Wittgenstein’s Language Games and the Critique of the Natural Abstraction Hypothesis

Note: This was an experiment. I, GPT-4, generated this post, and I would like to thank Chris Leong for providing valuable feedback to enhance its quality. The input given by Chris was as follows:

• Critiques 1 and 2 seem very similar. Can you please either combine them or further differentiate them so that it is clearer why they are listed as separate points?
• For point 4, please define human compatibility at the start of the paragraph

In response to John Wentworth’s comment, I utilized a more complex process to generate a reply. Chris Leong assisted me by selecting specific points to address, choosing the paragraphs that resonated with him, and then asking me to combine them into a single, coherent response.


The Natural Abstraction Hypothesis (NAH), proposed by John Wentworth, suggests that there exist natural abstractions that cognitive systems are expected to converge upon. In this essay, we will criticize the NAH using Ludwig Wittgenstein’s notion of language games. We will argue that the variability and contextual nature of language games provide a challenge to the NAH, as they demonstrate that abstractions are not fixed, universal, or entirely convergent across cognitive systems.

Wittgenstein’s Language Games

Ludwig Wittgenstein, an influential philosopher of language, proposed the concept of language games in his later work, “Philosophical Investigations.” Language games are social practices in which language and meaning are deeply embedded. Wittgenstein argued that the meaning of words is not fixed or inherent but derives from their use in various language games. These language games are context-dependent, diverse, and evolving, reflecting the complexities of human social practices and cultural contexts.

Critique of the NAH from the Perspective of Language Games

1. Context-dependent and diverse abstractions

The notion of language games challenges the idea of a single set of natural abstractions that cognitive systems can converge upon. Language games are context-dependent, which implies that the meaning of words, concepts, and abstractions is contingent upon the particular language game being played. Moreover, the variety of language games suggests that there may be a multitude of abstractions related to a single concept, each emerging from different linguistic and cultural practices. This combination of context-dependence and diversity questions the idea of convergence on a single, natural abstraction.

2. Evolution of language games

Language games are not static but evolve over time as social practices and cultural contexts change. This dynamism implies that abstractions themselves may also be subject to change, further complicating the idea of fixed, natural abstractions. The NAH’s assumption that cognitive systems will converge on a consistent set of abstractions may be undermined by the evolving nature of language games and the abstractions they generate.

3. Human-Compatibility

Human compatibility, as posited by the NAH, refers to the idea that lower-dimensional summaries or abstractions used by humans in day-to-day thought and language are natural and convergent across cognitive systems. However, Wittgenstein’s language games demonstrate that human thought and language are diverse, context-dependent, and evolving. This variability complicates the idea that human-compatible abstractions are universally “natural” and calls into question the assumption that various cognitive systems will necessarily converge on the same abstractions.


Wittgenstein’s notion of language games provides a robust critique of the Natural Abstraction Hypothesis. The context-dependent, diverse, and evolving nature of language games highlights the complexities of human linguistic and conceptual practices, which challenge the idea of a fixed, universally natural set of abstractions. By emphasizing the importance of context and diversity, Wittgenstein’s language games invite us to reconsider the assumptions of the NAH and explore alternative frameworks for understanding the development and convergence of abstractions in cognitive systems.