The attention schema theory in a neural network agent: Controlling visuospatial attention using a descriptive model of attention

"In experiment 3, in all games, whether training games or test games, the attention schema was disabled. To disable it, we turned off all the pixels in the attention schema array. The agent therefore never received any information from the attention schema array about the state of its attention. The blue line in Fig. 5A shows the results from the first test category (Attention Score, Attention Normal). ... Compared, however, to the agent in experiment 1 that was trained with a functioning attention schema (Fig. 3A, blue line), the agent in experiment 3 (Fig. 5A, blue line) was severely impaired. The difference in performance is large compared to the SE."

Results, p. 6

Disabling the attention schema causally reduced the agent’s ability to route attention, demonstrating a gating/control role consistent with selective routing mechanisms in both brains and AI systems .

Figures

Fig. 3 (p. 5) : Performance drops when attention is disabled, illustrating how gating of information flow (attention) controls task behavior in the agent .

Fig. 4 (p. 6) : When the attention schema itself is ablated, performance degrades, evidencing a causal role for the schema in controlling routing/attention .

"In the same manner, according to AST, we “know” about how our cortex is allocating resources to select items, because higher cognition has access to some of the information in the automatically constructed, lower-order attention schema. In AST, because the attention schema is a simplified model of attention, a kind of shell model that lacks the implementation details of neurons and synapses and pathways in the brain, it provides us with the unrealistic intuition that we have a nonphysical essence of experience, a feeling, a kind of mental vividness, that can shift around and seize on or take possession of items. In AST, the information in the attention schema is the source of our certainty that we have a nonphysical essence of experience."

The Relationship between the Attention Schema and Subjective Awareness, p. 8

This passage articulates a self-model that grounds reportable awareness in a simplified internal representation of attention, aligning with self-model and reportability markers in consciousness research for AI and brains .

"At the beginning of each game, the agent had an empty 200 × 10 pixel short-term memory buffer (corresponding to the 20 × 10 input array shown in Fig. 1 over 10 timesteps). ... Since the short-term buffer contained information not just from the current timestep, but from previous timesteps, the trajectory of the ball could be inferred and the task was rendered solvable. ... The agent was trained over the course of 1,500 iterations."

The Training Procedure, p. 3

The explicit attention schema array and temporally structured memory define the agent’s representational geometry for control, illustrating how structured internal representations support access and computation in AI models analogous to population codes in brains .

Figures

Fig. 1 (p. 3) : The figure makes explicit the representational format (attention schema array) that the agent uses to encode the state of attention, a concrete instance of representational structure in AI .

"If AST is correct—if subjective awareness in people indicates the presence of a model of attention—then the control of attention should depend on awareness. ... When a person has no subjective awareness of a visual stimulus, attention can still be exogenously attracted to the stimulus, and the degree of attention is not necessarily diminished (29, 30, 43–53). In that case, the person can no longer effectively control that focus of attention."

The Relationship between the Attention Schema and Subjective Awareness, p. 8

Summarizing convergent human findings, the authors highlight that conscious access is tied to the ability to endogenously route attention, aligning with selective routing as a consciousness-relevant control mechanism .