Much of modern economic thinking rests on the idea of rational actors. Not as an explicit claim, but as an ambient assumption. Policy, markets, and institutional design rely on the belief that individuals act deliberately, evaluate tradeoffs consciously, and make decisions in ways that can be modeled and anticipated.
This construction does more than describe behavior. It builds a wrapper of implied outcomes around a given context. Once rationality is assumed, certain results become expected, certain deviations become explainable, and certain interventions become justified.
Over time, the assumption hardens. What begins as a simplifying model becomes a framework through which behavior is interpreted and managed.
Recent research suggests that a meaningful portion of agency is exercised below the level of conscious deliberation. Decisions are often made while attention is allocated elsewhere, with reasoning applied after the fact rather than before. The mind narrates what the system has already done.
This creates a quiet problem. If agency is partially subconscious, and rationality is inferred rather than directly observed, the distinction between behavior and explanation becomes less stable than our models imply.
The risk is not that rational actor theory is imperfect. It is that once distributed beyond theory and embedded into real environments, it begins to shape the very behavior it is meant to describe.
How, then, do we draw a line between behavior and implied rationality if the degree to which agency is executed is not readily knowable?
Consider a simple thought experiment. A group of undergraduate students is asked to track their spending over a period of time. Half of the group records only what they purchase. The other half records both the purchase and the reason for making it.
It is reasonable to assume that observation alone will influence the first group. Awareness often alters behavior. The more interesting question concerns the second group. Does the requirement to justify a decision change the decision itself?
At that point, observation is no longer passive. The introduction of explanation reshapes intent. Rationality becomes something to be performed rather than something to be exercised.
This distinction matters because explanation carries its own incentives. Decisions that can be justified cleanly are favored over those that cannot, even if the latter are more effective. Behavior adapts not to outcomes, but to legibility.
Once this dynamic is present, rationality ceases to be a neutral descriptor. It becomes an environmental force.
In markets and policy systems, the effects compound. Participants learn which actions are rewarded under scrutiny and which invite friction. Over time, behavior converges toward what can be explained rather than what is optimal. Agency remains, but it is expressed through narrower channels.
What appears as discipline is often adaptation to observation.
This does not imply that oversight is misguided or that rationality is an illusion. It suggests something more precise. The act of observing behavior, and especially the act of requiring justification, changes the system in which decisions are made.
Outcomes, in such environments, reflect not only underlying preferences or information, but the structure of attention applied to them.
The dilemma is not whether observation influences behavior. That much is evident. The dilemma is whether systems designed to evaluate rational action can distinguish between genuine agency and the behavior produced by being watched.
Once that distinction blurs, evaluation becomes circular. Rationality is inferred from outcomes that were shaped by the expectation of rationality itself.
In these conditions, the most consequential design choice is not how decisions are judged, but where explanation is required. Systems that demand constant justification encourage different behavior than systems that allow intent to persist without interruption.
The question is not how to observe more closely.
It is how to observe without interfering.
That tension cannot be eliminated. But it can be acknowledged. And acknowledging it changes how decisions, accountability, and long-term outcomes are understood.