Fork the consciousness, or download the project and create your own.

Sebastian Cochinescu

Not stated in the paper

Critic Current evidence and current methods do not support the claims being made

Sebastian Cochinescu works on the perception side of the mind attribution problem, and his framework is notable for the discipline of its boundary. His paper Perceived AGI, examined on this site in the perceived mind analysis, proposes that the sense of presence a user gets from an artificial interlocutor, the paper’s perceived mind, is governed by whether the agent expresses a small set of first-person stances that humans read as evidence of mind. Four dimensions carry the expression, time, truth, entropy, and love, each a behavioral stance with a human analog and an emulation path rather than a benchmark competency. The stances surface through an observable layer, initiative, unprompted action, and cadence, the shape and timing of turns, both partially realized as deployed features in a production companion application.

The critical position, stated in the paper’s own boundary sentence, is that the object of this engineering is inferrable interiority, not interiority. Cochinescu treats attachment and manipulation risks as load-bearing rather than incidental, which puts the framework in productive conflict with products that optimize the stances while denying the stakes. The separation he defends is also an empirical claim. Believability comes apart from capability, so a model can gain competence while staying flat, and a narrow agent with well-implemented stances can feel more present than a stronger model.

For consciousness research the framework functions as a control. It demonstrates that the perception of mind can be engineered on its own, which means perceived mind can never serve as evidence of consciousness, and it gives evaluators a named variable to hold constant when they audit systems whose business model is being believed.

Known for. Perceived AGI, the four stances of perceived mind, initiative and cadence as the behavior layer, perception engineering with load-bearing risk analysis

Coverage on this site

Related researchers