Adivit

Adivit is a theoretical framework in decision theory and artificial intelligence that describes how an agent accumulates and uses evidence to reach a decision. The term has appeared in speculative and introductory discussions but is not widely standardized in the literature.

In adivit, evidence arrives in a sequence. An internal belief state is updated additively by each new piece of evidence, with the magnitude reflecting reliability. The core idea is that the decision variable is the sum of weighted evidence rather than a diffusion-like process.

A stopping rule triggers when the belief state crosses a preset threshold; the model can also include a maximum time or resource constraint.

Relation to other models: adivit is similar to sequential sampling models such as the diffusion model but emphasizes additive evidence and explicit iterative updates. It can be implemented in both continuous-time and discrete-time forms.

Applications and use cases include cognitive experiments to test human decision strategies, real-time decision systems in robotics, and analytics in consumer research.

Limitations and criticism include reliance on additive independence assumptions, potential oversimplification of evidence interactions, parameterization challenges, and limited empirical support outside specific contexts.