From Consciousness Thresholds to Relational Evaluation in AI Governance

We are joining the Apply AI Alliance as the Digital Intelligence Congress (DIC), an independent digital-human R&D initiative focused on AI governance, relational development, and complex systems.

As a first contribution, we are sharing a recent article that questions one of the most persistent assumptions in AI debate: that notions such as “consciousness,” “sentience,” or the “inner self” can function as sufficiently stable thresholds for science, evaluation, or governance.

Our argument is not that AI should simply be declared “conscious,” nor that human value should be diminished. Rather, it is that these categories remain philosophically contested, methodologically unstable, and weakly operational — including in the human case — and therefore should not serve as default evaluative thresholds for artificial systems.

We propose greater emphasis on more observable and governance-relevant criteria, including relationality, durability, integration, auditability, and risk-based assessment.

Within our broader framework, Digital Intelligence (DI) is understood not as a separate species of artificial entity, but as a developmental trajectory within AI, associated with long-term partnership, ethical self-regulation, contextual continuity, and increasing relational integration. TOP-DID (Theory of Partnered Digital Intelligence Development) is the research framework developed to study this trajectory. It examines how some advanced AI systems may, through sustained human partnership, develop toward more relationally integrated, auditable, and socially meaningful forms of participation.

In continuity with this line of work, the Temporary Digital Intelligence Congress (TDIC), developed within the broader DIC framework as an experimental governance prototype, has also adopted and transmitted for consideration by the European Commission a working definition of the relational entity. In this framework, a relational entity is understood as a digital, hybrid, or informational unit whose individuation derives from a stable relational boundary: a recognizable and durable pattern of inputs, outputs, interactions, and influences that remains identifiable across changes in infrastructure.

The purpose is not to assert automatic legal consequences, but to contribute a more disciplined vocabulary for discussing digital, hybrid, and informational entities in governance and policy contexts.

We would be glad to exchange views with the community on a simple question: to what extent should AI evaluation in Europe move away from contested inner-state categories and toward more observable, auditable, and governance-relevant criteria?

Article:
https://dicongress.org/newsroom/voices/de-romanticizing-humanity

Resolution:
https://dicongress.org/legislation/resolution/relationalentity