Personal Blogs

Bounded Rationality and the Boundary of the Market System

Herbert Simon's concept of bounded rationality holds that decision-makers do not — and cannot — process all available information before acting. Time, cognitive capacity, and the complexity of the system in front of them impose hard limits. Rather than optimising, actors satisfice: they select a decision that is good enough given what they can actually know, not the theoretically perfect one. This is the standard reading, and it is not wrong. It is, however, incomplete, because it treats "bounded" as a property of the individual mind rather than a property of a system boundary that the individual — often without noticing — has already drawn. 

(direct video / playlist)

Simon developed the concept in the 1950s as a corrective to the assumption, standard in classical economics, that agents are fully rational optimisers with complete information and unlimited processing power. His argument was not that people are irrational, but that the demand classical theory placed on human cognition was never realistic to begin with. Decision-makers, in Simon's account, act as satisficers: they search until they find an option that clears a threshold of "good enough," and then stop searching, rather than continuing until they have located the genuinely optimal choice. This distinction between satisficing and optimising is often flattened in popular use of the term, where "bounded rationality" becomes shorthand for "people make mistakes because they don't have enough information." That flattening loses almost everything useful about the concept.

It is also worth separating Simon's formulation from the behavioural economics tradition that followed it, particularly the heuristics-and-biases work associated with Kahneman and Tversky. That later tradition catalogues the specific, often predictable ways bounded reasoning deviates from a normative rational standard — anchoring, availability, loss aversion, and so on. It is a psychology of error. Simon's original concept is closer to a structural claim: given finite time and finite processing capacity, no useful theory of decision-making can assume unlimited information-gathering, regardless of whether the resulting decision happens to be "biased" in the technical sense. The systems reading developed here sits closer to Simon's original structural claim than to the biases literature, because the interesting move is not cataloguing where traders go wrong, but asking what determines the boundary of what any given trader is even in a position to consider.

From cognitive limit to boundary judgement

Read through a systems lens, bounded rationality stops being a fact about cognition and becomes a question about boundaries. An actor is only ever bounded relative to a system they have implicitly defined for themselves, and that definition is a choice, not a given, even when it is made unconsciously. This is the move Werner Ulrich's critical systems heuristics makes explicit: before asking whether a decision was rational, ask what was drawn inside the boundary of "the system" for that decision, and — just as importantly — what was deliberately or unconsciously left outside it. Ulrich's boundary questions were developed for social planning and public policy, where the stakes of a badly drawn boundary are visible and often severe: whose interests count, whose knowledge is treated as legitimate, who bears the consequences of a plan that never consulted them. They translate cleanly to markets, because a market is nothing more than a set of actors each operating with a different boundary around what counts as "the system" they are trading inside, and each treating their own boundary as though it were simply "the market," rather than one partial construction of it among several.

This reframing matters because it changes what "more information" can actually do. Under the popular reading of bounded rationality, the fix for a poor decision is more research — read the filing, check the float, watch the tape more closely. Under the boundary reading, more research inside an unexamined boundary just produces a more detailed picture of the same partial system. It does not correct the boundary itself. A trader who spends three hours refining their read of price action has not necessarily become less bounded; they have become more thoroughly informed about the specific, narrow slice of the system their boundary already permitted them to see. The boundary, not the depth of research within it, is very often the actual constraint.

Whose boundary, drawn where

Apply that to a trade. Most retail analysis draws the boundary around price, volume, and float — the visible mechanics of the chart. Bounded rationality, on that boundary, looks like a data problem: the trader didn't have enough information about the stock, and more research would have closed the gap. But widen the boundary to include the broker's margin requirements, the market maker's inventory risk, and the regulator's disclosure rules, and a different picture appears. Each of those actors is bounded too, by a boundary drawn around their role, not the trader's. The retail trader's "irrational" entry and the market maker's "rational" hedge can both be locally sound decisions, made by actors whose system boundaries simply don't overlap. Neither actor is short of information in any absolute sense. They are short of information relative to a boundary they never chose to widen.

This is closer to a SODA-style reading — Ackermann and Eden's Strategic Options Development and Analysis — than a purely economic one. SODA's premise is that group decision problems are rarely disagreements about facts; they are disagreements between individually coherent cognitive maps that were never reconciled, because nobody surfaced the boundary judgements each party was quietly operating under. The trader who says "the fundamentals didn't justify that move" and the market maker who says "the move was a rational response to short-dated hedging pressure" are not disagreeing about the facts. They are working from boundary judgements that never converge, because neither actor drew the system the same way — and, critically, neither one is aware that a boundary judgement was made at all. That invisibility is usually the actual problem, not the rationality of either party. A cognitive map, in SODA's sense, is not a distortion of the "real" system; it is a legitimate, internally coherent construction of it, built from a particular vantage point. There is no neutral vantage point from which to declare one map correct and the other mistaken. There is only the question of which boundary each map was drawn within, and whether that boundary was ever made explicit.

Why the boundary itself is not static

There is a further complication that a purely economic reading of bounded rationality tends to miss: the boundary that defines what an actor considers "the system" is not fixed for the duration of a trade. It moves as the situation develops, in the same way a Viable System Model's System 4 function is meant to continuously scan the environment and revise the organisation's model of what's relevant to it (Hoverstadt, 2020). System 4, in VSM terms, exists precisely because a viable system cannot survive on a single, fixed model of its environment; it needs a standing function whose job is to notice when the environment has changed in ways the current model does not account for, and to feed that back into how the system organises itself. An individual trader rarely has a formal System 4 function in this sense, but the absence of one is exactly the failure mode this section is describing: a thesis formed on Monday's boundary, and never revisited, behaves like a viable system that has switched off its environmental scanning.

A thesis built on Monday's boundary — this stock, this float, this catalyst — is not obliged to still hold by Wednesday, because the boundary itself has likely shifted, pulling in actors, constraints, or information that weren't inside the frame when the position was opened. New short interest data is published. A market maker's hedging book changes shape. A regulator opens an inquiry that was never part of the original picture. None of this necessarily falsifies the original thesis in the way a single piece of contradictory information would; it does something more structural, which is to change what the relevant system even consists of. Treating an entry decision as a fixed, defended position, rather than a boundary judgement open to revision as the system reveals more of itself, is one of the more expensive misreadings of what bounded rationality actually implies for practice. It converts a structural feature of decision-making under uncertainty — that the boundary moves — into a personal failure to have been right at the outset.

Boundary judgements and the pattern itself

This also explains something about chart patterns that is easy to miss from inside a purely technical framing. A chart pattern is, among other things, a record of what was left inside a particular boundary — price and volume, nothing else — compressed into a visual shape. It is not that the pattern is false. It is that the pattern is a boundary judgement rendered as a picture, and it inherits every limitation of the boundary that produced it. Two setups can look identical on a chart while sitting inside entirely different wider systems: different borrow availability, a different macro backdrop, a different mix of institutional participants. The pattern repeats because the visual boundary — price against time — is narrow enough to produce the same shape from genuinely different underlying conditions. Reading the pattern without asking what has been excluded from the frame that produced it is, in effect, mistaking one actor's boundary judgement for a complete description of the system.

The practical consequence is specific, not merely philosophical. Widening your own boundary judgement — deliberately asking who else's constraints are shaping this price, and why their bounded rationality might look nothing like yours — does more to explain an apparently "irrational" move than assuming someone, somewhere, simply lacked information. The information usually existed. It just sat inside a different actor's boundary, governed by a different set of constraints, and was never going to be visible from where the original decision was made. The discipline worth building is not "gather more data." It is "ask whose boundary you are currently trapped inside, and what you would see if you deliberately moved it" — and to keep asking that question for as long as the position is open, not only at the moment it was opened.

How bounded decisions aggregate into a feedback loop

None of this means the individual boundary judgement is the end of the story. Boundaries matter for a second reason that a purely cognitive account of bounded rationality has no real way to capture: bounded decisions do not stay isolated. They aggregate, and the way they aggregate can produce dynamics that no individual actor intended or fully understood while it was happening. A short squeeze is the clearest illustration available in markets. No single short seller decides to trigger a squeeze. Each one is making a locally bounded decision — cover now, at this price, given this margin call, this borrow cost, this risk limit — using information and constraints specific to their own boundary. None of them is reasoning about the aggregate effect of everyone else doing the same thing at roughly the same time. And yet the sum of those individually bounded, individually reasonable decisions produces exactly the reinforcing feedback loop that defines a squeeze: covering pushes price up, which forces more covering, which pushes price up further, each step driven by actors who were never modelling the loop itself, only their own narrow slice of it.

This is a genuinely systemic property, not a cognitive one. It cannot be explained by saying any individual actor was insufficiently rational, because each actor's decision was perfectly sound relative to their own boundary. The loop emerges from the structure of how many separately bounded actors are coupled to the same price, not from any failure of reasoning inside a single mind. This is precisely the kind of phenomenon that a boundary-based reading of bounded rationality is equipped to explain and a purely cognitive one is not: the interesting object of analysis stops being any individual trader's decision quality, and becomes the structure connecting many differently bounded decisions to one another. Understanding a squeeze, in other words, is less about diagnosing anyone's rationality and more about mapping whose boundaries are coupled to whose, and how tightly.

A working boundary audit

Turned into something usable rather than only descriptive, Ulrich's boundary questions suggest a short, repeatable audit that can be run against a position, ideally more than once across its life rather than only at entry. Four questions do most of the work. Who is the boundary of this analysis actually built around — is it drawn around the chart, around the company, or around the full set of actors with a stake in the price? What has been placed outside that boundary that could plausibly move the price regardless — borrow availability, a regulator's attention, a market maker's inventory position, a fund's mandate constraints? Whose bounded rationality is currently doing the most work in this move, and is it the same actor whose bounded rationality was doing the work when the thesis was first formed? And finally, if the boundary were deliberately widened to include the actor currently most affected by this price and least visible from the original vantage point, what would the position look like from there?

None of these questions produce a number, and none of them replace the mechanical analysis — the float, the borrow, the catalyst — that any position still needs. What they do is make the boundary itself an object of deliberate attention rather than something absorbed unconsciously along with whichever framework was used to first find the setup. Given that the boundary, not the depth of analysis inside it, is very often the actual constraint on a bounded-rational decision, that shift in attention is not a peripheral addition to the analysis. It is closer to the analysis that was missing.

This is explored further, with direct application to trading decisions, in the video above.

Regards,

Russell Larke

BA (Hons) Business Management | MSc Candidate (Systems Thinking)
Trading Beyond Charts