Why ask why?

2 July 2026

This is the latest version of an essay for EECS 470 students at the University of Michigan. This class on computer architecture is for the next generation of engineers, who are starting to familiarize themselves with the technologies and narratives around computer engineering. My other objective is for the readers to reflect on where and whether these narratives continue to fit as the future unfolds.

If you listen to podcasts from people like Sam Harris or Scott Galloway, you might have heard them implore the audience to hold two or more, often contradictory, ideas in their minds at the same time. They do this to encourage their audience to see the bigger picture and think more deeply about a subject. F. Scott Fitzgerald wrote that the test of a first-rate intelligence is the ability to hold two opposing ideas while still retaining the ability to function. He followed it with a personal application: that one should, for example, be able to see that things are hopeless and yet be determined to make them otherwise.

My emphasis is different: being able to hold multiple mental models about a subject in your mind is necessary to improve your own knowledge. It is also a requirement if you want to solve complex problems and become good at what you do. It is a route to creativity, which, according to David Deutsch, leads to new and better explanations: “The real source of our theories is conjecture, and the real source of our knowledge is conjecture alternating with criticism.”

The significance of what Deutsch says is that there is no need to have access to all the facts up front to develop better explanations, as long as one engages in a critical process to extract them from the imagined. The converse of this is also true: the facts don’t just get handed to us on a silver platter; we have to get to them through criticism, otherwise, how do we know they aren’t just imaginings? What we need to look for is coherence between hypotheses and the facts, and be ready to update mental models if they don’t tell a coherent story.

In essence, being good at juggling mental models helps both with hypothesis generation and their creative destruction, enabling better and faster problem-solving.

Why tell you this?

For one thing, I don’t think the importance of creativity in problem-solving has been sufficiently advocated. Applying creativity usually saves time and effort and improves results. However, some people think they can’t be good at it because they just weren’t born that way. John Cleese disabuses you of that notion and shows that creativity is “a way of operating.” It’s a mood that you can put yourself in, not a special gift that you are born with. And John Cleese should know, as his father’s last name was originally Cheese, which was creatively changed to Cleese out of embarrassment.

For another, I don’t think the scientific process is well understood by most people, despite being a driving force of our world since the Enlightenment. We usually hear of it as something that serious scientists do, but most people would probably be hard-pressed to answer what makes a process scientific and another not. People engage in this process even when they gossip: they form conjectures and evaluate their conclusions, resulting in shared mental models. And if they make predictions that could, in principle, be proven wrong, then one could argue that what they are engaging in is a direct application of the scientific process (Karl Popper, in The Logic of Scientific Discovery, distinguishes genuine science from pseudoscience by this criterion of falsifiability).

Third, we learn and reason through narratives. Sure, we can memorize facts, but those facts don’t translate into knowledge unless we know how the pieces of information are connected. Not only do these stories make most facts easier to remember, but they also serve as templates for new conjectures that help make sense of new information. In an important sense, the narratives we understand are more important than the facts they are based on, as they help us make educated guesses about situations we haven’t encountered before. Understanding the historical context of ideas helps us spot opportunities for building on them in future contexts.

To illustrate this, I will start with a sketch of the world from behind the Iron Curtain as it existed almost fifty years ago. You might think that the personal stories are self-indulgent or that this bygone era is irrelevant today, but if you look closely, many of them have modern parallels. You will need to see if you can spot them in your own lives. When I talk about the advantages of information-poor environments, you should also understand that all environments, even in the West, were that way relative to today, and the only way to attempt to recreate their advantages is through judicial filtering.

Useful narratives often transcend domains: I will also link some of the observations to concepts in computer science and engineering. If you are not yet familiar with them, they are worth investigating and incorporating into your mental models.

The 1980s were the decade in which computer technology became widely accessible, even behind the Iron Curtain. This is the era in which much of the collective subconscious of expectations about how technology would evolve began to take shape. That subconscious is still firmly installed in many heads and shapes actions. So understanding the narratives and the world that gave rise to them is important to predict the future.

East vs. West

Not having access to facts was a well-understood part of life in Budapest, Hungary, in the 1980s. At that time, it was a communist country behind the Iron Curtain, with limited access to information. The world was divided into two starkly different places: the free and democratic West, where people could discuss their ideas, travel, and live in prosperity, and the communist dictatorships that controlled what people were allowed to do, sanctioned people’s thoughts and movement, and left people living in relative poverty. There was a wall dividing the two worlds in Europe, called the Iron Curtain, which was an empty border area patrolled by guards and mined to prevent people from escaping from the east to enjoy the luxuries of the west. The official reason for it was different: it was to prevent the imperialist Western aggressors from invading the peace-loving people of the Eastern bloc.

The Eastern bloc was controlled by the Soviet Union, the West by the United States of America. What prevented the two sides from going to war was the concept of mutually assured destruction, enabled by the vast arsenals of nuclear weapons they built up, and Game Theory. The idea behind its application was to impose rationality on the two blocs, not to start a war, by credibly assuring the other side that, if they did, they would be wiped out as well. The war machines were automated such that there would always be a second strike if either side were foolish enough to make the first. The war was cold because it avoided direct conflict by establishing well-defined boundaries, as both sides knew how dire its consequences would be. By the way, in case you didn’t know, MAD is the doctrine still in place today.

Mutually Assured Destruction led to a relatively stable world. By 1972, the SALT I treaty was signed to limit the growth of nuclear weapons (why build more weapons if they are already sufficient to destroy the world?), which also limited missile defenses; the logic of MAD only holds if the balance of power is maintained.

In the 1980s, it was becoming obvious that the West had won the Cold War and that the East was going broke, but no one anticipated that the Eastern Bloc would dissolve by the end of that decade. There were short-lived revolutions before: Hungary in 1956, Czechoslovakia in 1968, but the Solidarity movement took hold in Poland in the early 1980s, and it stuck; the government couldn’t make it go away. In 1989, Hungary dismantled the Iron Curtain on its borders without retaliation from the Soviet Union, and the Berlin Wall fell later that year.

When you live in an environment like that and are interested in understanding your world, you can only do that by holding and challenging multiple mental models. For one thing, there was a model based on personal experience. There were other models put forth by the party, others by the Western messages that filtered into the country, and then models from people around us, from many generations and life experiences.

You could make simple observations, and there was no one to give you a definite and objective answer. And if they did, you would try to model what their agenda could be with that answer. Here is an example: how come countries that seem like problem places and dictatorships call themselves democratic, while no countries that seem like functioning democracies do? This, in turn, can lead to envisioning what that name conveys from the perspectives of the dictators, their peer groups and enemies, the people they rule, and pretty much everyone those words could influence. You hit gold if, in an investigation, you found out something that didn’t fit your previous models: you could then test it by trying to convince others that yours was more accurate.

Different perspectives exist in all environments. What made the situation different back then was an intrinsic experience that all perspectives are wrong and that better perspectives can be developed by pitting the models against each other. People were “pitted” because none of the sources of “truth” could be trusted.

The 1980s

Television sets in Hungary were bulky and showed only black-and-white pictures. Signals on the aerial were affected by weather patterns and other mysterious forces, so channel tuning had to be done continuously. There were two TV channels: one showed documentaries; I recall grainy shots of World War II, and the main one, which operated most days except Mondays. In the evenings at 7pm, there was “Exercises for the family,” a program during which the audience could get off the couch and do some aerobics in front of the TV. At 7:15pm, there was a cartoon or two for children, and at 7:30pm, the evening news. At 8pm, there was a movie, often imported from the West, that offered a welcome distraction from reality and an opportunity for small talk the next day. Everyone watched the same things if they could afford a TV.

The evening news was a key way the party could influence perception and mood in the country. There was only one party: the Hungarian Socialist Workers’ Party, the name signaling that, while the country was committed to the Marxist-Leninist communist utopia, it hadn’t quite arrived there yet. Nonetheless, people referred to them as the communists. People knew that the news was crafted to influence them, so people went to great lengths to discern why certain things were said and, more importantly, what the significance of what wasn’t said was.

Despite control of the communication channel, the government could not fully control the message. One problem they faced was having to tune the message to multiple constituencies simultaneously. For one thing, there were the Soviet supervisors who monitored and evaluated how the Hungarian communists ran things. For another, they had to manage how different groups of people would react to the news. What would the workers think? What would the young people hear? What about the communist elites in the 2nd district? Most people knew that they were being manipulated, and figuring out the intentions behind the words became a game.

By the 1980s, the regime was pushing boundaries with the Soviet Union on how to set policy to keep its population happy. While the news was controlled top-down by the party, people had become quite good at reading between the lines. In fact, it seemed that even the news anchors would use double-speak to convey official messages unofficially so that their Soviet censors (and others) wouldn’t notice.

I vividly remember my grandfather’s reactions while watching the evening news, and his commentary helped me understand that there were multiple versions of the truth, each describing aspects of the world around us. These truths emphasized different points of view but were usually not in direct conflict with one another when presented. However, if I could somehow detect that these messages were aimed at different audiences, I could build my own mental model for each and then extrapolate to what would make sense for someone holding that view.

After a while, I became pretty good at predicting both the news narrative and my grandfather’s take on things. A little later, the Western perspective on the same topics also became increasingly accessible. Having access to diverse perspectives allowed me to develop my own models of the world. But instead of having to align myself with one or the other, the environment was conducive to maintaining a balance, as society was generally disillusioned with “THE TRUTH.”

Game Theory

It wasn’t until about a decade later that I found out that losing control of their message wasn’t just due to the party’s ineptitude; full control was actually impossible to achieve. The difficulty of the manipulator was first noted in 1947 by Neumann and Morgenstern, who laid the foundations of Game Theory. They wrote in “The Theory of Games and Economic Behavior” that in general two or more functions cannot be maximized at once, as “in general, one function will have no maximum where the other function has one.” This means that if there are multiple audiences to a story, and the storyteller intends to please all of them, some of the audience will be more disappointed than others. Neumann and Morganstern gave the following example:

A particularly striking expression of the popular misunderstanding about this pseudo-maximum problem is the famous statement according to which the purpose of social effort is the "greatest possible good for the greatest possible number." A guiding principle cannot be formulated by the requirement of maximizing two (or more) functions at once.

Such a principle, taken literally, is self-contradictory, (in general one function will have no maximum where the other function has one.) It is no better than saying, e.g., that a firm should obtain maximum prices at maximum turnover, or a maximum revenue at minimum outlay. If some order of importance of these principles or some weighted average is meant, this should be stated. However, in the situation of the participants in a social economy nothing of that sort is intended, but all maxima are desired at once by various participants.

Unfortunately for wanna-be manipulators, a message’s audience participates in receiving the message: they get to refine their own interpretations. Sometimes this aligns with the intended message, sometimes it does not. The lucky thing for me was to be in an environment that wasn’t too scared and was interested in looking behind the curtain to discern the truth so that I could witness this predictive, second-guessing process firsthand. I came to understand that there are usually multiple versions of the truth and that I could develop techniques for discerning which were closer to reality.

The method to accomplish this was simple: I just had to keep asking why. Obviously, our TV wouldn’t answer, but I could gather opinions from people around me to build internal models of what reality the message was intended to obscure.

The Zipper

To visualize this process, the image that comes to mind is a double zipper, with the two sliders moving from the zipped center toward the open edges. One slider represents the new information acquired, and the other the context that makes sense of it (the whys). The zipper itself represents a particular topic. Each tooth that’s locked is a nugget of information (a belief) that expands one’s knowledge. The more information is received, and the more questions are answered, the longer the region zips. If, however, there is new conflicting information, the zip goes no further; in fact, it may open regions that had been locked before.

So why are there two sliders in my metaphor? One slider represents the information that someone gives me; it is under the control of the source: an advisor, a teacher, a reporter, a marketer, or perhaps a master manipulator. The information pieces build on each other and grow as the slider closes the teeth. The other slider is under my own control: I can check for consistency in the information I receive, create and compare hypotheses about it, and close some teeth if my mental model seems to fit. All I have to do is challenge the information received. The easiest way to do that is to ask questions. The information-receiving slider is imagined as a passive process: the sender curates the messages, sends them to the receiver, and hopes they land as intended. The information synthesis slider creates hypotheses about the messages, fills in the blanks, and checks for consistency. If the messages and the hypotheses are consistent, then the sliders can be moved. If a later consistency check fails, the teeth can be opened, the received information discarded, and new hypotheses generated.

Surprise

The process I describe using zippers can be observed in children of a certain age. There is a period, roughly ages 3 to 5, when children ask relentless "why" and "how" questions as they build a model of cause and effect. It's driven genetically to enable cognitive development. Children selectively probe gaps in their causal understanding and even evaluate who gives good answers. The modern term for this is explanation-seeking curiosity. Researchers like Bonawitz, Shafto, Gopnik, and Schulz model the questioning child as a Bayesian learner whose questions are actions chosen to reduce uncertainty. They use the term Bayesian Surprise to describe a situation in which an observation has low probability under the child's current model; such surprising evidence prompts children to explore and learn more. Beyond surprise, children question and explore more when several explanations remain roughly equally likely. Asking allows the refinement of better internal models.

The same idea of surprise is at the core of Claude Shannon’s Information Theory: the more surprising an event, the more information it carries. This idea is used in many places in computer systems: data compression spends more bits on surprising patterns (as surprising events are, by definition, less frequent), being able to accurately predict things leads to better data compression, and flash drives become reliable by deploying error correction as a defense against being surprised by incorrect data. Notice that these examples are not just about the surprise itself, but about what the system does in response to it: all have models of the “world” and try to plan for eventualities in it. How closely the world model matches reality determines the technique's success.

Social machines

Adaptive modulation is a wireless technique in which the transmitter adjusts the number of bits per symbol based on the channel quality. If the channel is good, cram in more bits; if it’s bad, back off to fewer bits so the data survives. A channel can be bad for a million reasons; among these are too many devices trying to send data at the same time, the signal being reflected off buildings and arriving at the receiver multiple times, or the receiver not sensing the signal and mistaking it for background noise…

Analogous to the problems the party faced in the 1980s. There was (more or less) a single official communication channel, and they had to send multiple messages to different recipients simultaneously through it. As this was primarily one-way communication, they had to do their best to anticipate the impact on different constituencies and encode the information so that everyone would receive only the intended spin. They had some ability to error-correct by adapting future messages, but these corrections could also convey unintended messages, so they had to be used sparingly. The receivers, on the other hand, have an asymmetric advantage: they can discuss their interpretations, compare notes, and predict the party's future behavior. All that was needed was curiosity on the receivers’ end, an attempt at a plausible answer, which, even if not completely correct, could improve their models of the world. Being less wrong is better; being completely right is not actually the goal; it may be impossible.

Wireless communication protocols implement a very similar process when sending and receiving bits of information over a communications channel. These protocols aim to create the illusion that only a pair of devices is communicating, even though they might be communicating over a channel shared with a million other devices. If you were to break the system into building blocks, you would find similar abstractions for hypothesis generation, verification, prediction, and out-of-band communication. In fact, all of you using WiFi or mobile phones use systems that implement a digital version of the process that plays out at a higher level in society.

Information-poor environments

In Bayesian learning, priors encode initial beliefs about parameters before observing data. Priors are combined with the likelihood via Bayes' theorem to produce the posterior distribution, which represents updated beliefs after seeing data. The strength of the prior's influence diminishes as more data is observed: weak priors are quickly overwhelmed by data. This is great news: bad initial predictions can be corrected quickly given new, correct data!

The 1980s, especially behind the Iron Curtain, were an information-poor era. This meant we had ample time to analyze the information we received, develop our own theories and hypotheses, and fill in the gaps in our knowledge. This was true both for politics and for technologies. It also meant that priors, often just an incorrect starting point, signified curiosity about a subject rather than a fixed, unmovable belief. Arguing with someone was about creating new and better models of the world, rather than debating articles of faith.

Mental models are never perfect. There is always an opportunity to improve them by identifying inconsistencies and examining models from different perspectives. In fact, having multiple, conflicting mental models is a surefire way to speed up their refinement. An information-poor environment is actually great for that because it means the individual can focus on fewer things and filter out less noise.

Not a walk down memory lane

I am not suggesting that you should move to Cuba to experience something closer to 1980s Hungary to become a better thinker. I’m not discouraging it either, but my main point is different: the route to knowledge is through forming narratives, second-guessing perspectives, and refining your own. I chose my examples to illustrate that ideas in one realm have parallels in others, and you can and should learn across them. Don’t expect knowledge to be handed to you on a silver plate; you can only make it your own if you become a five-year-old child again, who keeps on connecting the dots by incessantly asking why. In an information-rich environment, most information is deliberately crafted to propagate, which means distracting you from something else, so you have to become good at filtering. Attention is all you’ve got.

Coda on creativity

The best scientists and engineers are all very creative. Usually, one wouldn’t expect this, as they don’t tend to apply creativity to their wardrobes or haircuts. However, when faced with a daunting number of constraints on engineering tasks, they find ways to resolve them and still deliver products on time. They find plausible narratives that map onto a problem and decide what to do based on a “creative hunch,” a mental muscle they’ve chiseled over their lives.

Nature’s way of engineering is through evolution, an excellent system if time is not of the essence. However, humans tend to think otherwise, and creativity ends up being a helpful shortcut to problem-solving. Instead of trying all possible potential solutions to a problem, creativity prioritizes experiments in certain smaller areas. The most creative solutions often break the underlying assumptions they started with, making the results even better.

John Cleese has talked about creativity for decades and published a book titled *Creativity: A Short and Cheerful Guide *in 2020. It’s not surprising that he has a view, as he is one of the comedic geniuses in Monty Python, and also needed to top up his bank account after several expensive divorces. An entertaining story for getting corporate gigs, I’m sure, helped with that. However, what he says resonates and is worth trying.

He distinguishes between two mental states. Closed mode: the anxious, purposeful, slightly stressed state we're in most of the time when we're trying to get things done. Good for execution, bad for original thought. Open mode: relaxed, playful, curious, less fixated on a goal. This is the state in which ideas actually show up, and connections between them are made.

Cleese gives a recipe to slip into the open mode on purpose:

• Space: Seal yourself off from everyday demands and interruptions. Create a quiet "oasis" where you won't be disturbed.
• Time: Give yourself a clear start and finish (1.5 hours). A fixed window helps keep the outside world out; too short, and ideas can't develop; too long, and focus wanders. Cleese says it takes him about 30 minutes to calm down, leaving an hour of real creative play.
• Patience: Spend time sitting with a problem rather than grabbing the first solution. He argued that the most creative people tolerate the discomfort of not resolving things quickly and let the problem stew. When you sit down, your mind throws up trivial to-do-list thoughts; let them pass.
• Confidence: Don't pre-judge your ideas. You can't be playful if you're afraid of making a mistake: to play is to experiment, and any drivel may lead to a breakthrough.
• Humor moves us from the closed mode into the open mode faster than almost anything, because play and laughter loosen us up.

A few other ideas he returns to often:

• Trust your unconscious, give it material and time ("sleep on it”), and it'll keep working on problems while you're not looking.
• Separate creating from editing, because generating freely and criticizing at the same time strangles ideas.
• Capture ideas as they come, since they're slippery and easy to lose.
• And guard against interruptions, which he treats as the natural enemy of the open mode.

July 3, 2026