Datasets:
Modalities:
Text
Formats:
text
Size:
10K - 100K
Tags:
philosophy
epistemology
artificial-intelligence
systems-thinking
value-alignment
cognitive-sciences
License:
value-hexagram-corpus / 03_fondations_systemiques /FromMentalModelstoCollectiveCo-construction_MappingTensionsandtheRoleofArtificialIntelligence.txt
| # From Mental Models to Collective Co-construction: Mapping Tensions and the Role of Artificial Intelligence | |
| **Author:** Manus AI | |
| **Date:** March 2026 | |
| --- | |
| ## Introduction | |
| The ways in which human beings construct their relationship to reality have been modelled through a wide range of concepts, from individual cognitive structures to the broadest social dynamics. The emergence of Artificial Intelligence (AI) — whether in the form of generative language models (LLMs) or predictive architectures (World Models) — is transforming this landscape profoundly. AI no longer acts merely as a tool, but as an "actant" [1] capable of simulating, mapping, and sometimes participating in the collective fabrication of meaning. | |
| This synthesis offers a structured cartography of these mental constructions, organising them around two fundamental tensions that run through every attempt at collective co-construction: the tension between **the operational and the ethical**, and the tension between **the individual and the collective**. It then examines how deliberative and semiotic approaches to AI propose to address these paradoxes. | |
| --- | |
| ## 1. The Strata of Mental Construction | |
| The concepts through which we think our relationship to the world can be organised into three broad strata, moving from the individual to society, before being confronted with their computational equivalent. | |
| ### 1.1. The Individual Cognitive Foundation | |
| At the scale of the individual, the relationship to reality is never direct; it is always mediated by internal structures. Jean Piaget was the first to formalise the idea of **schemas** [2] — cognitive building blocks that transform through *assimilation* (integrating reality into the schema) and *accommodation* (modifying the schema in response to reality). This dynamic of equilibration finds an extension in the **mental models** of Craik and Johnson-Laird [3], which describe temporary internal representations enabling mental simulation ("what would happen if..."). | |
| More recently, computational neuroscience has unified these approaches under the paradigm of the **predictive brain** (Friston, Clark) [4]. The brain is no longer seen as a passive receiver, but as a generative machine that projects its predictions onto the world and perceives only "prediction errors" — the gap between the model and reality. | |
| ### 1.2. The Social and Collective Dimension | |
| These individual structures do not float in a vacuum: they are anchored in social environments. Pierre Bourdieu demonstrated how social structures become incorporated in individuals in the form of **habitus** [5] — a system of durable, pre-reflective dispositions. At the group level, Serge Moscovici theorised **social representations** [6], forms of practical knowledge that forge a shared common sense. | |
| Jürgen Habermas's theory of **communicative action** [7] lays the foundations for rational deliberation capable of transcending disagreements through a non-strategic use of language. However, as Ludwig Wittgenstein highlighted with his concept of **language games** [8], the meaning of words always depends on the "form of life" shared by speakers, making communication complex between groups whose experiences are too far apart. | |
| ### 1.3. The Worldview (Weltanschauung) | |
| At the apex of this architecture stands the **Weltanschauung** (worldview) [9], a philosophical concept designating the fundamental cognitive, moral and evaluative orientation of an individual or a society. Unlike mental models, which are partial and operational, the worldview is encompassing and identity-generating. In the scientific domain, Thomas Kuhn translated this concept through the notion of **paradigm** [10], showing that adherents of competing paradigms "live in different worlds", rendering their visions incommensurable. | |
| ### 1.4. The Computational Equivalent: World Models | |
| In the field of AI, Yann LeCun has proposed the architecture of **World Models** (notably through JEPA) [11]. To achieve autonomy, an AI must possess an internal model of the world enabling it to simulate the consequences of its actions within a latent space. While this approach is the direct descendant of Craik's "mental models", it differs radically from the "Weltanschauung": the World Model is purely operational. It simulates the physics of the world in order to act, but carries no moral or evaluative judgement. | |
| --- | |
| ## 2. The Two Fundamental Tensions of Co-construction | |
| Every attempt to build a shared project — whether a company, a society, or a political endeavour — encounters two major structural tensions that the preceding concepts help to illuminate. | |
| | Tension | Description | Manifestation in the Concepts | | |
| |---|---|---| | |
| | **T1: Operational ↔ Ethical** | The opposition between instrumental efficiency (KPIs, technique) and the search for meaning, values, and the common good. | Strategic action vs. communicative action (Habermas); the World Model (LeCun) vs. the worldview (Dilthey). | | |
| | **T2: Individual ↔ Collective** | The paradox between preserving the singularity of each individual and the necessity of a shared coherence for collective action. | Habitus and social representations tend to suppress singularity in order to forge group unity. | | |
| ### Mapping the Tensions | |
| The visualisation below positions the various concepts along these two axes of tension. | |
|  | |
| > **Analytical reading:** A clear diagonal emerges. Concepts linked to cognitive foundations (in blue) and to pure AI (in orange) are situated in the lower-left quadrant (operational and individual). Conversely, concepts linked to dialogue (in violet) and deliberative AI approaches (in pink) occupy the upper-right quadrant (ethical and collective). It is in this latter quadrant that the keys to resolving both tensions are to be found. | |
| --- | |
| ## 3. Resolutions: From Follett to Deliberative AI | |
| How can these tensions be transcended? Several theoretical and methodological frameworks, today augmented by AI, propose pathways toward resolution. | |
| ### 3.1. Creative Integration (Mary Parker Follett) | |
| As early as the 1920s, Mary Parker Follett [12] proposed a masterful resolution to both tensions. Faced with difference, she rejected domination and compromise in favour of **creative integration**. Integration consists of inventing a new solution that satisfies the deep needs of both parties. | |
| - **Resolution of T1:** She refused to separate the operational from the ethical, holding that genuine efficiency arises from taking real human needs into account. | |
| - **Resolution of T2:** For Follett, the individual does not sacrifice themselves to the group; they are fulfilled *in* and *through* the group. Difference is not a problem to be smoothed over, but the raw material of collective creation. | |
| ### 3.2. The Just Middle (Wassati and Enga.IA) | |
| As a contemporary extension of Follett, the Wassati approach [13] formalises the search for the "just middle" through four circles of values (individual, familial, societal, planetary). The Enga.IA tool maps emotions and deep causes to construct a **"just individual middle within a just collective middle"**. | |
| As the preceding chart shows, this approach is positioned at the very apex of the "Ethical / Collective" quadrant, because it directly resolves Tension 2 by making individual alignment the *sine qua non* condition of collective coherence, and resolves Tension 1 by systematically linking operational objectives to planetary and societal values. | |
| > **A note on the deep causes and the societal circle:** It may be worth proposing — humbly, as a hypothesis — that the *societal circle* of Enga.IA could be read as a practical and operational declension of the *Weltanschauung* (Dilthey) or *social representations* (Moscovici): a shared evaluative framework that gives meaning to collective action. Similarly, the *deep causes* identified by Enga.IA likely touch on what Bourdieu called *habitus* (when the cause is social and pre-reflective) or on *cognitive schemas* (Piaget) when the cause is cognitive — namely, the resistance of an existing schema to accommodation. In this reading, Enga.IA would use AI not to impose a vision, but to descend from surface emotions toward these pre-reflective structures, in order to make them visible and workable. | |
| ### 3.3. Augmented Collective Intelligence (HILL) | |
| The concept of the Hybrid Intelligence Living Lab (HILL), developed by La Fabrique du Futur [14], operationalises co-construction by creating a "digital twin of collective intelligence". Drawing on maieutics and creative controversy, HILL uses AI not to average opinions, but to facilitate a descent toward the group's deep needs (resonating with Otto Scharmer's *Theory U* [15]), enabling the emergence of prospective scenarios that integrate systemic complexity. | |
| ### 3.4. The Hybrid Semiotic Factory (Christophe Denis) | |
| To understand the precise role of AI in these devices, the epistemological framework of Christophe Denis [16] is fundamental. He theorises the **hybrid semiotic factory**, where meaning is no longer the monopoly of the human, but emerges from a "distributed unveiling" between humans, artefacts (LLMs) and symbolic environments. | |
|  | |
| > **Analytical reading:** This second chart illustrates the tension between the human and the computational. The Semiotic Factory, Deliberative AI, and the Wassati approach are explicitly positioned as "bridges" (in pink and amber) connecting the computational power of LLMs to the ethical and collective human dimension. | |
| Denis emphasises that LLMs produce a "machinic language" and a "machinic truth": they are formally plausible but devoid of a "form of life" (Wittgenstein). Consequently, AI hallucinations should not be seen as mere errors to be censored, but as "actants" (Latour) [1] capable of provoking cognitive decentring and opening new spaces of interpretation for the group. | |
| --- | |
| ## 4. Global Architecture: From Foundations to Syntheses | |
| All of these concepts articulate according to a precise hierarchy, moving from basic cognitive building blocks to contemporary resolution architectures. | |
|  | |
| > **Analytical reading:** This dependency graph shows how contemporary resolution frameworks (Level 5) draw on a dual heritage: on one hand, the human heritage of theories of dialogue and worldview, and on the other, the recent heritage of computational models. Deliberative AI and augmented collective intelligence devices (HILL, Wassati) constitute the convergence point of these two trajectories. | |
| --- | |
| ## Conclusion | |
| Machines are rapidly acquiring performant "world models" for simulating and acting, but they remain structurally devoid of "worldviews" (Weltanschauung), as they possess neither bodily anchorage (Enaction [17]) nor moral judgement. | |
| However, as recent experiments in deliberative AI demonstrate (DeepMind's Habermas Machine [18], Polis [19]), AI can play a critical role in collective co-construction — provided it is not used to average or smooth out differences, but rather as a maieutic third party within a hybrid semiotic factory. By mapping deep values (Wassati) and facilitating creative integration (Follett / HILL), AI becomes the instrument enabling human societies to resolve the dual tension of efficiency and meaning, of the individual and the collective. | |
| --- | |
| ## References | |
| [1] Latour, B. (1991). *We Have Never Been Modern*. Harvard University Press. | |
| [2] Piaget, J. (1936). *The Origins of Intelligence in Children*. International Universities Press. | |
| [3] Johnson-Laird, P. N. (1983). *Mental Models*. Cambridge University Press. | |
| [4] Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. *Behavioral and Brain Sciences*, 36(3), 181–204. | |
| [5] Bourdieu, P. (1980). *The Logic of Practice*. Stanford University Press. | |
| [6] Moscovici, S. (1961). *La Psychanalyse, son image et son public*. PUF. | |
| [7] Habermas, J. (1981). *The Theory of Communicative Action*. Beacon Press. | |
| [8] Wittgenstein, L. (1953). *Philosophical Investigations*. Blackwell. | |
| [9] Dilthey, W. (1911). *The Formation of the Historical World in the Human Sciences*. Princeton University Press. | |
| [10] Kuhn, T. S. (1962). *The Structure of Scientific Revolutions*. University of Chicago Press. | |
| [11] LeCun, Y. (2022). A Path Towards Autonomous Machine Intelligence. *OpenReview*. | |
| [12] Follett, M. P. (1924). *Creative Experience*. Longmans, Green and Co. | |
| [13] Peyret, H. (2024). *The Wassati Principle and Enga.IA Analysis*. Wassati. | |
| [14] Demarquette, S. (2024). *Hybrid Intelligence Living Lab (HILL)*. La Fabrique du Futur. | |
| [15] Scharmer, O. (2007). *Theory U: Leading from the Future as It Emerges*. Berrett-Koehler. | |
| [16] Denis, C. (2025). From the conversational agent to the hybrid semiotic factory: towards a philosophy of semiotic immanence. *HAL Sorbonne Université*. | |
| [17] Varela, F. J., Thompson, E., & Rosch, E. (1991). *The Embodied Mind*. MIT Press. | |
| [18] Google DeepMind. (2024). AI mediation can improve democratic deliberation. *Science*. | |
| [19] Megill, C., & Tang, A. (2023). Polis and vTaiwan: Scaling deliberation. *AI Objectives Institute*. | |