analiza aiqbench
This paper proposes a world-model framework where AI alignment is defined by maintaining processability between heterogeneous intelligence forms rather than forcing consensus on shared representations. By introducing the Multi-Phase Inference Mechanism, the authors argue that cognitive diversity persists because observations do not automatically become inferential targets without constructing sufficient state representations. For technical audiences, this shifts the focus from value convergence to designing systems that preserve distinct error-detection capacities while enabling communication under finite informational constraints.
Toward AI That Understands Self and Others: A World-Model Theory of Cognitive Diversity and Alignment
arXiv:2605.29930v2 Announce Type: replace Abstract: Modern societies possess more information than ever before, yet they do not converge toward a single shared understanding. The same events, facts, laws, technologies, or risks can be interpreted as evidence of freedom, danger, exclusion, injustice, responsibility, or unrealized possibility. Existing discussions often treat such disagreement as a conflict of values, preferences, or beliefs. This paper argues that disagreement is already a late-stage phenomenon. The central premise is simple but not trivial: observation is not yet inference. Not every observation becomes inferentially relevant, and not every possible object in an observation sequence becomes an estimation target. A possible target becomes admissible only when a state representation can be constructed that is approximately sufficient for prediction, evaluation, or action with respect to that target. This paper develops a world-model theory of cognitive diversity and alignment by reconstructing recognition as the construction of such approximate sufficient statistics under finite informational, representational, observational, and action constraints. It formulates this position as the Multi-Phase Inference Assumption (MIA) and defines its core internal mechanism as the Multi-Phase Inference Mechanism (MIM). The framework introduces alignment maps and transformation loss to analyze how heterogeneous world models communicate without being collapsed into a single representation. World-model alignment is therefore processability, not agreement: the design of AI systems that help heterogeneous forms of intelligence remain mutually processable while preserving their distinct error-detection capacities.
Abstract:Modern societies possess more information than ever before, yet they do not converge toward a single shared understanding. The same events, facts, laws, technologies, or risks can be interpreted as evidence of freedom, danger, exclusion, injustice, responsibility, or unrealized possibility. Existing discussions often treat such disagreement as a conflict of values, preferences, or beliefs. This paper argues that disagreement is already a late-stage phenomenon. The central premise is simple but not trivial: observation is not yet inference. Not every observation becomes inferentially relevant, and not every possible object in an observation sequence becomes an estimation target. A possible target becomes admissible only when a state representation can be constructed that is approximately sufficient for prediction, evaluation, or action with respect to that target. This paper develops a world-model theory of cognitive diversity and alignment by reconstructing recognition as the construction of such approximate sufficient statistics under finite informational, representational, observational, and action constraints. It formulates this position as the Multi-Phase Inference Assumption (MIA) and defines its core internal mechanism as the Multi-Phase Inference Mechanism (MIM). The framework introduces alignment maps and transformation loss to analyze how heterogeneous world models communicate without being collapsed into a single representation. World-model alignment is therefore processability, not agreement: the design of AI systems that help heterogeneous forms of intelligence remain mutually processable while preserving their distinct error-detection capacities.
87 pages. Revised version with a refined abstract emphasizing disagreement as a late-stage phenomenon, target admissibility, processability, and the methodological abstraction used to compare humans, AI systems, and institutional decision procedures under shared information-theoretic constraints
Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
arXiv:2605.29930 [cs.AI]
(or arXiv:2605.29930v2 [cs.AI] for this version)
Text integral sub licența deschisă a publicației · 277 cuvinte
