoherence, and hallucination. This presentation introduces Fisher‑Generative Informational Realism (FGIF), a unified analytic framework that provides an external structural scaffold capable of stabilising and constraining the outputs of current AI‑LLM systems. FGIR integrates Fisher Information Field Theory, Imperative Theory, and Butterfly Theory to detect informational drift, quantify instability, and map branching future trajectories. Central to this approach is entropy‑aware governance: the capacity to sense rising disorder in information fields, identify early signs of systemic drift, and intervene before volatility cascades into institutional or societal failure. Because LLMs cannot form intentions, integrate feedback, or self‑monitor, FGIR supplies the missing structural properties of agency (goal‑orientation, recursive correction, and coherence maintenance) by transforming weak prompts into strong, high‑constraint inquiry structures. The framework demonstrates how governance must shift from behavioural patching to structural, entropy‑sensitive oversight, enabling human institutions to steer increasingly powerful but fundamentally non‑agentic AI systems toward coherent, resilient, and ethically aligned trajectories.