Herbert Simon: a wealth of information, a poverty of attention

A product review begins. On the screens there are more dashboards than anyone is watching: cohorts, funnels, heatmaps, one line going up and another going down without it being clear which one matters. An assistant has summarised four hundred pages of user interviews in seconds. The backlog holds more "validated" opportunities than the team could build in a decade. There is, in short, an abundance of everything. And yet the meeting ends without a decision, or with a worse one than it would have reached on half the data and twice the silence.

The promise was the opposite. For years we took the scarcity of information to be the problem, and assumed that piling up more —more metrics, more signals, more content— would let us decide better. Now that artificial intelligence lets us generate information without bottom, that promise has quietly collapsed: we have everything we asked for and we decide worse. Someone named the reason more than half a century ago, before the internet, before the phone in our pocket, in an essay almost no one in product has read.

That someone is Herbert Simon. He is worth returning to, not to collect a fulfilled prophecy, but because his work holds an idea we still have not digested: that an abundance of information does not solve our problems but creates a new and harder one.

A man who did not fit inside a discipline

Simon (Milwaukee, 1916 - Pittsburgh, 2001) is one of those thinkers for whom the word "economist" is too small, even though he won the Nobel Prize in Economics in 1978. He also won the Turing Award, computing's highest honour, for his pioneering work in artificial intelligence. He founded, almost in passing, cognitive science. He taught for decades at Carnegie Mellon, writing interchangeably about organisation theory, the psychology of thought, the philosophy of science, and design. He did not change subject out of restlessness: he was always chasing the same one, decision-making, wherever it appeared.

And he wrote against a very specific figure. The economics and management theory of his time took for granted a character Simon spent his life dismantling: economic man, the perfectly rational decider who knows every alternative, computes its consequences, and chooses the optimum. That creature, Simon said, does not exist. It is not that people are irrational; it is that our rationality has a scale, that of a finite mind working with incomplete information, in limited time, and with a modest capacity to compute. His whole body of work comes from taking that limitation seriously rather than idealising it away.

His ideas spread so widely that they ended up diluted: today we repeat watered-down versions of Simon without knowing they are his. He is being read again now, whole and by name, for two reasons that converge. Because we built the information-rich world he described and ran straight into the limit he foresaw. And because, in building machines that decide, we have again stumbled on the very borders he drew for the human mind.

Rationality has limits

Simon's central idea, the one that holds up all the others, is called bounded rationality. Against the omniscient decider of the models, he described how people in an organisation actually decide: no one surveys the universe of options. A few are considered —those close to hand, those experience suggests— they are compared against a threshold of "this will do", and the first that clears it is chosen. We do not optimise. We cope, intelligently, within what the mind can hold.

Anyone who has led product recognises the mechanism, even if the methods we dress ourselves in say otherwise. No team lists every possible feature and computes which one maximises some value; three or four plausible paths are weighed against what the team knows and against what there is time to build, and a bet is placed on one. The prioritisation frameworks we venerate are, almost always, impoverished echoes of Simon: attempts to give the appearance of optimal calculation to what remains a judgement under constraints.

From here comes his view of organisations, which runs deeper than it looks. For Simon, an organisation is above all a structure for making decisions: hierarchy, communication channels, standard procedures are not inert bureaucracy but devices for decomposing problems too large for a single head and distributing them among many. When someone leading product decides which team owns which decision, or how strategy translates into priorities, they are doing exactly what Simon described: drawing the cognitive boundaries within which a knowingly bounded rationality will operate. It is a lens that helps you not to mistake a structural problem for a people problem, which is one of the costliest errors in scaling an organisation.

Satisfice, don't optimise

To the way of deciding I have just described Simon gave a name that caught on: satisficing, a cross between satisfy and suffice. To seek not the optimal but the good-enough. It sounds like surrender, like the conformism of someone who gives up too early. It is the exact opposite: it is the rational way to decide when optimising is impossible, because the cost of continuing to search for the perfect option far exceeds anything that perfection would add.

The distinction has an edge for our craft. Much of contemporary metrics culture is a hymn to optimisation: to squeeze out one more point of conversion, one more decimal of retention, as if improving a number were always good by definition. Simon invites a different question: has this metric already crossed the threshold at which it stops being the bottleneck? If so, continuing to optimise it is squandering the scarce resource —the team's attention— on polishing something that is already good enough, instead of redirecting it to the next constraint that truly limits the product. The team chasing the marginal optimisation of an indicator that has already done its job is not being rigorous: it is satisficing its own need for control, not the product's needs.

A wealth of information, a poverty of attention

If there is one line of Simon's that has travelled all the way to us, even if almost no one knows it is his, it is the one that gives this piece its title. He wrote it in 1971, in an essay with an apparently arid title —Designing Organizations for an Information-Rich World— at a time when talking about an excess of information seemed eccentric.

What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information sources that might consume it.

— Designing Organizations for an Information-Rich World (1971)

The move has a pure economic elegance. In a world poor in information, more information is wealth. But when information becomes abundant, it reveals a hidden scarcity: that of whatever information consumes. And what it consumes is obvious the moment it is said: the attention of the recipient. The real cost of information is not in producing it or transmitting it —that tends to zero— but in receiving it, in the time and the mental capacity the recipient spends. Designing organisations, or products, in such a world stops being a problem of producing more and becomes one of filtering better.

I have written elsewhere about this attention crisis that now defines almost any digital product, and this line was already at its centre. What Simon saw in 1971, with a handful of computers in the world, has become the condition of our age. Every feed, every recommendation system, every search engine negotiates its constraint, whether it names it or not. And artificial intelligence takes it to the extreme: when generating plausible content costs practically nothing, the only variable that remains scarce, expensive, impossible to manufacture, is the human attention that content lays claim to. Producing more information in an already saturated world is not creating value; it is adding demand to a resource that can no longer keep up.

Everyone designs

The last piece of Simon I want to recover is the most philosophical and, for product, the most liberating. In The Sciences of the Artificial (1969) he argued that there is a class of objects —the artificial ones: products, organisations, software— that are not governed by natural laws but by the fit between their inner constitution and the outer environment they serve. Studying nature is not enough to understand them, because they do not describe what is but what could be and how to bring it about. What is needed, he said, is a different science: a science of the artificial. And at its core lies design.

Everyone designs who devises courses of action aimed at changing existing situations into preferred ones.

— The Sciences of the Artificial (1969)

The definition is deliberately broad, and that is its force. Under it, the engineer tracing a bridge, the physician prescribing a treatment, the person redrawing an organisation, and the one leading a product all design alike. Design stops being a minor, decorative craft subordinate to "real" engineering and reveals itself as the central intellectual activity of all the professions. Anyone leading product should feel addressed: what they do is not to administer a factory of features but to design an artificial system —the fit between what the product is on the inside and what its environment demands on the outside— under the limits of a rationality that never fully encompasses it.

From that view comes his explanation of why the complex systems that survive are almost always organised into semi-independent modules, each able to evolve and to fail without dragging down the whole. It is the deep reason why the organisations that make software structure themselves, when they work, around bounded domains: not out of fashion, but because a system —technical or organisational— whose pieces cannot change separately is too fragile to adapt in time.

Where it falls short

It would be unfaithful to Simon to turn him into an oracle; he, who distrusted omniscient rationality, would be the first to point to the edges of his own.

The most visible lies in his treatment of attention. Simon thinks of it as a scarce but homogeneous, measurable resource, almost like a currency handed out in interchangeable units. But not all attention is equal: the distracted attention of someone thumbing through a feed and the deep attention of someone solving a hard problem are not the same substance in different quantities, they are different things. His model, moreover, says little about power: about who has the right to claim whose attention, and how hierarchy and politics decide that allocation long before any calculation of efficiency.

There is a second crack, subtler. By taking the limits of the mind seriously, Simon tends to naturalise them, to treat them as a fixed given of the problem. But often the most valuable work is not deciding well within given boundaries, but moving those boundaries: reframing the problem until the constraint that seemed impassable dissolves. Bounded rationality describes beautifully how to cope inside a box; it says less about when and how to break it.

And there is a third, familiar to anyone who makes product. Simon's science of the artificial assumes a relatively stable, well-specified environment to which the artefact must fit. The digital is rarely like that: software again and again dissolves the illusion of a problem that holds still, because the artefact and its environment transform together, and it is often the product itself that changes the environment it then tries to adapt to. In that dance, "designing the fit" is chasing a target that moves because we move it.

Why it matters today

None of this invalidates Simon. It situates him, which is what he would do. And it explains why, situated, he remains one of the most useful thinkers for understanding the present moment.

His most immediate lesson is the one about attention. If the scarce resource is not information but the capacity to attend to it, then the good product is not the one that offers the most but the one that respects that limit: the one that filters, orders, and protects the attention of the person using it instead of auctioning it to the highest bidder. In an economy that makes money by capturing attention, designing to save it is almost a countercultural gesture, and probably the most honest line of differentiation a product has left.

The second lesson is more uncomfortable, and artificial intelligence brings it. It is tempting to believe that machines abolish bounded rationality: at last, a decider that really can process everything. But today's models do not escape Simon, they industrialise him. They have a finite context window, they decide on incomplete information, they make do with the good-enough answer rather than the optimal one. We have not built economic man in silicon; we have built, at another scale, another bounded rationality, and at the same time we have multiplied the flood of information that saturates our own. Simon's question —how to allocate a scarce attention across an abundance that keeps growing— does not vanish with AI: it becomes the central design problem of the decade.

Even the vocabulary hides an irony. The architecture that made these models possible was introduced in 2017 in a paper titled, without ceremony, Attention Is All You Need. The machine born of the information flood carries at its technical core the very word Simon placed at the centre of the problem —though they do not mean the same thing: for Simon, attention is the scarce resource that abundance exhausts; for the transformer, the mechanism that decides which parts of the input to look at. That both land on the same word does not feel entirely accidental: each, in its own way, is about allocating a focus that never stretches to everything.

And there remains the deepest lesson, the one that gives the craft its dignity. Leading product is not managing a queue of tasks: it is designing an artificial system under constraints, deciding with a limited mind in an environment that will not hold still, choosing what to give the scarce attention available. Simon tells us that this is not a lesser evil to be overcome with more data or more machines, but the very condition of any serious design. You work with what the mind can hold, not with what an ideal mind could.

What is still scarce

Let us return to the meeting at the start, with its overflowing screens and its decision that never comes. The abundance that overwhelms us is not a failure of our tools: it is precisely the world Simon foretold, the information-rich world, fulfilled with a fidelity he did not live to see. What we failed to hear was the second half of his sentence: that such a world inevitably brings a new poverty with it.

The scarce resource was never information. Now that we can manufacture it without bottom, this has stopped being a theoretical subtlety and become the practical question that orders the craft. Simon left us not a recipe for producing more, but a warning about what is truly missing. And what is missing, today more than in 1971, is a human capacity to attend and to judge that no abundance can manufacture. The only question that matters, at the end of that meeting, is not how much more information we can gather. It is what deserves the scarce attention we have left.