Should I Buy It?
Where does the mind stop? At the skin? At the skull? The obvious answer conceals a serious philosophical question — and a direct practical consequence for how we think about cognitive tools.
The Clark & Chalmers thesis (1998)
Andy Clark and David Chalmers published “The Extended Mind” in Analysis in 1998, proposing what they call active externalism: the mind is not limited to intracranial processes. Some cognitive processes extend beyond the skull, into artifacts, tools, other people.
The founding quote: “Where does the mind stop and the rest of the world begin? The question invites two very different answers. One says that it stops at the skin, or perhaps at the skull. The other says that it ain’t necessarily so.”
Inga and Otto
The heart of the argument is a thought experiment.
Inga wants to go to the museum. She recalls it’s on 53rd Street. She consults her biological memory and goes.
Otto has Alzheimer’s disease. He carries a notebook where he records all important information. He consults his notebook for the museum’s address and goes.
Clark & Chalmers ask: what is the functional difference between the two cases? None. Inga believes the museum is on 53rd Street — and so does Otto, in the only sense that matters for explaining his behavior.
The conclusion: Otto’s notebook is his memory, not a memory aid. The difference is not one of degree — it’s a constitutive difference. Otto’s mind extends into the notebook.
The parity principle
Clark & Chalmers formulate the parity principle: if a part of the world fulfills a cognitive function that, if it were happening in the head, would be recognized as cognitive, then it is part of the cognitive system.
This principle erases the intuitive distinction between “using a tool” and “thinking.” The boundary is not physical — it is functional.
The four conditions of integration
For an external artifact to count as part of the extended mind, four conditions must be met:
- The resource is constantly available and easily accessible
- Output is automatically endorsed without systematic critical evaluation
- Content is easily accessible when needed
- Content was consciously endorsed in the past — the information was deliberately placed there
These conditions distinguish Otto’s notebook from a library book: one is constitutive, the other is an ordinary external resource.
Extension to LLMs
In 2025, Smart, Clowes & Clark published “ChatGPT, extended: LLMs and the extended mind” in Synthese — Clark himself applies his thesis to large language models.
LLMs fulfill several conditions of the parity principle: reasoning, planning, solution generation. The question is no longer “am I thinking with them?” but “under what conditions do they constitute a legitimate cognitive extension?”
A new type of extension
The Memory Logger bot does more than Otto’s notebook. It doesn’t just store — it narrates. It doesn’t return raw information — it contextualizes, connects, and structures it.
This is a cognitively different type of extension: not an auxiliary memory, but an auxiliary sense-making process. The distinction from Wegner’s Transactive Memory matters: Wegner describes the delegation and coordination of memory between individuals — distributed memory. Clark & Chalmers describe constitution: the external artifact is the memory, not merely a repository to consult.
Links with other concepts
The extended mind is the philosophical framework that legitimizes Transactive Memory: if the mind can extend into artifacts, it can extend into other minds. And it responds to Reconstructive Memory: if biological memory reconstructs and distorts, externalization is not an inferior substitute — it is sometimes a more faithful support.
Sources: Clark, A. & Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19. Smart, P.R., Clowes, R. & Clark, A. (2025). ChatGPT, extended: LLMs and the extended mind. Synthese.