About

Next Insight Lab

Next Insight is the experimental lab of GRDprocess Sàrl, a Swiss research company developing structured frameworks for AI-assisted analysis. The lab is where theoretical research becomes operational — each tool translates a formal methodology into a testable prototype.

Why Structured Analysis Matters

Language models are trained on human text. That text carries implicit value systems, unexamined assumptions, and systematic conflations between distinct concepts. When a model produces an « objective analysis » of a normative question, it inherits these blind spots and presents them as reasoned conclusions.

The problem is not that AI models lack analytical capacity. The problem is that they lack procedural structure to separate what they know from what they assume — and to make that separation visible to the user.

Intellectual Foundations

Our work draws from several traditions in logic, argumentation theory, and philosophy of reasoning:

Informal Logic

Unlike formal logic which deals with deductive validity, informal logic studies the patterns of reasoning people actually use — including fallacies, presumptive inferences, and burden of proof. It provides tools to evaluate arguments that are neither purely deductive nor purely inductive, which is precisely the territory where normative claims operate.

Argumentation Schemes (Douglas Walton)

Walton identified recurring patterns of defeasible reasoning — argument from expert opinion, argument from analogy, argument from consequences, and dozens of others. Each scheme comes with a set of critical questions that, when asked, reveal whether the reasoning holds or collapses. This approach treats arguments not as true or false, but as presumptively acceptable until challenged.

Commitment Stores (Charles Hamblin)

Hamblin’s work on formal dialectics introduced the concept of tracking commitments over time within a dialogue. A participant who asserts P at time T is committed to P — and any later assertion incompatible with P creates a detectable inconsistency. This temporal dimension is absent from current AI safety approaches, which evaluate outputs in isolation rather than tracking coherence across exchanges.

Dialectical Systems

From Aristotle’s Topics through medieval Obligationes to modern dialogue logic, the dialectical tradition treats reasoning as an interactive process between parties with different roles. Analysis emerges from structured exchange — question and response, assertion and challenge, claim and justification — rather than from monolithic inference.

Value Theory and Axiological Analysis

Normative claims always rest on value premises, whether explicit or hidden. Drawing from philosophical value theory, we treat values not as abstract ideals but as operational principles that can be identified, classified, and mapped against each other. The goal is not to judge which values are correct, but to make the value structure of any position fully transparent.

Our Methodological Principles

• Value-agnostic by design — our frameworks map axiological positions without judging them
• Procedural over parametric — structure produces consistency, not model size
• Interactive over monolithic — user interaction is a correction mechanism, not a convenience
• Traceable over persuasive — every classification is justified, every change is documented
• Falsifiable over definitive — frameworks are treated as coherent and productive, not as final truth
• Bottom-up over top-down — conclusions emerge from observations, not imposed by assumptions

Distributed Intelligence

Beyond tool development, our broader research explores why current approaches to artificial general intelligence may be fundamentally limited. Scaling parameters within a single system produces better pattern matching, but not emergence. Emergence requires distribution — multiple interacting agents with distinct roles, feedback loops, and the capacity for genuine self-correction over time.

This perspective informs everything we build: our tools are designed as distributed processes where different components (decomposition, classification, interaction, consolidation) operate as distinct stages with explicit interfaces, rather than as a single prompt producing a single output.

The Company

GRDprocess Sàrl is an independent Swiss research company. Over 15 years of experience in data analysis across telecom, energy and industrial sectors inform our pragmatic approach to turning theoretical frameworks into operational tools.

Based in Valais, Switzerland.
UID: CHE-385.988.552

Contact: contact@nextinsight.org