Review | The meaning and origin of goal-directedness: a dynamical systems perspective

Introduction

What counts as an adequate scientific description of life has long hinged on how we balance reductionism and holism without drifting into ontological dualisms. Vitalism, with its appeal to an extra-physical élan vital, offered a historical placeholder for life’s apparent purposiveness; but precisely because it postulated an unknowable force, it faded as physics and chemistry learned to explain order far from equilibrium. Reductionism, in turn, delivered powerful analyses of parts and mechanisms, yet it often falters when asked to account for the circular causation, stability under perturbation, and anticipatory behavior that living systems so plainly exhibit. The challenge is not to re-mystify life, but to model its directedness without smuggling in final causes externally provided.

This is where careful distinctions matter. In Toward an ecological physics and a physical psychology, Turvey and Shaw argued that progress requires abolishing explanatory dualisms while still acknowledging forms of epistemological (not ontological) irreducibility and complementarity, i.e. limits of certain formalisms to capture closed, self-producing organization without a change of descriptive level. That spirit animates Francis Heylighen’s proposal: retain forward, mechanistic causation, but shift the explanandum from “mysterious purposes” to dynamical organization: the geometry of attractors, basins, and feedback that make purpose-talk testable rather than metaphysical.

The contemporary landscape makes this reframing urgent. On one side, teleological rhetoric reappears in new guises (e.g., Platonic “morphospaces” or “irruptions” of motivation), risking a slide back into ontological dualisms. On the other hand, purely programmatic accounts (“the gene as blueprint”) fail to explain plastic persistence under noise and unbounded creation of novelty. The middle path is to naturalize goal-directedness as a property of open, far-from-equilibrium systems organized so that many different beginnings converge on the same end and deviations are actively countered—without any backward causation or intelligent designer. That is the core wager of The meaning and origin of goal-directedness: a dynamical systems perspective, and it is the angle from which this review engages the paper.

What is goal-directedness?

Heylighen begins by clearing conceptual underbrush. Goal-directedness, in his usage, does not rely on a vital force, Platonic morphospace, irruptions of motivation, conscious intention, or “future causes.” It is instead a testable property: if we perturb a system and its trajectory robustly re-converges on the same end state, we are justified in modeling it as goal-directed. The classical worries—anthropomorphism, time-reversal, determinism—are addressed in turn. Irreversibility and nonlinearity make exact trajectory prediction in open systems unrealistic; what matters is that the family of trajectories originating across a region of state space ends in the same “goal,” despite disturbances. That is equifinality made operational.

With the pathologies removed, the paper collates five operational hallmarks: equifinality (many starts, one end), plasticity (adapting to varied initial conditions), persistence (returning to course after perturbation), negative feedback (counteracting deviations), and concerted action (coordinated multi-component dynamics). Each hallmark can be probed empirically (move the cheetah or the bacterium, watch whether it still eats or finds the gradient) and then mapped to dynamical structure. This is the key methodological move: replace teleology with geometry.

Cybernetics supplies the canonical mechanism: circular causation via negative feedback. A thermostat neither foresees nor represents a setpoint; its wiring ensures that sensed deviations trigger counteractions that reduce error. Natural systems generalize the same pattern without an external programmer: regulatory loops are intrinsic, targets are defined by the system’s own dynamics, and “purpose” names the stable outcomes of those loops. Crucially, the feedback is spatially circular (output re-enters as input), not temporally retrocausal.

This yields a crisp, non-mystical definition: a system is goal-directed when its dynamics embed an attractor that (i) lies far from equilibrium (requiring ongoing work to reach/maintain), (ii) has a large basin (supporting equifinality/plasticity), and (iii) is stabilized by feedback (supporting persistence). Purpose talk then abbreviates a causal claim about trajectories in a shaped landscape: the system’s own rules make it act “as if for” a goal.

Relatedly, cybernetic negative feedback can be married to contemporary inference stories. Kiverstein’s account of nested Markov blankets portrays autonomy as mutual enabling across coupled components; at the dynamical level this is precisely concerted action in a sparse, structured network. The point is not that organisms compute elaborate models, but that their control structure keeps each subcomponent in high-probability states, preserving the system’s integrity under perturbation; another way of specifying a big, forgiving basin.

Naturalizing goal-directedness

Heylighen formalizes the picture with standard dynamical-systems tools. States are vectors, evolution is an iterated map (or flow), and phase portraits reveal the arrows of motion. In reversible (idealized Newtonian) regimes, distinct beginnings diverge; in the irreversible, nonlinear regimes that typify biology, attractors and basins appear. The attractor is not a force “pulling”; it is a region that trajectories enter and cannot leave. Equifinality becomes a theorem about flows, not an anthropomorphic gloss.

Two refinements matter biologically. First, attractors need not be static points (homeostasis): limit cycles capture rhythmic maintenance (metabolic or predator–prey cycles), and extended manifolds approximate homeorhesis (staying on a growth trajectory). Second, resilience—the ability to “bounce back”—is basin geometry: broad basins, gentle boundaries, and error-suppressing feedback make persistence probable; brittle basins do not. In short, effective goal-directedness is resilience in the phase portrait.

What about concerted action? High-dimensional coordination across limbs, organs, or molecular pathways is represented as covarying components along a trajectory: sequential, parallel, recurrent, hierarchical, and networked dependencies. The cheetah’s run, the cell’s response, or an organism’s thermoregulation are all instances where many “knobs” turn together to keep the system on course. That is why concerted action is not an extra ingredient but a structural property of flows in large state spaces.

Still, if evolved systems look as though they “aim” at survival, where did such aiming come from? Here the paper points to chemical reaction networks and Chemical Organization Theory (COT): in reaction systems, organizations (closed, self-maintaining sets of species/reactions) correspond to attractors; sufficiently rich networks tend to evolve toward such organizations; and random perturbations push systems out of fragile attractors into more robust ones with larger basins. This frames the origin of goal-directedness as the statistical outcome of selection for self-maintenance in open reaction networks, not as the imprint of a designer.

On the biochemical side, COT’s “organizations” dovetail with classic closure ideas—cycles that regenerate their own catalysts and boundaries. The paper’s argument is strongest here: far-from-equilibrium closure explains why certain reaction webs are both self-maintaining and purpose-like in behavior. Where more work is needed is in quantifying concerted action (how much coordination is “enough”?) and in specifying boundary conditions (where is the system/environment cut?), issues that tie directly to individuality debates and holobiont cases.

Conclusions

Heylighen’s essay succeeds in relocating goal-directedness from metaphysics to mechanics. By anchoring equifinality, plasticity, persistence, negative feedback, and concerted action in attractors and basins of far-from-equilibrium systems, it makes “purpose” an empirical, perturbation-testable property. Teleology becomes unnecessary; dualism, avoidable; and “goals” become names for the stable outcomes of circular organization that expends work to resist disorder.

Two strengths stand out. First, the synthesis integrates cybernetics and dynamical systems with origin-of-life motifs: from thermostats to chemotaxis to general organizations, the same logic of circular causation and error correction explains directed behavior. Second, the resilience lens—goal-directedness as robust basin geometry—invites measurement: how wide are basins, how steep are boundaries, how fast are returns? Those are tractable questions for models and experiments alike.

Similarly, the dynamical picture clarifies a live worry: Does goal-directedness (one attractor) preclude open-endedness (ongoing novelty)? Not necessarily. Biological systems inhabit multi-attractor landscapes with hierarchies of ends: transient “subgoals” (feeding) lie in the basin of higher-order cycles (metabolism, reproduction). Novelty can arise by bifurcation (parameter change reshapes basins) or attractor switching (noise or learning moves the system between regimes). The picture is not “one basin to rule them all,” but layered control across timescales.

The account is admirably non-mystical yet leaves fruitful openings. It shows how goal-talk can be cashed out as attractor geometry and feedback, but it does not by itself solve questions about learning, innovation, or rule-level change (when a system alters its own dynamics). Those belong to the frontier where dynamical-systems descriptions meet syntax-open formalisms for self-modification. Still, as a naturalization of “purpose,” the present framework is a clean baseline on which such extensions can be built. Particularly, Hofmeyr’s (F, A)-systems allow the modelling of syntax-open systems that are operationally closed, which can be directly related with COT.

More open questions remain. We need principled metrics for concerted action in high-dimensional systems; sharper treatments of boundaries and individuality; and bridges to open-endedness where systems not only follow rules but also change them. None of these diminish the paper’s contribution. Rather, they mark the path forward: keep the causal arrow pointed from present to future, model organization in landscapes that reward self-maintenance, and treat “purpose” as a compact description of how resilient, closed-to-constraint systems navigate their worlds. On that program, this paper is both a clarifier and a useful launchpad.