Odotat

Odotat is a theoretical framework in cognitive science and human-computer interaction that describes a method for tagging perceptual objects across multiple sensory modalities to support inference and decision making in artificial agents and human observers. It frames perception as a process of assigning structured object representations, called odotat vectors, to elements in a scene.

The name odotat is a coined term chosen to evoke the idea of tagging objects; it does not correspond to any single language root.

In odotat, each perceived object is associated with a multi-dimensional odotat vector encoding attributes such as spatial location, appearance, motion, temporal consistency, and a measure of confidence. The vector is updated as new sensory evidence arrives, typically using probabilistic or evidence-accumulation rules, allowing cross-modal fusion and resilient tracking under uncertainty.

Applications include robotics perception, virtual reality, and simulation environments where robust cross-modal object tracking is important. Implementations are often described as probabilistic graphical models, neural-network modules with tagging heads, or hybrid architectures that combine symbolic and subsymbolic components.

Odotat is primarily discussed in theoretical literature and in a small body of experimental work as a conceptual tool. Critics note that it remains underspecified in empirical practice, with challenges around scalability, interpretability, and validation across domains.

Related concepts include multi-modal perception, object tagging, Bayesian inference, and perceptual data fusion.