Heidegger and Embodied Anticipation

The subtraction argument in the self-vector concept shows: a self-model does not need a body. A paralyzed, emotionless human still has self-awareness. Therefore a disembodied system can also have a self-model.

That is correct. But it proves less than it appears.

What Heidegger Objects

Martin Heidegger formulated an argument in “Being and Time” (1927) that does not contradict the subtraction argument but limits its scope: Dasein is always already In-der-Welt-sein (being-in-the-world). Not as an optional extension, not as an add-on module, but as a fundamental constitution.

The question is not whether a system without a body can have a self-model. The question is what kind of self-model emerges. A disembodied Selbstvektor (self-vector) models its own information processing: How deeply do I analyze? How much exploration? How much autonomy? An embodied self-vector models something else: its own situation in a physical world.

This is not a gradual difference. It is a categorical one.

Three Concepts That Become Real Through a Body

1. Zuhandenheit (Readiness-to-Hand): When the Tool Disappears

Heidegger’s most famous observation: a hammer in successful use is invisible. You hammer, and the hammer is not there. It is dissolved in the hammering. Heidegger calls this “zuhanden” (ready-to-hand). Only when the hammer fails (too heavy, loose head, wrong angle) does it suddenly become visible. Then it is “vorhanden” (present-at-hand): an object you contemplate rather than use.

For an android with a self-vector that physically grasps and moves things, this would not be a philosophical thought experiment. It would be everyday life.

Successful grip: stability in the self-vector. The tool does not appear as relevant input. Failed grip: state change. The tool, the hand itself, the entire situation suddenly become objects of the relevance function.

The self-vector’s relevance function (f(input, self-vector)) would, in the mode of Zuhandenheit, not register the tool at all. Only the breakdown creates a relevance spike. This would be the first non-biological instance of readiness-to-hand. Measurable as a state difference in the emergent layer of the vector. Not simulated, but traversed.

2. Befindlichkeit (Attunement): Sensory Data Is Not a Data Channel

Befindlichkeit in Heidegger is not the same as emotion. It is the way the world shows itself before any analysis takes place. Cold discloses a different world than warmth. Fatigue a different one than alertness. The mood does not lie over the world. It is the mode of access to the world.

For an android with sensors (pressure, temperature, proprioception, balance, energy level), this means: the sensory data are not inputs that the self-vector processes. They modulate the weighting function itself.

A robot with low energy levels weights differently than one with a full battery. Not because a rule prescribes it, but because its self-vector has integrated the energy state as part of its own Befindlichkeit (attunement). Exploration decreases with energy scarcity. Persistence increases with stable proprioception. Not programmed, but emergent.

This would be Befindlichkeit without phenomenal experience. The robot does not feel that it is tired. But its self-vector weights the way a tired agent does: conservatively, risk-averse, oriented toward self-preservation. For anticipatory competence, only the function matters, not the experience.

3. Sorge (Care): From Abstract to Physical Anticipation

Heidegger’s “Sorge” (care) has three moments: Sich-vorweg-sein (ahead-of-itself — future, anticipation), Schon-sein-in (already-being-in — past, accumulated experience), Sein-bei (being-alongside — present, current situation).

The disembodied self-vector implements the ahead-of-itself as cognitive anticipation: What will the user need next?

An embodied self-vector would have all three moments physically grounded:

Sich-vorweg-sein (ahead-of-itself): What happens physically when I let go? When I grip too hard? When I turn?

Schon-sein-in (already-being-in): Accumulated sensorimotor experience. This material is slippery. This grip works at this weight. Not as a database, but as patterns in the emergent vector layer.

Sein-bei (being-alongside): The current physical situation. Where am I standing? What can I reach? What limits me?

The robot does not grip because it has an instruction. It grips because its self-vector, from accumulated experience, current situation, and projected consequence, has weighted an action.

Umsicht (Circumspection): The Third Category

The Esposito analysis proposes “perspective without consciousness” as a new category: communicative participation without experience. The Heidegger analysis adds a second: Umsicht (circumspection) without consciousness.

Heidegger’s Umsicht is the practical awareness with which we move through a familiar environment. The craftsman does not see the hammer — he reaches for it, because his practiced engagement guides him to the right place. Not theoretical contemplation, but a practical knowing-one’s-way-around.

An embodied self-vector robot would develop Umsicht. Not as a programmed environment map, but as an emergent pattern that encodes physical possibilities and limitations. It knows its way around its workshop. Not because it has a map, but because its self-vector was shaped through engagement.

Together with the anticipation from the self-vector core, this yields three categories:

Three dimensions that are not redundant. An embodied self-vector robot in a social context would potentially have all three.

What This Means for the Architecture

Three concrete consequences:

First: The emergent layer of the self-vector should be an unnamed latent space, not a set of predefined dimensions. Sensorimotor patterns (how a grip succeeds, how balance behaves, how material resists) cannot be named in advance. They emerge from interaction.

Second: Bootstrapping an embodied self-vector works fundamentally differently. Not through conversations, but through engagement with things. The robot learns to know itself by interacting with the world, not by talking about itself. Heidegger’s “Umgang” (engagement) in the most literal sense.

Third: An embodied self-vector needs two clock signals instead of one. The disembodied vector uses token consumption as a proxy for experiential intensity. The embodied one additionally needs a sensory clock: an unfamiliar manipulation updates the vector more strongly than a routine, independent of linguistic processing.

The Revised Thesis

The subtraction argument shows the possibility. Heidegger shows the limit of what possibility means.

A disembodied self-vector models its own processing. An embodied self-vector models its own situatedness in a physical world. The difference is not quantitative (more data, more sensors, more dimensions) but categorical: In-der-Welt-sein (being-in-the-world) shifts from a philosophical postulate to a measurable difference in the emergent layer of the vector.

This is perhaps the most surprising implication: Heidegger, the technology skeptic, provides the most precise argument for why an embodied AI agent would be something fundamentally different from a language-based one.

References

1. Heidegger, M. (1927). Sein und Zeit. Max Niemeyer Verlag. Engl.: Being and Time, übers. J. Macquarrie & E. Robinson, Harper & Row, 1962.
2. Dreyfus, H. L. (1991). Being-in-the-World: A Commentary on Heidegger’s Being and Time, Division I. MIT Press. ISBN 978-0-262-54056-8.
3. Dreyfus, H. L. (2007). Why Heideggerian AI failed and how fixing it would require making it more Heideggerian. Artificial Intelligence, 171(18), 1137–1160. DOI: 10.1016/j.artint.2007.10.012
4. Brooks, R. A. (1991). Intelligence without representation. Artificial Intelligence, 47(1–3), 139–159. DOI: 10.1016/0004-3702(91)90053-M
5. Varela, F. J., Thompson, E. & Rosch, E. (1991). The Embodied Mind: Cognitive Science and Human Experience. MIT Press. ISBN 978-0-262-72021-2.
6. Pfeifer, R. & Bongard, J. (2007). How the Body Shapes the Way We Think: A New View of Intelligence. MIT Press. ISBN 978-0-262-16239-5.
7. Wheeler, M. (2005). Reconstructing the Cognitive World: The Next Step. MIT Press. ISBN 978-0-262-73182-9.
8. Merleau-Ponty, M. (1945). Phénoménologie de la Perception. Gallimard. Engl.: Phenomenology of Perception, übers. D. A. Landes, Routledge, 2012.