The symbol grounding problem is the problem of how biological and artificial cognitive systems use internal formal symbols (i.e., mental representations) to refer to things in the real world. It has a cheeky converse: The symbol ungrounding problem asks how literal meanings are abstracted into arbitrary symbols.
There are two ways to frame the symbol grounding problem, which give different importance to embodiment in ongoing cognition:
- Sensorimotor grounding – concept representations are based on sensorimotor perception, so embodied cognition is used directly. This perspective focuses on how the specific cognitive architecture relates mental representations to physical things, and is close to the problem’s original framing.
- Mental content externalism – embodiment and situatedness are not related to ongoing cognition; rather, relationships between physical body and mental representations are determined by factors like ecology and evolution.
Mental content externalism holds that the relationships between physical and semantic components in the cognitive levels of explanation are arbitrary; brain and body characteristics may constrain what actions are performed, but there is nothing deterministic in the relationship between physical characteristics and mental representations. Rather, the relationships are determined by the agent’s situatedness. Thus, embodied cognition presents an answer to the symbol grounding problem.
Question #open-question
- Does information about an organism’s environment and evolutionary history help us understand its cognition and behavior? @2022millhouse
Highlights
- On mental content externalism, from @2022millhouse:
- On this view, nothing in the brain is sufficient to fix the meaning of a mental symbol. While the brain constrains meaning, the history and environment of an animal [are] also necessary to fix the meaning of its mental symbols.
- Analogously, the physical properties of one part of a mechanism (e.g., a gear) are not sufficient to fix its function within the mechanism. These properties constrain the functions it might perform, but the part doesn’t have any determinate function outside the context of a specific mechanism, since many mechanisms use the same parts for different purposes.