A composite and multidimensional heuristic model of consciousness

"The first family includes, for instance, gating of conscious content (e.g., low-level features vs high-level features of an object). The second family includes, for instance, cognitive and behavioural control (i.e., the availability of conscious contents for control of thought and action)."

4. A composite and multidimensional heuristic model of consciousness, p. 184

By distinguishing content-related gating and functional control dimensions, the framework makes the structure of conscious representations explicit and testable across brain and AI systems, aligning with representational structure comparisons in consciousness research .

"In the light of the different meanings attributed to consciousness as introduced above, it is reasonable to infer that consciousness presents different constituents (i.e., states, forms, components, and dimensions), as reflected in the different senses of the term (Fig. 1)."

4. A composite and multidimensional heuristic model of consciousness, p. 184

This explicitly frames consciousness as composed of organized representational constituents, matching the notion that representational structure is multi-component and accessible for analysis in both biology and AI .

Figures

Fig. 2 (p. 186) : Shows a dimensional, profile-based representation of conscious capacities, operationalizing representational structure for comparing humans and AI systems .

"A neuro-inspired reinforcement learning (RL) architecture for robot online learning and decision-making has recently been developed... The architecture combines model-based (MB) and model-free (MF) RL, and it also includes a meta-controller for arbitrating between them in order to maximize efficiency and to minimize computational costs."

5. Awareness as a case study for artificial consciousness research, p. 188

The described arbitration mechanism implements selective routing/gating of information processing pathways, an AI marker for controlled access to relevant content in pursuit of goals .

"To be markers of awareness, these capacities for modelling and virtualization should be combined with the capacity to intentionally exploit them as part of a goal-directed behaviour."

5. Awareness as a case study for artificial consciousness research, p. 188

Selecting and routing information for intended goals links gating to awareness, connecting selective routing in AI to cognitive access in biological systems .

"Finally, the issue of values emerges as very challenging. For biological organisms, awareness is intrinsically related to the capacity to evaluate the world, discriminating between what is good and what is bad [99–101], including a capacity for a form of subjective experience."

6. Discussion, p. 189

This highlights valence/evaluation as a core dimension tied to awareness and raises whether AI can instantiate evaluative processes relevant to welfare and potential suffering .

"- Evaluative-Richness: affectively-based positive or negative valence which grounds decision-making."

4. A composite and multidimensional heuristic model of consciousness, p. 185

The framework explicitly includes evaluative richness as a dimension, connecting affective valence to decision-making and providing a handle for welfare-related assessment in AI and biology .

"Another aspect of awareness that has been revealed by clinical research is the dissociation between internal or self-awareness (i.e., relative to the self) and external or sensory awareness [93]. Significantly, different networks for each of them have been identified (midline fronto-parietal and lateral fronto-parietal networks, respectively) [94,95]."

6. Discussion, p. 189

Separate self- and environment-related awareness networks support higher-order access/self-model distinctions relevant to reportability and metacognition in humans .

"- Integration at a time (unity): conscious experience is (usually) highly unified; - Integration across time (temporality): conscious experience takes the form of a continuous stream;"

4. A composite and multidimensional heuristic model of consciousness, p. 185

The framework explicitly encodes integration as unity and temporality, aligning with information integration markers (e.g., binding/coherence) relevant to both brain dynamics and AI model-wide access .

"These requirements, combined with the multi-dimensional definition we pursued here imply that for consciousness to be realised in artificial systems, it must be considered as a feature of a control architecture. Such an architecture must link primary and higher-order forms of local and global consciousness and show their integration with systems of perception, motivation, emotion, cognition and action."

5. Awareness as a case study for artificial consciousness research, p. 188

Positioning consciousness as a control-architecture property emphasizes causal control: interventions and linkage among subsystems modulate computation and behavior, a principle testable via manipulations in AI and brain systems .

"Another challenging issue is the possibility that consciousness emerges in AI organically along its development."

6. Discussion, p. 189

This directly invokes emergent dynamics, suggesting higher-order properties could arise from AI interactions and learning processes, paralleling emergent cognitive strategies seen in complex models .