Philosophical Foundations of Learning: Insights from Wittgenstein

and Heidegger on AI Cognition

Zhifei He

College of Arts — School of Humanities, University of Glasgow, Glasgow, U.K.

Keywords: Wittgenstein, Existential Engagement, Language and Thought, AI Ethics, Situated Learning.

Abstract: This paper attempts to develop a perspective on how Wittgenstein and Heidegger can be fruitfully synthesized

to bear upon the questions involved in designing AI. Currently, when AI technologies are basically good at

track following and note making, they lose traits toward social and existential intelligibility. The loss comes

in comprehending meanings and intentions. Indeed, the “language games” concept of Wittgenstein tends to

emphasize that meaning is negotiated through dynamisms induced by social practices on the one hand, though

Heidegger argues on the other that understanding comes from engagement emanating from the body in the

world. The context with elements of learning and adaptation that AI has today is seen as limitations, and hence

the paper unearths philosophy-based methods that can be used to improve the performance of AI through real

social and environmental embedding.

1 INTRODUCTION

The stunning pace of artificial intelligence has

effectively elevated capacities pertaining to certain

miracles of language generation, image

identification, and autonomous systems. However,

the success of AI significantly relies on

representational models; in other words, the system

crutches enormous data to leverage patterns and

create outcomes that link statistically. Of course,

these approaches produce results but not as much as

human cognition does in terms of significance,

context, and intention. All these hereby claim a shift

in paradigm from data-driven pattern recognition

toward models of engagement, where AI systems

dynamically interact with the world and its manifold

complexities.

Both philosophers, Wittgenstein and Heidegger,

have put forward insights as to where the future

development of AI should lead. They push the theory

of representational AI toward the relational,

contextual, and existential aspects of being.

“Language games” explain how meaning is

constructed through practical use in shared practices,

while Heidegger, through the notion of “Being-in-

the-world,” stresses that understanding is acquired

through an engaged and embodied relationship with

an environment. These views highlight the limitation

of present-day AI systems but also serve as markers

on the road toward building machines that might

more humanly engage with the world. One of its most

glaring deficits is that AI cannot be social and

existential in creating meaning. According to

Wittgenstein and Heidegger, human language and

thought derive fundamentally from social praxis and

life experience. Meaning, for Wittgenstein, comes

from the common rules and practices of individual

communities—rules that are acquired actively, not

passively, through shared participation. Heidegger,

within an ontological framework of reference,

demonstrates for us that human interpretation is

impossible without context. Take an example, the

carpenter perceives a hammer not through abstract

knowledge about how it is made but through

knowledge about use and specifically how it is

employed within the work of building.AI lacks the

ability to engage with social and physical contexts,

making its operations rigid compared to human

cognition. Modern AI language models, such as GPT,

are very good at simulating human text generation,

but they completely lack an understanding of the

intentions, the cultural normativity, or even the

existential consequences of what they are doing. A

machine can produce a very grammatically correct

and seemingly relevant response to something deeply

sensitive but cluelessly miss out on the emotional or

even social nuance underpinning the conversation

because its building corpus does not account for it.

This is because AI operates within static datasets and

He, Z.

Philosophical Foundations of Learning: Insights from Wittgenstein and Heidegger on AI Cognition.

DOI: 10.5220/0013977700004912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Innovative Education and Social Development (IESD 2025), pages 231-238

ISBN: 978-989-758-779-5

231

representational logics that end up taking the world as

symbols and data points rather than engaging with it

as a dynamic and relational field.

Furthermore, Heidegger’s reflections highlight

another key aspect of interaction embodied with the

environment. While artificial intelligence in the

current period is increasingly abstract in nature, the

same intelligence, when Heideggerian, would possess

sensorial and relational sensitivities—enabling robots

to understand their environments contextually. The

confluence of philosophical thought between

Wittgenstein and Heidegger surfaced incredibly

substantial ethical considerations—to place the two

cultures at such a level of interface where the very

design of AI must consider respect for cultural

diversity, norms, and existential values. An AI system

trained to detect and make a joke at its own expense

must be able to appreciate various meanings that

could attach to the same expressions in a different

setting. Likewise, in the quest for AI that is

empathetic, it should be made sure that machines do

not take advantage of human emotions to manipulate

them. Therefore, these developments can be guided

by the challenges through a philosophy that

synthesizes such human-centered principles, to help

build systems that over all better the quality of human

life rather than undermining it.

This paper examines how philosophy can be

synthesized with AI to create a future in which

machines can interact with the world in a more

meaningful way. Wittgenstein and Heidegger,

through criticism of representation, stress context,

practice, and relationality, and this paper correlates

their arguments to some limitations and possible

routes toward an AI that can surpass mere intelligence

to be adaptive and sensitive to context and even be

ethically grounded. A good integration between

philosophy and technology can make it possible for

AI to evolve from its purely functional orientation to

the level of cognition and values that humans also

have.

2 IMPLICATIONS OF

PHILOSOPHICAL

FOUNDATIONS FOR AI:

LEARNING BEYOND DATA

Human learning is not just a linear process of pattern

recognition but active and reflective engagement with

the world. It is dynamic. Human cognition comes to

grips with new situations by constantly reflecting on

meaning and interpreting reality in ever-changing

ways. This makes human learners much more than

information gatherers and users in an adaptive

process of meaning seeking and reality making. The

Heideggerian "being in the world" sees purpose,

emotional states, cultural meanings, and lived

experiences all determine and be part of the learning

process through informal learning. However, much of

machine learning, especially in today’s AI, rests with

large-scale pattern processing in data. A scientific

theory of cognition and the possibility of a single

worldview have been hindered by the

phenomenology of AI and robotics, which has

ignored representational features (Gómez, Bravo,

2009). It is training machines to identify and then

output statistical correlations, based typically on

statistical relationships abstracted from any

meaningful context. These models learn rather

sophisticated syntax, picking up on n-grams

(sequences of words) and probabilities that allow

fluency text production.

The constraints regarding AI pertain to a lack of

real understanding. Ai can simulate intelligence

because it provides apt answers or executes tasks in

an efficient manner, yet this “knowledge” is entirely

utilitarian, not involving any reflective or conscious

thought. An output from the AI results from the

optimized algorithm, not from any reflection on the

content. Humans have a better understanding of

things; they perceive reality and can ask

metacognitive questions about their learning. Take,

for instance, a model like GPT. Such models do very

well in predicting the next word in a sequence, based

on a huge amount of data, or generating output that

sounds convincing. But they do not have a deep

understanding of the very essence and intentions of

humans influencing language use to express thoughts.

For Heidegger, AI stands apart from the realm of

meaning, unable to grapple with existence and what

it means for being in the world. When humans

articulate words using their language as an example

for empathy, humor, or sarcasm, behaviors closely

linked to social practice and lived experience,

language-using machines only issue token responses

to simulate these same expressions. They do so

without having a glimmer of what those words signify

or any concern about their meaning. This cognitive

limitation underscores the deeper philosophical

problem that AI systems are essentially

representational in their approach to learning. They

reduce reality to data and symbols, and thus never get

to grips with the world in a practical, intentional, or

empirical manner. Consequently, their "knowledge"

remains disconnected from that environment which

gives human understanding its depth and plasticity.

IESD 2025 - International Conference on Innovative Education and Social Development

232

These do not detract from the fact that the

philosophical orientations of Wittgenstein and

Heidegger can very usefully be pressed into service

to direct AI toward more contextually aware and

meaningful learning. By learning the philosophy of

meaning and conditions of the possibility of

understanding, the developers may eventually come

to engineer such AI that can more closely

approximate the specific faculties of human

reasoning.

3 WITTGENSTEIN’S LANGUAGE

GAMES: MEANING AS SOCIAL

PRACTICE AND CHALLENGES

FOR AI

The idea of “language games” was introduced by

Wittgenstein to basically dismiss the old theory that

language is a set of symbols operating within a fixed

system of rules. What he is actually trying to get at is

that meaning does not spring forth from matched

definitions or some sort of symbolic formalism, but

rather from the practical use of words within concrete

social relationships. It is from this view that language

becomes an ever-changing and active engagement

positioned within different “forms of life”—shared

and interacting cultural practices, customs, and ways

that ground the meaning of words. Take, for example,

the word “apology.” This term will have meaning

only as its usage is ascribed within a given culture and

specific relationship with reference to mending a

social relationship that has been broken by some

offense—not from any dictionary definition. The

human richness of language depends on parasitic

sensitivities such as tone, intention, or mere

relationality with established cultural standards.

Naturally, speakers adjust themselves to those games

for languages to which they belong by offering

appropriate utterances that those linguistic agents

might expect and conventionally perform.

Wittgenstein teaches us philosophical psychology

challenges the view that language is a device for

communicating independently constituted thoughts,

suggesting that language determines thought rather

than just communicating independently constituted

thoughts (Proudfoot, 2009), and with the concept that

language is a tool for conveying freely formed ideas

is contested by Wittgenstein's philosophical

psychology, which contends that language causes

thought rather than merely conveying independently

formed ideas. And its most fundamental level,

language is a practice that is social—it is not 'known'

or 'understood' through learned rules or patterns but

through active participation in situations where

meaning is constructed and negotiated.

Since the 1950s, post-Wittgensteinian philosophy

has shaped AI research, ignoring Margaret

Masterman's contributions to ordinary language

philosophy (Liu, 2021). Given this view of language,

it becomes clear that much as a large part of modern

AI might try, it does not understand language.

Modern AI processes language quite different from

the way data and statistics are used. Present AI is

based on recognizing patterns in a giant amount of

information stored in datasets; their outputs are

probability-based and not necessarily indicative of

true understanding or engagement with context.

While it might produce sentences that are

grammatically and semantically correct, the use of

language by AI fails to achieve any of the dynamism

of language games as posited by Wittgenstein and

because it does not dynamically operate within real

ongoing human social activities. For example, in a

discussion on ethics: human participants regulate

their speech in terms of the norms of the specific

culture, the moral principle involved, or even the tone

of feeling that is under consideration. AI would be

cautious with words when speaking on sensitive

matters; this will be applied as metaphors or hedging

words to avoid conflict. Such adjustments are not

formulaic but arise from an intuitive understanding of

the shared norms that govern ethical conversation. AI

does not have this power of interpretation to adapt

language in a socially meaningful way. It can only

simulate these through patterns that have been learned

during training, but it is not truly “aware” of the rules

of the game nor can it comprehend the deeper

significance of ethical discourse. For example, it may

provide an apparently appropriate response to a

sensitive topic, using an algorithm that lacks the

emotional or even cultural resonance needed to give

the conversation human depth. The reason for this is

simple: AI models are not attached to lived

experiences or cultural practices because they

constitute the very foundation of Wittgenstein’s

forms of life and, therefore, any meaning for

language. Consequently, AI perceives language as

bare information, stripped of those human

relationships that normally interpret its content. In the

opinion of Wittgenstein, meaning at all is essentially

coterminous with its use, a view that contemporary

AI is forced to take and can only ever really achieve

the appearance of use.

The existing AI models have been designed to

learn from the language used in static texts that float

in a vacuum. A big leap forward would be to have

Philosophical Foundations of Learning: Insights from Wittgenstein and Heidegger on AI Cognition

233

systems that can learn language use within socially

situated environments, where words are used in

actual, practical, and interactive contexts. Nelson

(2009) mentioned that the significance of communal

activities for language acquisition is highlighted by

contrasting Wittgenstein's ideas on word learning and

meaning in group activities with contemporary

cognitive and social-pragmatic theories. Through

this, the machine can attempt to replicate the kind of

practical understanding comprising diverse elements

which humans derive from actual lived interactions.

Just like humans learn by trial and error with social

validation, the AI can implement reinforcement

learning in the context of real environments. Through

interaction with a user and either explicit or implicit

feedback, the AI could develop that intuitive feeling

of when its response corresponds to the “rules of the

game.” For instance, it could come to perceive that

jokes are fine for informal contexts but not for the

gravitas needed in formal interactions or notice that at

times it should prioritize empathizing over just being

factually correct.

4 HEIDEGGER’S BEING-IN-THE-

WORLD: MEANING AS

EXISTENTIAL ENGAGEMENT

AND CHALLENGES FOR AI

Martin Heidegger’s concept of “Being-in-the-world”

presents a fundamental challenge to traditional views

of cognition. Heidegger repudiates the

representational model, which takes understanding to

be the production of internal constructions that

somehow correspond to an external reality.

Positively, he asserts that human cognition comes

into being as active engagement with the world —

such an approach underscores the focus on meaning

as located, contextual, and relational. According to

this view, therefore, understanding does not stand

detached from mere observation but gets assimilated

as a practical engagement conducted through

embodiment. Humans, as per Heidegger, do not "gain

access" to the world by means of abstract symbols;

they are always already in some way involved in it,

drawing their significance from those contexts in

which their action and existence find themselves.

Imagine a carpenter with tools. In daily practice,

a carpenter uses these tools without breaking down

their description into individual parts. Instead of

defining the hammer as a “wooden utility with a

handle and a steel striking head”, the carpenter gives

it a role—for instance, to drive in nails. Such

equipment acquires its definitive sense from its

practical integration into the totality of equipment

within which it operates. Heidegger has termed this

mode of being “readiness to hand,” in which meaning

is already built into our dealings with things,

stemming from their bearing on the tasks we set for

ourselves. This engaged relationship stands in stark

contrast with the theoretical attitude, which naturally

demands the isolation and abstraction of objects from

their settings and reduces them to individual

components. Unlike humans, AI systems perceive

their surroundings abstractly and separate from the

world that they act in; humans see where they are

going based on lived experience. Even in advanced

robotics research, where machines can move from

one place to another for the most part, those

movements are determined by pre-specified

algorithms rather than by the robot having a built-in

appreciation of its presence in such space. For

instance, a robot may traverse a cluttered room

successfully with reliance on sensor data and

pathfinding algorithms, yet always within that space

without ever truly comprehending its place and

purpose within it. For example, to move an obstacle

rather than navigate around it or why objects within

the room have some emotional or functional

importance to them. Lacking existential grounding,

AI is fundamentally unmoored from the very contexts

within which it operates, thereby reducing any

possible engagement with the world as input-output

functioning.

One of the fundamental challenges in attempting

to recapture the Heideggerian notion of Being-in-the-

world results from the basic nature of AI’s design. but

by addressing the Cartesian presumptions of modern

cognitive science, a Heideggerian approach can act as

the "conceptual glue" for future cognitive science

research (Wheeler, 2005). AI is designed upon

representational architectures that treat the world as

information to be processed, not as a meaningful

relational whole. These systems are based on

statistical relationships and hard-and-fast rules that

take inputs as they appear without the possibility of

accumulating the kind of contextual and relational

understanding that arises out of human cognition. Nor

does AI have that kind of intentionality, the

background upon which human dealings with the

world rest. A more Heideggerian approach might

enhance AI research and development, while

Heideggerian AI failed because it lacked a

comprehensive knowledge of the human mind

(Dreyfus, 2007). For humans, every action and every

perception is molded by some goal or anxiety, which

Heidegger has termed Dasein's "care," influencing

IESD 2025 - International Conference on Innovative Education and Social Development

234

how one interacts with things and interprets their

environment. Insight is a mysterious, unconscious

phenomenon that can be explained in causal terms, as

demonstrated by the emergence of the 'cognitive

unconscious' in the 19th and early 20th centuries

(Shanker, 1995). To approach the AI systems closer

to Heidegger’s ideal contextual understanding, the

researcher needs to make a machine that would take

into consideration the sensation and the environment

that is around it and the basic concept of human

activity and its prospects for further study are

illuminated by Heidegger's Dasein philosophy and its

applicability to cognitive science. (Kiverstein,

Wheeler, 2012). That means moving from

representational processing to one that is more

relational and adaptive engagement. There are

multiple avenues to be pursued: beyond just

immediate sensory inputs, the AI system must be able

to learn long-term patterns of interaction with the

environment and adapt its behavior over those scales.

AI systems should develop feedback loops to be able

to refine their understanding of the environment

through interaction with humans and other agents.

Preston (1993) raised that although it needs more

refinement and comprehension, Heidegger's

alternative method of analyzing intelligent behavior

provides insightful information for AI and cognitive

research. This draws Heideggerian insights in AI

design away from the classical frames of computation

models, turning to prioritize relational engagement,

contextual awareness, and intentionality. In this way,

digital systems could come a little closer to achieving

the existential and situational features of human

cognition. Perhaps, attaining true Being-in-the-world

may be far from reach, but such developments will

make the AI system practice more meaningful and

adaptive interactions with the environment, hence

drawing it nearer to Heidegger’s vision of

understanding as a dynamic process that involves an

active context.

5 CONTRASTING YET

COMPLEMENTARY

PERSPECTIVES:

WITTGENSTEIN AND

HEIDEGGER ON MEANING

AND CHALLENGES FOR AI

The conception of meaning of the two philosophers,

Wittgenstein and Heidegger, assimilate different yet

reciprocal aspects. For instance, Wittgenstein

primarily focuses on the external social practices that

form the basis for any meaning emerging out of

shared human activities. According to him, language

is not a formation but a dynamic process; it is a game

that depends on forms of life belonging to cultured

and social human beings. Wittgenstein's "Lecture on

Ethics" continues his exploration of ethics by delving

into Heidegger's ideas of Being and dread (Murray,

1974). Hence, words get their meaning not just from

any particular circumstance but from their use within

socially defined practices. For example, the concept

of making a promise only has meaning according to a

culture where there is already an understanding of

trust and responsibility. So, as Dastur (2010) said that,

Wittgenstein refuses metalanguage and questions his

own position, while Heidegger questions the ante-

predicative foundation of propositional language and

its connection to Dasein's transcendence. On the other

hand, Heidegger conveys that meaning arises in the

context of active embodied engagement with the

world. And so human understanding is based on the

actual physical, social, and existential contexts of life.

Meaning arises not only through social conventions

but through a direct, willful engagement with one's

surroundings. For instance, for a carpenter, the

concept of a hammer goes beyond what is described

in the language or how it is used within a society; it

emerges from the embodied experience of this tool as

an extension of his purpose and intentions for a

specific task. The orientation toward Being-in-the-

world pressed by Heidegger elicits how meaning is

fundamentally something deeply relational and

contextualized, emerging from human existence.

Social practices and existential engagements

underscore fundamental lacunae in the AI paradigm.

Most AI systems are essentially about symbols,

abstractions, and deriving patterns from data—

largely missing out on all of what learning and

meaning-making are socially and existentially about.

For example, it is capable of producing text in various

languages or holding a simple conversation, yet A.I.'s

comprehension consists strictly of functionality and

not human cognitive depth. Integrating insights

advanced by Wittgenstein with Heidegger

researchers will open up, through their work, the

manners in which AI can be brought toward a richer

and more cohesive understanding of meaning.

Wittgenstein and Heidegger's views on mind/action

can be combined to form a holistic view of life,

combining their views on the constitution of states

and the flow-structure of the stream of life (Schatzki,

1993). Since, according to Wittgenstein, meaning is

something concocted socially, therefore AI systems

can be developed to gather its learning experience

Philosophical Foundations of Learning: Insights from Wittgenstein and Heidegger on AI Cognition

235

from interaction throughout a wide range of human

settings. Existing AI models are rather grounded on

fixed datasets and hence display difficulty in

adaptation toward the socially contingent and

contextualized nature of practice. Such, including

dynamic and interactive frameworks, could place

artificial intelligence systems in a position to perceive

and absorb implicit rules of human interaction. An

example scenario could be where conversational AI

observes and fine-tunes responses according to live

users, thereby gradually learning norms of context

regarding politeness or humor. Being-in-the-world

would be applied as per Fecht’s principle which

indicates the embedding environment is instrumental

to meaningful understanding. AI systems which are

designed for the processing of symbolic data as well

as physical, sensory, and relational inputs may

acquire a better-rounded understanding of their

environment. One example can be a household robot

which notices the warmth coming from the sun. By

combining sensory information, such as the warmth

of sunlight or the sound of a certain person’s voice,

coming from the environment, the robot will be able

to relate to its environment in terms of Heideggerian

relationality. To develop AI, this is exactly what the

integration of embodied practices with social

practices entails.

6 ADVANCING MACHINE

COGNITION: TOWARD A NEW

PARADIGM

The future of artificial intelligence requires a

transition from representational models— where AI

identifies patterns based on data, to an engagement

model defined as context-aware, cognition-driven

through interaction. Today, representational models

are basic in the AI systems that are built around

extracting a statistical pattern from a large dataset and

generating output based on this abstraction. Duarte

(2019) said the techniques based on machine learning

can successfully mimic human behaviour in games,

producing results that are almost identical to those of

trained players in traditional card games like Hearts.

Great for doing things like advancing our capabilities

in natural language processing or image recognition,

it turns out to be extremely bad for getting a handle

on the world in any deep way because it lacks proper

engagement. An engagement model must mean that

AI systems are not only data processing but also

dynamic interaction with the environment, thereby

being capable of adaptation to the concrete context

and learning through embodiment experience.This

version of Hubert Dreyfus's seminal work

underscores the need for AI researchers to investigate

philosophy and intricate mental models, as well as the

incapacity of disembodied machines to replicate

higher mental activities (Dreyfus, 1992).

An engagement model gives priority to data as

contextual and relational knowledge rather than

abstract symbols. Thus, Future work should focus on

interactive games where communication in Natural

Language is crucial for understanding semantics and

physical embodiment is essential for developing

grounded meanings in neural models (Suglia,

Konstas, Lemon, 2024). And information has to be

embodied within real contexts. Therefore, for

instance, an interactional approach for conversational

AI would relate not only to the capacity to generate

grammatically correct language responses but a

further accommodation of tone, phrasing, and

information to the emotional status, cultural

environment, and social dynamics of the individual

being spoken to. Such dynamics in interaction barites

a striking resemblance to human communication,

where the meaning flows not just from words but

from the network between language use, intention,

and context. At the application level, this requires

adaptive feedback loops that depend on continuous

learning from experience. Perspectives of philosophy

give a grounding to such transformations, involving

critiques on existing limitations as well as offering

constructive ways to help build systems that could

comprehend more meaningful cognition. Both

Wittgenstein and Heidegger could be relevant to the

development of context-aware AI. Heidegger argues

that for a basic understanding of human experience,

one must consider a complex set of interrelations

among different entities over time. Wittgenstein’s

approach through language also offers an interesting

perspective: in his later work, he placed increasing

emphasis on the social aspects of language, and the

notion that rules should be seen as emergent

phenomena evolving in response to a need within

human practice. The training of AI should keep this

in mind and gradually place more importance on

datasets that reflect the dynamism and contextual

grounding for its learning. Machine learning

techniques, such as neural networks, evolutionary

computation, and reinforcement learning, can

enhance digital game AI by improving game agent

behavior and creating more engaging and entertaining

experiences (Galway, Charles, Black, 2008). For

example, an AI system trained using Wittgensteinian

principles in datasets representing diversity across

human social practices might acquire the ability to

IESD 2025 - International Conference on Innovative Education and Social Development

236

understand humor, together with politeness or

sarcasm, by recognizing and adapting its responses

according to contextual norms. The focus for such an

AI is on developing pragmatic adaptability as it learns

the evolving, dynamic “rules” of communication

through real-world experience rather than through

fixed algorithms. 'Being-in-the-World’: Heidegger’s

philosophy underscores the embodied practice and

referential context. Such an AI Heidegger would be

programmed to dynamically engage its environment,

drawing knowledge not only from datasets but also

from embedded sensory encounters. For instance, a

robotic system in a care home might learn how to

understand the implicit emotional and physical

signals given by the residents and adapt its responses

to become more personalized and caring. In such a

system, meaning would be that which results from its

relationship with the world, rather than as an isolated

computational result. These philosophical intuitions

are essentially what is meant to underscore the

impetus toward building future AI that can truly

perform within lived contexts, be those through social

interaction (Wittgenstein) or environmental

embedding (Heidegger). This is how AI can be

designed to engage with the world more in line with

human understanding.

As AI systems become increasingly integrated

into human contexts, ensuring that they align with

social norms, cultural differences, and existential

values becomes essential. The research of Heidegger

and Wittgenstein can help educators better

comprehend cognition in science classes, improve in-

class science instruction, and guide the creation of

innovative learning environments (Roth, 1997). The

move toward the model of engagement makes the

need for ethical safeguards even greater because AI

systems will come to bear direct influence upon and

interaction with human lives. Philosophy has laid

down important motivations for ensuring human-

centric ethical AI design through the principle of

making sense of the world. For example, in his claims

about language and meaning, Wittgenstein more than

once pressed that the rules of social practice be

respected in building an AI system. A robot trained to

know what is funny in one culture must appreciate

that these utterances may be used very

inappropriately or be misunderstood in another

setting. Heidegger posits that `being with others as a

condition of being is a significant aspect of being.

With consequences for ontological and

epistemological commitments, the paradigm shift in

AI towards enactive cognition centers on

embodiment, development, interaction, and emotion

(Vernon, Furlong, 2006). The AI will have to be

empathetic and be involved, especially for delicate

domains such as healthcare or education. Ethical

considerations, as autonomy achieves by the AI, must

also demand careful respect. As AI becomes more

capable of dynamic engagement, so too must design

ensure that the machine's autonomy is bounded by

humans watching it and keeping it accountable. An

AI system should therefore make explicit its decision-

making processes to allow for reasons for

understanding and challenging actions when

necessary.

7 CONCLUSION

The idea of incorporating AI into design reflects

limitations, but also untapped potential for

innovation. Wittgenstein's interpretation of

Heidegger challenges the representational and data-

driven nature of AI, the construction of AI and the

continuing development of its theory that can move

beyond mere efficiency to meaningful and ethical

engagement with the human environment. The results

are fine in terms of output, but these models miss the

deeper layers of meaning - contextual, social and

existential; These are all components of human

understanding. In practical social usage, the

philosophical principles of interaction, adaptability,

and situational awareness are prioritized and

established by humans in a shared framework. This

means embedding AI systems in a relational

environment where meaning is no longer thought of

as some abstract piece of data, but as something that

emerges and develops through concrete interactions

with the world. Philosophically, this has pushed the

boundaries of AI from mere functionality to the

ability to adapt meaningfully in real-world complex

situations.

The next generation of AI should also go beyond

these levels of interaction to capture social

interactions, environmental contexts, and even

philosophical reflections to better fit into these

contexts. This will enable AI to achieve not only

operational efficiency, but also the richness and

adaptability of human cognition. The insights raised

from philosophy will hopefully provide new ways to

build intelligent and meaningful AI that aligns with

human life.

REFERENCES

Dastur, F. (2010). Language and Metaphysics in Heidegger

and Wittgenstein. , 319-331.

Philosophical Foundations of Learning: Insights from Wittgenstein and Heidegger on AI Cognition

237

De Almeida Rocha, D., & Duarte, J. (2019). Simulating Hu-

man Behaviour in Games using Machine Learning.

2019 18th Brazilian Symposium on Computer Games

and Digital Entertainment (SBGames), 163-172.

Dreyfus, H. (2007). Why Heideggerian AI Failed and How

Fixing it Would Require Making it More Heideggerian.

Philosophical Psychology, 20, 247 - 268.

Dreyfus, H. (1992). What computers still can't do: a critique

of artificial reason. Leonardo, 27, 83.

Galway, L., Charles, D., & Black, M. (2008). Machine

learning in digital games: a survey. Artificial Intelli-

gence Review, 29, 123-161.

Gómez, J., & Bravo, R. (2009). The Unbearable Heaviness

of Being in Phenomenologist AI.

Kiverstein, J., & Wheeler, M. (2012). Heidegger and Cog-

nitive Science.

Liu, L. (2021). Wittgenstein in the Machine. Critical In-

quiry, 47, 425 - 455.

Murray, M. (1974). A Note on Wittgenstein and

Heidegger. The Philosophical Review, 83, 501.

Nelson, K. (2009). Wittgenstein and contemporary theories

of word learning. New Ideas in Psychology, 27, 275-

287.

Preston, B. (1993). Heidegger and Artificial Intelligence.

Philosophy and Phenomenological Research, 53, 43-

69.

Proudfoot, D. (2009). Meaning and mind: Wittgenstein's

relevance for the ‘Does Language Shape Thought?’ de-

bate. New Ideas in Psychology, 27, 163-183.

Roth, W. (1997). Being-in-the-World and the Horizons of

Learning: Heidegger, Wittgenstein, and Cognition. In-

terchange, 28, 145-157.

Schatzki, T. (1993). Wittgenstein + Heidegger on the stream

of life. Inquiry: Critical Thinking Across the Disci-

plines, 36, 307-328.

Shanker, S. (1995). The nature of insight. Minds and Ma-

chines, 5, 561-581.

Suglia, A., Konstas, I., & Lemon, O. (2024). Visually

Grounded Language Learning: A Review of Language

Games, Datasets, Tasks, and Models. J. Artif. Intell.

Res., 79, 173-239.

Vernon, D., & Furlong, D. (2006). Philosophical Founda-

tions of AI. , 53-62.

Wheeler, M. (2005). Reconstructing the Cognitive World:

The Next Step.

IESD 2025 - International Conference on Innovative Education and Social Development

238