Toward a Symbiotic Society with Generative AI: For Living Together into the Future

Since the release of ChatGPT in late 2022, large language models (LLMs) have become the central focus of AI research. Among the various paths toward Artificial General Intelligence (AGI), beginning with LLMs now appears to be the most viable. However, as discussed in chapter “Large Language Models and Distributional Semantics: Do Large Language Models Understand Language?”, these models carry significant challenges, such as bias and hallucination—the generation of information that is factually incorrect, unfounded, or syntactically incoherent.

Efforts to align such AI models with human ethics, values, and goals constitute the field of AI alignment. As generative AI becomes increasingly embedded in society, this field is gaining critical importance. But here arises a deeper question: Are human ethics and values truly so fixed and universal that they can serve as reliable training targets or “correct labels” for AI? In reality, human values are plural and diverse. As LLMs are applied globally, reliance on a few dominant, linguistically biased platforms risks reinforcing a new form of globalization—the homogenization of values. This in turn could suppress cultural, ethical, and ideological diversity.

From the viewpoint of symbol emergence systems theory, however, language—as well as the ethics and laws expressed using language—is not static but emergent, context-dependent, and historically contingent. Language is inherently arbitrary, always shifting. Ethics and norms, too, evolve across time, culture, and environment.

Within the context of AI alignment, two fundamental concerns emerge:

1. 1.

   How LLMs may reshape human language and thinking.

2. 2.

   How human ethics and value systems themselves evolve and adapt.

In March 2023, GPT-4 was released. While using it for English proofreading, several months later, I (the author) noticed it frequently suggested the phrase “delve into,” a relatively uncommon expression I hadn’t often used before. Curious, I checked Google Trends and found a clear uptick in global usage of “delve into” starting that very month.^{Footnote 1} While seemingly minor, this example subtly illustrates how LLMs are already influencing our linguistic habits. Language is shaped by usage; we adopt expressions we are exposed to. Thus, our language—and by extension, our thinking—is already changing through exposure to generative AI.

This insight leads to a crucial realization: AI alignment must account for humans themselves as evolving participants in the system. We need a logic of mutual, dynamic alignment—not just aligning AI to us, but co-adapting together in a symbiotic framework.

A second issue lies in human ethics itself. As generative AI proliferates, so too do risks—not only from AI hallucinations, but from human misuse, such as fabricating false texts or images. Even if we constrain AI’s hallucinations, we cannot prevent malicious use when intent originates from humans.

In the age of AI-integrated society, these problems are fundamentally interconnected. AI alignment, though rooted in technical disciplines, must serve as an interface to philosophy, ethics, sociology, and political science. Across these boundaries, we are called to envision and build a new symbiotic society, where humans and AI co-evolve ethically, socially, and cognitively.

In the 2020s, the rapid progress of generative AI has brought humanity face-to-face with a new identity crisis. Up until the twentieth century, what made human intelligence unique was often said to be rational and linguistic thought. The use of recursive language structures was considered a marker of human distinctiveness. Creativity—manifested in painting, composing music, or storytelling—was also often held up as something only humans could do. But many of these long-standing pillars are now being dismantled by generative AI and large language models.

Yet, there is a hidden bias in how we define “human-likeness.” Discussions of intelligence frequently focus on higher-order cognitive abilities, overlooking the lower-order skills that all humans, including toddlers, possess. Tasks such as proving theorems, playing high-level Go, painting, composing, or speaking multiple languages—things most people cannot do—have somehow become the benchmark for what it means to be human.

This bias is reflected in Moravec’s Paradox, which highlights the counterintuitive reality that what is easy for humans—like walking, grasping, or sensing—is often hardest for AI, while what’s cognitively demanding for humans is relatively easy for machines. Even in this era of generative AI, we still don’t have domestic robots that can reliably help with everyday chores—an example of AI’s enduring limitations in sensorimotor capabilities.

Historically, social perceptions of “intelligence” were shaped through contrast with animals. Because animals can perform sensorimotor tasks, such capabilities have not been included in what defines “human-likeness.” Instead, humanity has placed its identity in language, logic, and creativity—areas now increasingly mastered by AI. In fact, from AI’s earliest days, logic and games were among its primary targets; IBM’s Deep Blue defeated the world chess champion as early as 1997. While people once claimed AI could never match human creativity, generative AI now produces images and music that surpass those created by most humans.

We must now update our concept of “human-likeness.” First, we must acknowledge that animals challenge us from below and AI from above. “Human-likeness” is being caught in a cognitive pincer movement. To articulate what makes us human, we must look between and across high- and low-order cognition—and beyond them. Perhaps it is precisely in these connecting layers, such as System 0 and System 3 (as described in the previous section), that human distinctiveness truly lies. This is a view made possible by symbol emergence systems theory.

Second, we must confront the functionalist assumptions embedded in AI research itself. Functionalism—the view that mind and intelligence are defined by what something can do—makes AI development possible. However, as AI becomes deeply embedded in society, it promotes a worldview where human value is similarly reduced to functionality. Yet our worth as humans is not defined by what we can do alone.

In market economies, people are often valued according to their capacity to produce or perform. When this tendency aligns with the functionalist foundations of AI, “human-likeness” becomes vulnerable. The 2020s may mark a period when our shared sense of humanity is increasingly threatened. It is now imperative that we consciously move beyond functionalism, to recognize and cultivate human dignity not only in our abilities, but in our being—our shared presence, meaning, and relationships.

Symbol emergence systems theory presents a bottom-up perspective on language: that symbol systems are emergent, formed socially through collective predictive coding. However, the essence of the theory lies not only in bottom-up formation, but also in how these emergent symbol systems exert top-down influence on individuals. In the language of complexity science, this is known as the micro–macro loop (see chapter “Symbol Emergence Systems: Toward a World Where Humans and AI Co-discover Meaning”). This final section explores the philosophical implications of this loop.

Across the humanities and social sciences, debates surrounding the nature of knowledge, self-awareness, and social phenomena have often been shaped by a dichotomy: realism versus social constructionism. While the former considers phenomena and categories as objectively existing entities, the latter emphasizes that knowledge and meaning are socially constructed through cultural context, historical contingency, and symbol systems.

For example, in gender theory, to view the difference between men and women as biological is essentialist, while to see it as socially constructed is the constructionist perspective. In philosophy of science, realism holds that scientific discoveries uncover truths that exist independently of social influence, while social constructivism argues that scientific knowledge is shaped by the beliefs, values, and interactions of scientists themselves.

Symbol emergence systems theory does not take sides but seeks to encompass both views. Consider scientific knowledge: it is formed as a process in which distributed, tacit, embodied experiences are externalized into formal representations. This aligns with collective predictive coding, where symbol systems evolve as emergent representations grounded in both individual cognition and cultural context. The beliefs and values of scientists can be seen as a distributed memory within System 3. In this light, realism and constructivism are not incompatible, but interwoven.^{Footnote 2}

In daily scientific practice, the realist and constructivist perspectives continually collide, producing the tension and dynamism that define System 3. This tension is not a flaw, but a feature of human society: a system that persists in adapting through collective symbol emergence.

In gender debates, for instance, the categorization of “male” and “female” is often treated naively as either completely arbitrary or entirely biological. But symbol emergence systems theory sees these categories as emergent from relationships among symbols, cultural context, and collective experience. Arbitrary does not mean meaningless; it means flexible within constraints. The key is to recognize the symbol system as a historically learned structure that carries functional weight in adaptation.

We as humans can generate new symbols at the micro level. But from a macro-level perspective, we are also bound by existing symbol systems. Psychoanalyst Jacques Lacan described this as the “Big Other”—the symbolic order into which we are born, which structures our understanding of self and other. Symbol emergence systems interpret this as a natural consequence of collective predictive coding: the constraints of symbol systems enable us to indirectly share others’ experiences and function as a society.

To ignore symbolic systems entirely is to risk cultural collapse; to passively obey them without expression or critique is to halt symbolic evolution. We must both participate in the systems and continuously reshape them.

So what does it mean to live “authentically”? Should we break free from social expectations of the self, or accept them rationally as part of societal norms?

The answer, I believe, is both.

If the essence of being human lies not in solitary adaptation, but in the collective emergence of arbitrary symbol systems for shared adaptation, then we are necessarily participants in symbol emergence systems. In that participation, the micro–macro loop is inevitable. We are neither fully free from nor entirely bound by symbolic structures. The struggle within that loop is what makes us human.

As Generative AI becomes an inescapable part of this system, our task is to build technologies, theories, institutions, cultures, and societies that allow us to live as ourselves—to be human—within the new loops of symbol emergence.

Authors

1. Tadahiro Taniguchi

Cite this chapter

Taniguchi, T. (2026). Toward a Symbiotic Society with Generative AI: For Living Together into the Future. In: Taniguchi, T. (eds) Symbol Emergence Systems. Springer, Singapore. https://doi.org/10.1007/978-981-95-1327-7_35

Download citation

• .RIS
• .ENW
• .BIB

• DOI: https://doi.org/10.1007/978-981-95-1327-7_35

• Published: 17 March 2026

• Publisher Name: Springer, Singapore

• Print ISBN: 978-981-95-1326-0

• Online ISBN: 978-981-95-1327-7

• eBook Packages: Artificial Intelligence (R0)