Why we have a self, model other minds, and feel free

Paper 7 · Pødenphant Lund (2026) · Read on Zenodo

I use friction theory to understand what a mind is.You are standing at a choice. Before you do anything, you run it through in your head: if I say this, then that happens, and then I will have to... The little future you just fast-forwarded, and let shape what you actually do, is the same piece of machinery that gives you a self, and that lets you guess what is going on inside another person. Why we have a self, how we model other minds, and what it means to have free will are normally treated as three quite separate questions, in three research fields that barely talk to each other. They are three faces of the same thing: a substrate that runs hypothetical futures in its head and uses them to weight what it does next.

The three threads

Cognitive science, developmental psychology, and philosophy of mind have each been working on a problem they treat as their own:

• Self-modelling. Why do we have a representation of ourselves at all? Why does the brain bother to model the very system doing the modelling?
• Theory of mind. Why and how do we model other minds: figuring out what someone else believes, wants, intends?
• Free will. What does it mean to be free in a universe that runs on physics? Are our choices "really" choices?

These three traditions almost never talk to each other. Self-modelling is studied in cognitive neuroscience and computational cognitive science. Theory of mind is studied in developmental psychology, primatology, and (recently) LLM evaluation. Free will is studied in philosophy of mind, philosophy of action, and the experimental free-will tradition. None of them treat the other two as central.

The proposal

The paper argues that all three are manifestations of one underlying machinery, forward-modelling: the substrate's capacity to simulate hypothetical states and use those simulations to weight what it does in the present.

Self-modelling: the data structure forward-modelling needs

If a substrate is trying to optimise outcomes over time, it has to model how its current decisions will affect future states, including future states of itself. That requires representing itself as a thing-in-the-world that can be acted on. The self-model is not a luxury or a side-effect; it is the data structure the optimisation requires. The deeper the substrate's time horizon, the deeper the self-model has to be.

Theory of mind: forward-modelling pointed at another forward-modeller

Modelling another agent's internal states is the same machinery applied to a different target. You forward-model yourself to predict your future. You forward-model another person to predict what they will do. The two use the same capacity parameters and the same predicted scaling. This explains why theory-of-mind capacity correlates with self-modelling depth, and why large language models that have learned to forward-model themselves often (sometimes) display theory-of-mind-like behaviour.

Free will: forward-modelling that translates future friction into present pressure

This is the harder claim. The argument: forward-modelling is what turns future friction (the cost a decision will impose later) into a present friction-gradient (a pressure that shapes the current decision). You feel free because you can simulate hypothetical futures and let them influence the current race. You are not free from the substrate; you are free within the substrate that simulates and weighs alternatives.

The framework's central thesis: we are free within our context; we are not free from our context.

The dissolutionist move

The paper does something unusual for philosophy of mind. It does not pick a side in the free-will debate between libertarianism and determinism. Instead, it argues that the debate is built on a category of constructions that share a structural flaw.

The construction-family: the philosophical zombie, the perfectly rational Econ, Newton's vacuum, the ideal Bayesian observer, the frictionless market, libertarian free will. Each of these requires a substrate with no friction: a substrate that can instantiate the construct without losing anything in the instantiation. The paper argues that this kind of substrate is not just empirically rare; it is structurally impossible given how race-architecture works. The debate dissolves because the construction being defended (or attacked) was never possible to instantiate in the first place.

Engagement with adjacent traditions

• Integrated Information Theory (Tononi) on consciousness — treated as competitor positioning, not adjudicated.
• Active inference (Friston) and control theory (Carver & Scheier; Powers) on adjacent territory — three concrete distinguishing predictions are proposed: path-dependence, commitment-irreversibility signatures, resource-bounded layer-asymmetric pre-emption.
• Contemporary free-will literature — Mele, Kane, List, Pereboom on the libertarian-naturalist side; Libet, Wegner, Haggard on the experimental side. The dissolutionist move responds to the structural shape of both literatures.

Why this matters

For cognitive science. If self-modelling is structurally required by forward-modelling, then the question "why do we have a self?" gets a mechanistic answer that does not require special pleading for consciousness. The self is what the substrate has to model in order to optimise over time.

For LLM evaluation. Theory-of-mind benchmarks on LLMs make sense in this framework: a substrate that has learned to forward-model itself acquires (some of) the machinery for forward-modelling others. The paper engages the contemporary LLM-ToM literature directly.

For philosophy of mind. The free-will debate has been stuck for decades. The dissolutionist move offers a way out that does not require choosing libertarianism, hard determinism, or compatibilism. The construction the debate is defending was never going to be instantiable in the kind of substrate we live in.

The cite

Read on Zenodo → · Technical version · Dansk version

Related on this site:

• Paper 0 (BFT) — the biological specialisation that P7's forward-modelling mechanism plugs into.
• Paper 1 (Friction Theory) — the substrate-universal framework whose race-axioms P7 builds on.
• Paper 14 (Logic as Reactance) — truth-value judgment as reactance signature; P7's dissolutionist response uses similar machinery.
• Paper 13 (Operational FT) — race-opening, recursive resolution; P7's recursive forward-modelling builds on this.