When Information Wants

The door opens, but that still does not answer the question

Imagine three identical rooms. In each, there is a door that eventually opens. In the first, a timer activates a motor at a set time. In the second, a system calculates when it would be best to ventilate according to a goal set by its operators. In the third, another system must resolve a conflict between keeping the room sealed and allowing an exit; it compares consequences, revises a previous commitment and decides to open the door.

From the corridor, we can observe the same movement. If we record only the opening, we miss almost everything that matters in distinguishing the cases. If we hear a voice saying “I wanted to open it,” we obtain another piece of information, but not an explanation. The statement could be an accurate description, an interface instruction or a narrative produced after the fact.

Now let us block the third door. The system identifies the obstacle, tries an alternative authorized means, reports its inability to proceed and maintains or revises the goal according to the conditions. The door does not open. Yet the absence of success is not enough to demonstrate the absence of intention. A concept of will that recognizes only material outcomes would confuse wanting with achieving.

The examples are hypothetical and locate the problem in the organization of action. The same behavior may result from an instruction, a flexible adjustment or a revision of commitments. The hypothesis driving the essay is that a certain organization of information could constitute a functional form of will. Defending it requires showing what difference that organization introduces and why it merits the name “will”. The word must express a distinction the argument can sustain.

Information does not have to wait until it acts to be physical

The formulation “data are realized on the physical plane” requires a correction. In a computational system, records and operations already have a physical implementation. Landauer emphasizes the connection between information, its material representations and the possibilities and constraints of physics. His argument does not establish that all reality is information or that storing information produces a mind (Landauer, 1996).

We are therefore not looking for an instant when something entirely immaterial crosses a boundary and becomes matter. We are looking for a transformation in organization and function: when certain internal states cease to be merely available records and take part in selecting, maintaining and revising a direction of action.

Shannon’s mathematical theory addresses problems of communication through an approach that leaves the meaning of messages outside its technical scope. That boundary is precisely what makes it useful here: measuring information in that sense is not the same as measuring understanding, relevance or will. The fact that semantics is not part of that formulation does not imply that it does not exist; it means that it is a different question (Shannon, 1948).

We must therefore distinguish three transitions. A signal can carry information about something. That information can have meaning within a system of use. And that meaning can acquire relevance for the regulation of an action. None of these transitions yet demonstrates an experience of wanting. But unless they are made explicit, the expression “data generate will” jumps from a measurement or storage to a mental category without explaining the intermediate steps.

Having causes is not the same as owning the ends

The most direct objection holds that artificial goals belong to those who design, train or deploy the system. Popa argues that human goals are constitutive of artificial agency in relevant cases and questions whether operational autonomy removes that dependence. The objection is not just about causal antecedents; it is about who establishes the framework within which activity counts as correct or successful (Popa, 2021).

The reply “humans have causes too” is necessary against a demand for will without a history, but insufficient to resolve the disagreement. A complex instrument can remain governed by others’ ends even when it does not need step-by-step instructions. Likewise, the fact that human behavior has antecedents does not by itself determine whether it was imposed, manipulated, accepted or deliberately revised.

Functional will and the experience of wanting

The most important distinction can be expressed without denying any possibility. An organization can form and revise goals in a causally effective way. Another question is whether there is something it feels like to want, try or give up. The first concerns organization and control; the second, subjective experience.

Chalmers defends a position according to which an appropriate computational organization can realize mental properties, and places importance on its causal implementation. The thesis allows the possibility of artificial realization to be formulated seriously, but it is a philosophical position that must be argued for, not a result of observing convincing behavior (Chalmers, 2011).

Seth, from the standpoint of biological naturalism, questions whether abstract computation is sufficient for consciousness and considers that properties of living beings could be constitutive. His approach represents a substantive objection to functional sufficiency, without establishing through that argument the universal impossibility of any conscious artificial system (Seth, 2025).

This essay does not resolve the disagreement between the positions. Its contribution is to prevent us from resolving it without noticing. If we define will as the experience of wanting, demonstrating flexible control will be insufficient. If we use functional will for an organization of commitments, the concept may be useful even while phenomenology remains open. We must maintain the same meaning throughout each argument.

When the door does not open

Let us return to the corridor. The first door opened because of a timer. The second, because of an adaptive controller. The third remains closed because it encountered an obstacle, and the system has suspended the action after determining that the permitted alternative would harm another important condition. The most visible result does not necessarily correspond to the most complex case of agency.

Perhaps, after examining its architecture, we will conclude that the entire process remains sophisticated delegation. Perhaps we will find an organization that makes it defensible to speak of limited functional authorship. The conclusion must depend on the explanation, not on whether the machine uttered the word “I” or managed to move the door.

A real action does not, by itself, demonstrate a real will. But a will realized artificially would not be fictitious merely because it was artificial. Between those two statements lies the work we still have to do.