Original and Derived Intentionality

Words, maps, diagrams, computer states, and traffic signs can all be about things. Their aboutness appears derived from conventions, designers, users, or interpreters. Human thought is often said to possess original or intrinsic intentionality: beliefs mean what they do in their own right, not because another mind assigns an interpretation.

The distinction promises to stop a regress. If every representation means only because someone interprets it, the interpreter's thought must already mean something. But declaring biological minds original is not yet an explanation. Naturalistic and interpretivist theories challenge whether there is a sharp boundary at all: human content also depends on evolution, learning, language, and social attribution.

Paradigm cases of derivation

An ink mark means danger because a community uses it as a warning. A map line denotes a border because cartographic conventions and intentions connect the line with a region. A data structure represents a customer balance because a software system and its users assign fields that role.

Derived content has several features:

• another content-bearing practice explains it;
• the same vehicle could receive a different interpretation;
• misuse and correctness depend partly on conventions or purposes;
• absent interpreters or practices, the physical mark appears semantically inert.

The dependence need not be on a current reader. An unread book retains meaning because it was produced within a linguistic practice. Derivation can be historical and social rather than an occurrent act of interpretation.

The regress

Suppose an internal brain symbol means dog because a homunculus reads it as dog. The homunculus must possess a state that means dog, so the explanation repeats the problem at a smaller scale. A theory of mind cannot explain original content through an inner interpreter unless that interpreter's content is independently grounded.

The same regress threatens simple computational theories. Formal symbol manipulation is sensitive to shape or syntax. Semantics cannot be added merely by saying the machine "reads" a bit pattern as an instruction, since reading is the intentional capacity to be explained.

Naturalistic theories stop the regress with causal tracking, biological function, or autonomous consumer roles. Phenomenal-intentionality theories stop it with conscious presentation. Interpretivists deny that it must stop at an intrinsic fact inside one system.

Searle's biological intentionality

Searle distinguishes the intrinsic intentionality of conscious biological minds from the observer-relative intentionality of artifacts. A program has syntax assigned by designers and users; running the program does not by itself create semantics.

The position underlies the Chinese room argument. A person follows rules for manipulating Chinese symbols and produces appropriate answers without understanding Chinese. If the person and rulebook lack understanding, Searle argues, formal computation alone is not sufficient for original content.

The appeal to biology can be read in two ways. It may be an empirical hypothesis that specific biological causal powers generate consciousness and intentionality, or a stipulation that only biological systems have intrinsic content. Only the former is explanatory, and it requires identifying the relevant powers rather than citing carbon or neurons generically.

The systems reply

The standard response says the room's person does not understand Chinese, but the whole system - person, rulebook, memory, and operations - may. Components need not possess the property of the organised whole. A neuron does not understand English either.

Searle imagines the person internalising the whole system and still lacking understanding. Critics reply that first-person intuition from the component's perspective does not settle the system-level property. The debate turns on which causal and representational organisation constitutes understanding, not whether one rule-following component feels that it understands.

The systems reply is strongest when the organisation is connected to perception, action, learning, and correction rather than a static lookup. This motivates robot and embodied replies while leaving open whether such grounding yields consciousness.

The symbol-grounding problem

Harnad's symbol-grounding problem asks how symbols acquire content without every symbol being defined only through other symbols. A dictionary is a network of marks; a reader connects some of them to perception and action.

Grounding proposals tie basic symbols to sensorimotor categories learned through interaction, then build complex contents compositionally. This supplies causal contact and error correction. It does not automatically supply determinate reference or phenomenal understanding. Sensors can be interpreted as detecting many correlated conditions, and their categories depend on training and goals.

Grounding is therefore necessary on many naturalistic views but not sufficient by itself. Teleological history, consumer use, social practice, and inferential structure may all contribute.

Dennett's intentional stance

Dennett proposes an interpretive strategy. Treat a system as a rational agent with beliefs and desires when doing so yields reliable prediction. Belief is not an inner inscription carrying intrinsic semantic paint; it is a real pattern discernible from the intentional stance.

This dissolves the demand for an inner source of original content. A system has the beliefs that an ideal interpretation of its behaviour, environment, capacities, and history attributes to it. Content can be objective enough when the pattern supports compression and prediction, even if multiple interpretations remain possible.

Critics worry that interpretation makes mentality observer-dependent and assigns beliefs too liberally. Dennett answers that patterns are not invented at will: some attributions predict robustly and others fail. The residual question is whether predictive reality captures what a state means for the subject or only a useful observer-level pattern.

Radical interpretation

Davidson's radical interpretation derives a speaker's beliefs and meanings together from observable utterances and circumstances, constrained by truth and rationality. No prior semantic facts are read off one sentence in isolation. The interpreter maximises overall coherence and agreement while allowing local error.

Content is thus essentially embedded in a holistic pattern of worldly interaction and rational interpretation. The approach explains why meaning and belief cannot be assigned independently. It also creates indeterminacy: multiple interpretation schemes may fit all evidence.

Davidson treats this indeterminacy as compatible with all the objectivity content needs. Critics seeking intrinsic intentionality regard it as evidence that interpretation has not reached the underlying mental fact.

Degrees and chains of derivation

The original/derived distinction may be less binary than its examples suggest.

• A natural sign carries information without convention.
• An evolved signal has biological function.
• A trained neural state has learned consumer roles.
• A private mnemonic inherits content from a subject's intention.
• A public word inherits norms from a community and causal history.
• A generated token inherits linguistic structure while also entering new feedback and correction loops.

Each case depends on prior structures differently. Human thought itself is shaped by language, teachers, evolution, and environment. If dependence implies derivation, very little content is wholly original. A graded view asks how autonomous, world-sensitive, and internally norm-governed a representational system is rather than where meaning first appears from nothing.

Misreading and semantic constraint

Pure reader-conferral cannot explain misinterpretation. If a reader alone supplies content, there is no independent fact they can get wrong. Texts and system outputs constrain interpretation through syntax, causal history, inferential consequences, and public conventions.

This does not prove that the producing system has speaker meaning. Sentence meaning can be fixed by language while communicative intention is absent. A smoke alarm's signal can have a stable function and correctness condition without the alarm intending to warn.

The distinctions are:

Success at one level does not settle the next.

Artificial systems

Artificial systems sharpen the distinction because design and autonomous operation coexist. A model's outputs inherit conventions and training data, yet internal states can be causally responsive to new input, support systematic inference, guide tools, and receive correction from the world.

Three positions remain:

1. Strong derivation: all content belongs to designers and users; the system is an instrument like a book or map.
2. Functional originality: autonomous tracking, inference, and correction give the system content in its own right, regardless of consciousness.
3. Phenomenal originality: only conscious understanding supplies intrinsic content; functional success remains derived.

The LLM case study examines arguments from externalism, use, tool feedback, misreading, and reader conferral. Its central result is conditional: theories that readily grant content are precisely those that make content independent of consciousness.

Does original intentionality exist?

The case for original intentionality is that interpretation must ultimately answer to a state whose content is not borrowed. Conscious thought appears to provide such a state. The case against a sharp boundary is that conscious thought's content itself depends on causal history, public concepts, inferential role, and interpretation.

A layered naturalism can stop the regress without an intrinsic semantic atom. Basic tracking becomes representation through biological or learned function; networks gain conceptual role; social practices stabilise norms; consciousness supplies a distinctive mode of access. No layer creates meaning alone, and none requires a homunculus.

The alternative phenomenal view insists that the layers explain only conditions under which observers attribute content, not genuine presentation to a subject. The dispute therefore returns to consciousness and content: whether what-it-is-like is one contribution to meaning or its ultimate source.

Assessment

Original versus derived intentionality is not settled by pointing to obvious artifacts. The hard cases are systems with autonomous learning, world-sensitive correction, public language use, and uncertain consciousness. A useful account should distinguish signal, functional, sentence, speaker, and phenomenal levels rather than force one verdict for all of them.

The regress objection requires some non-interpretive ground for content. It does not decide whether that ground is consciousness, biological function, causal organisation, social practice, or a real pattern across them.

Selected references

• Davidson, Donald. Inquiries into Truth and Interpretation (1984).
• Dennett, Daniel C. The Intentional Stance (1987).
• Harnad, Stevan. "The Symbol Grounding Problem" (1990).
• Searle, John R. "Minds, Brains, and Programs" (1980).
• Searle, John R. Intentionality (1983).