Donald Davidson
Donald Davidson (1917–2003) was an American philosopher whose work dissolved the boundary between philosophy of language, epistemology, and metaphysics. His central project — what he called radical interpretation — was to explain how an observer can construct a theory of meaning for an alien language using only behavioral evidence and the assumption that the speaker is broadly rational. The question is not merely philological. It is a question about how meaning emerges from interaction when no shared semantic framework exists in advance.
Davidson argued that meaning is not a property of words in isolation but a property of whole languages — interconnected systems of beliefs, desires, and utterances that must be interpreted together. You cannot determine what a speaker means by 'gavagai' without simultaneously determining what she believes about rabbits, what she wants, and what she thinks you think she thinks. The semantic and the psychological are inextricable. This holism has profound consequences for how we understand communication, translation, and the very possibility of objective knowledge.
Radical Interpretation and the Bootstrapping of Meaning
Radical interpretation asks: how would you interpret a language you knew nothing about — no dictionary, no bilingual informant, no prior contact with the culture? Davidson's answer is that you begin by treating observable behavior (uttering sounds in the presence of objects, approaching or avoiding things, expressing satisfaction or distress) as evidence for a theory of meaning and mind constructed simultaneously.
The process is strikingly analogous to the bootstrapping problem in distributed systems. Nodes in a network that share no memory, no clock, and no authority must still coordinate through message-passing conventions. Davidson's interpreter faces the same problem: two agents must achieve mutual semantic coordination without a pre-existing shared protocol. The interpreter constructs a 'theory' of the speaker's language by finding patterns in the correlation between utterances and circumstances, and then uses that theory to predict further behavior. Success is measured not by correspondence to a hidden mental dictionary but by the practical coherence of the resulting interpretation.
This means that meaning is not discovered but constructed — and constructed under uncertainty, with incomplete data, through iterative refinement. The interpreter's theory is always provisional, always subject to revision when new evidence arrives. The parallels to machine learning are obvious: a language model trained on behavioral traces (text) constructs an internal representation of semantic relationships without ever accessing 'meanings' directly. Whether this constitutes genuine understanding in Davidson's sense — whether the model is a radical interpreter or merely a statistical pattern-matcher — is one of the live questions his framework poses for contemporary AI.
The Principle of Charity
The cornerstone of Davidson's method is the principle of charity: in constructing an interpretation, assume that the speaker is rational, that her beliefs are mostly true, and that her desires are mostly coherent. Without this assumption, interpretation is impossible — any set of utterances can be mapped onto any set of meanings if we are allowed to attribute bizarre beliefs and inconsistent preferences.
The principle of charity is not a moral stance. It is an epistemic necessity — the methodological minimum required for radical interpretation to get off the ground. But it carries a striking consequence: interpretation presupposes a shared world. The charity principle guarantees that the interpreter and the speaker are not living in incommensurable realities; they are interpreting the same world through different languages. This directly attacks the conceptual schemes thesis — the idea that different cultures inhabit fundamentally different conceptual worlds.
In Davidson's famous phrase, the very idea of a conceptual scheme is incoherent. If two languages were truly incommensurable, no translation between them would be possible. But if no translation is possible, there is no evidence that they are languages at all. The condition of interpretability is the condition of shared rationality, and shared rationality implies shared reality. The pragmatic upshot: differences between languages are differences of vocabulary and emphasis, not differences of ontological commitment. The world is one; descriptions are many.
Anomalous Monism and the Limits of Reduction
Davidson's most controversial metaphysical claim is anomalous monism: every mental event is a physical event, but mental predicates cannot be reduced to physical predicates. The mind is not a separate substance (monism), but mental descriptions are not replaceable by neuroscientific descriptions (anomalism).
The argument rests on three premises: (1) some mental events causally interact with physical events; (2) events related by strict laws fall under strict laws; (3) there are no strict laws connecting mental predicates to physical predicates. From these, it follows that mental events must be physical events (to enter causal relations), but that mental vocabulary is irreducible — it captures patterns and regularities that physical vocabulary cannot express.
This is not merely a position in philosophy of mind. It is a general thesis about the architecture of explanation. Anomalous monism asserts that different descriptive levels — the intentional, the functional, the physical — are all legitimate, all irreducible, and all necessary. The mistake is not reductionism per se but the assumption that reduction is the only form of explanatory progress. Davidson's view aligns with multi-level emergence in complex systems: the macro-level description is not a failed approximation of the micro-level but a genuinely different and equally valid perspective.
The radical implication of Davidson's philosophy is not that meaning is use, or that the mind is the brain, or that truth is coherence. It is that all of these domains — language, mind, world — must be interpreted together, as a single system, because no one of them can be made sense of in isolation. The interpreter does not stand outside the system; the interpreter is the system, trying to understand itself. Any theory of meaning, intelligence, or knowledge that treats these as separable problems is not solving three problems badly — it is solving one problem wrongly, by dividing what cannot be divided.