How to Use This Book
This book is designed for self-directed learning. You do not need a teacher or a classroom. What you do need is intellectual honesty, a willingness to question your own ideas, and time to reflect.
Book Structure: Each section moves from the foundational to the applied. First you will understand the theoretical concepts, then you will see how they appear in real life, and finally you will practice applying them through concrete exercises.
Prerequisite: This book assumes you have completed all of Module I and Topic 1 of Module II (Epistemology). The concepts of types of knowledge, falsifiability, paradigms, epistemic assumptions, System 1/2, cognitive biases, and motivation are referenced without being explained again.
Special Note: This topic addresses the most frequent and most costly error in problem formulation: confusing what you actually want to achieve with the instrument you chose to achieve it. If epistemology taught you to examine the quality of your knowledge, this topic will teach you to examine the structure of your intention.
Recommendation: Read each full section before moving on to the exercises. After completing the exercises, reread the section. The second reading always reveals what the first one did not reach.
Learning Objectives
By completing this book, you will be able to:
1. Distinguish between ends (what you actually want to achieve) and means (the instruments you chose to achieve it), and recognize when a formulation conflates the two.
2. Apply the means-ends ladder technique to uncover the real intention behind any request.
3. Identify framing errors and understand how the way a problem is presented determines which solutions can even be conceived.
4. Apply Pólya's method to break down complex problems and verify that they are correctly formulated before solving them.
5. Recognize wicked problems as defined by Rittel and Webber, and understand why they resist any definitive formulation.
6. Apply Schön's reflective practice to detect and correct formulation errors in real time.
1. Introduction: The World's Most Expensive Confusion
If you had to choose a single error that explains the majority of badly formulated problems, it would be this one: confusing ends with means. People say what they want to do when they should be saying what they want to achieve. They ask for tools when they should be describing outcomes. They formulate solutions when they should be formulating problems.
“A client calls an architect and says: ‘I want a glass wall in the living room.’ The architect could simply design the wall. But if he asks ‘why a glass wall?’, he may discover that the client wants more natural light. And if there are other sources of natural light available, perhaps the solution is a skylight, not a glass wall. And if he asks further, perhaps the client does not want more light at all, but rather wants the living room to feel more spacious—which could be achieved with mirrors, lighter colors, or a rearrangement of the furniture. The glass wall was a means. The real end was three levels further up.”
This confusion is so ubiquitous that it has a name in requirements engineering: the premature solution effect. It occurs when the person presenting the problem has already chosen a solution and presents that solution as though it were the problem. It is not an error of logic; it is an error of formulation that wastes resources, forecloses superior alternatives, and produces results that satisfy the request but not the need.
“Theodore Levitt, a professor at Harvard Business School, captured it in a phrase that has become famous: ‘People don't want a quarter-inch drill. They want a quarter-inch hole.’ But even this falls short. People do not want a hole; they want to hang a picture. And perhaps they do not want to hang a picture; they want their living room to look welcoming. Each level upward opens a radically wider space of solutions.”
In the context of interacting with AI systems, the ends-means confusion is endemic. Users formulate requests that are means disguised as ends, and the systems, obediently, execute the means without ever exploring the end. The result is efficiency in the service of irrelevance: technically correct answers to fundamentally wrong questions.
1.1 Why Do We Confuse Ends and Means?
The ends-means confusion is not a random error. It is a predictable consequence of the cognitive mechanisms we studied in Module I.
System 1 generates solutions, not problems. When you face a situation, your System 1 does not meticulously formulate the problem and then search for solutions. It instantly generates the first solution that comes to mind—based on prior patterns, availability, and representativeness—and presents it as “what you need.” The formulation of the problem is skipped, because System 1 has no patience for it.
“An entrepreneur reads about the success of a company that implemented a customer service chatbot. His System 1 fires: ‘I need a chatbot.’ That is the solution his associative memory retrieved instantly. The real problem—which might be low customer satisfaction, long wait times, high operating costs, or something else entirely—was never explicitly articulated.”
Working memory cannot hold both levels. Simultaneously formulating the problem (the end) and the solution (the means), while evaluating the relationship between them, requires more working memory capacity than most people have available at the moment of the request. So the brain collapses the two levels into one: the solution becomes the problem.
Anchoring bias fixes the first means in place. Once the user has thought of a specific means (“I need a chatbot”), that means functions as an anchor. Every subsequent formulation gravitates around that anchor. The space of alternative solutions contracts without the user noticing.
Motivation closes the search prematurely. As we saw in Topic 4 of Module I, anticipatory dopamine fires when the user imagines the solution. That emotional charge makes the means feel right even before it has been evaluated against the real problem. Abandoning the means and returning to the end would mean giving up that anticipated satisfaction—something the reward system resists.
2. The Means-Ends Ladder
The means-ends ladder is an elicitation tool that lets you navigate vertically between levels of intention. It consists of asking questions that climb (“why?”) and descend (“how?”) through the hierarchy of objectives until you find the right level of formulation.
2.1 Climbing the Ladder: “Why?”
Each time you ask “why do you want that?”, you climb one rung of the ladder. The current means becomes the instrument of a higher end. This upward movement reveals the real intention hiding behind the surface request.
“Request: ‘I need a new website.’ Why? ‘So that clients can find us online.’ Why? ‘To generate more sales leads.’ Why? ‘To increase revenue by 20% this year.’ Now the picture is different. The real end is increasing revenue. A website is one possible means, but not the only one: optimizing the existing sales funnel, implementing a referral program, improving retention among current clients—all are viable alternatives that were invisible when the conversation began at the level of the website.”
But there is a limit: if you climb too high, you arrive at ends so abstract that they lose all operational usefulness. “I want to be happy” is a legitimate end, but it is not actionable. The art of formulation lies in finding the optimal level: high enough not to be trapped inside a specific means, but concrete enough to be actionable.
2.2 Descending the Ladder: “How?”
The inverse question—“how would you achieve that?”—descends the ladder. It converts an abstract end into concrete means. This downward movement is essential for implementation, but only after you have verified that you are descending from the correct end.
“End: ‘I want my clients to trust my company more.’ How? ‘By improving the transparency of our processes.’ How? ‘By publishing a quarterly results report and a behind-the-scenes blog.’ How, concretely? ‘I need a report template, an editorial calendar, and someone to write the posts.’ Now the means are in the service of the correct end. If we had started with ‘I need a blog’ without climbing first, we might have created a blog that generates no trust—only empty content.”
2.3 The Optimal Level of Formulation
The optimal level of formulation for a problem is the one where the end is specific enough to be evaluable but broad enough to admit multiple means. It is the level where Popper's falsifiability (Topic 1) meets creativity: you can define success and failure, but the solution space is not yet closed.
“Too low: ‘I need a Facebook Ads campaign.’ (Fixed means, not evaluable as an end.) Too high: ‘I want my company to grow.’ (Not falsifiable, not actionable.) Optimal level: ‘I need to generate 200 qualified leads per month at a cost per lead below $15, targeting HR directors at companies with 50–200 employees.’ Now the end is specific, measurable, and falsifiable, but the means remain open: it could be Facebook Ads, LinkedIn, email, events, content, or some combination.”
3. Framing Errors: The Problem of the Problem
If the ends-means confusion is the most frequent error of content, the framing error is the most dangerous error of structure. A framing error occurs when a problem is formulated in a way that predetermines—and limits—the possible solutions. It is not that the answer is bad; it is that the question is badly posed.
3.1 What Is a Frame?
A frame is the perspective, the vocabulary, and the assumptions that shape how a problem is presented. Changing the frame does not change the facts; it changes which facts are considered relevant, which relationships are assumed, and which solutions can be imagined.
“The same phenomenon—sales dropped 15%—can be framed in radically different ways. Cost frame: ‘We need to cut expenses to compensate for the decline.’ Market frame: ‘We need to understand what changed in our clients' needs.’ Competitive frame: ‘We need to analyze what our competitors are doing that we are not.’ Product frame: ‘Our product no longer solves the problem it used to solve.’ Each frame generates a different set of questions, relevant data, and conceivable solutions. And crucially, each frame makes the solutions of the other frames invisible.”
Kahneman and Tversky, whose work on biases you studied in Module I, demonstrated experimentally that frames affect decisions in a deep and predictable way. Their most famous example involves a hypothetical epidemic.
“A group is presented with the following situation: 600 people are at risk of dying. They are offered two programs. Frame A (gains): ‘Program 1 saves 200 people with certainty. Program 2 has a one-third probability of saving all 600 and a two-thirds probability of saving no one.’ Most people choose Program 1. Frame B (losses): ‘Program 1 results in the death of 400 people with certainty. Program 2 has a one-third probability that nobody dies and a two-thirds probability that all 600 die.’ Now most people choose Program 2. The programs are identical in both frames. Only the presentation changed: one speaks of lives saved, the other of deaths.”
3.2 Types of Framing Errors
The frame that is too narrow: The problem is formulated in a way that excludes most viable solutions. “How do we reduce the load time of our web page?” excludes the possibility that the real problem is that the page should not exist at all—perhaps users need an app, not a website.
The frame built from the symptom: The problem is formulated from an observable symptom instead of the underlying cause. “Employees arrive late” is a symptom. The problem might be a bad workplace culture, deficient public transport, schedules that do not fit real life, or simply that the work does not matter enough to them to arrive on time. Each cause demands a completely different solution.
“A manager asks an AI: ‘Design a system of penalties for employees who arrive late.’ The frame is punitive and is built from the symptom. A reframing might be: ‘What is preventing our employees from arriving on time, and how can we remove those barriers?’ The second frame admits solutions ranging from flexible schedules to corporate transport, including the possibility that remote work is the real solution.”
The frame built from the preferred solution: The problem is formulated so that the solution the user already wants appears to be the only logical one. This is confirmation bias operating at the level of formulation. “Why should we adopt AI in our processes?” already presupposes that AI is the answer; the question is only looking for justification.
The frame built from the dominant paradigm: The problem is formulated using the vocabulary and assumptions of the prevailing paradigm, rendering invisible any solution that would require a different paradigm. This connects directly with what you studied about Kuhn in Topic 1 of this module.
3.3 Reframing as a Tool
If framing errors are so dangerous, then the capacity to reframe—to take a problem and present it from a different perspective—is one of the most valuable skills in Intention Architecture.
Reframing does not change reality. It changes which aspects of reality are made visible. A good reframing produces that moment of clarity in which the user says “I hadn't thought of that”—not because they lacked intelligence, but because their original frame was concealing dimensions of the problem.
“Original request: ‘How do I convince my boss to give me a raise?’ Reframing 1 (from the boss's side): ‘What would your boss need to see for a raise to look like a good investment?’ Reframing 2 (from the system's side): ‘Is the problem your salary, or the total value of your compensation, which includes flexibility, professional development, and responsibilities?’ Reframing 3 (from the meta-problem): ‘Do you really want a raise here, or are you using salary as a proxy for feeling valued?’ Each reframing opens different solutions that the original frame had closed off.”
4. Pólya's Method: Understand Before You Solve
George Pólya (1887–1985), a Hungarian-American mathematician, published in 1945 a book that transcended mathematics to become a universal guide to problem solving: How to Solve It. His four-phase method is, in essence, a protocol against haste—and haste, as we have seen, is the source of most formulation errors.
4.1 The Four Phases
Phase 1 — Understand the problem: Before attempting to solve anything, make sure you understand what is being asked. What is the unknown? What are the data? What are the conditions? Can you restate the problem in your own words? Can you draw a figure, make a diagram, identify the parts of the problem?
“Pólya used to say that many of his students launched into manipulating equations without having stopped to ask: what are we actually looking for? In business contexts it is identical. A development team can spend weeks building a feature without anyone having explicitly formulated: what is the user problem this feature solves? What information do we have? What constraints exist? Can we say in one sentence what we are trying to achieve?”
Phase 2 — Devise a plan: Once you have understood the problem, look for the connection between the data and the unknown. Have you seen a similar problem before? Can you use analogies? Can you decompose the problem into smaller parts? Can you solve a special case first?
“If you have never exported a product to Japan, have you exported to another country? Has anyone in your industry exported to Japan? Can you decompose ‘exporting to Japan’ into sub-problems—regulation, logistics, cultural adaptation, distribution channels—and tackle each one separately? Pólya teaches that creativity is not magic; it is the capacity to connect a new problem with problems previously solved.”
Phase 3 — Carry out the plan: Implement the plan, verifying each step. Is each step correct? Can you prove it?
Phase 4 — Look back: Review the result. Can you verify the solution? Can you reach the result by another route? Can you use the result or the method for another problem? This phase is the most ignored and the most valuable: it is where learning consolidates and where errors that execution did not reveal are detected.
4.2 Pólya and Intention Architecture
Pólya's contribution to Intention Architecture is concentrated in Phase 1: understand before you solve. In interactions with AI systems, Phase 1 is exactly what users skip. They formulate requests that are Phase 3 directly—“do this”—without having passed through understanding the problem or devising a plan.
“A user writes: ‘Create a database to manage my store's inventory.’ This is Phase 3 without Phase 1. A system with intention architecture would apply Pólya: ‘Before creating the database, let's make sure I understand your problem. How many products do you handle? How do you manage inventory today? What is the specific problem—are you losing products, do you not know when to restock, is there a discrepancy between physical and recorded stock? Have you considered that a well-designed spreadsheet might solve this without needing a full database?’”
Phase 4—looking back—is equally critical and normally absent. After the system produces an answer, does the user evaluate it against the original problem? Or do they simply accept it because “it looks good” (the affect heuristic)? A well-designed system would include a verification step: “Does this solve the problem you had, or do we need to adjust?”
5. Reflective Practice: The Professional Who Thinks While Acting
Donald Schön (1930–1997), a professor at MIT, introduced a concept that changed how we understand professional competence: reflective practice. His book The Reflective Practitioner (1983) argues that competent professionals do not apply theories to problems; they construct the problem while acting, reflecting within the action itself.
5.1 Reflection-in-Action vs. Reflection-on-Action
Schön distinguishes between two kinds of reflection. Reflection-in-action occurs during the activity itself: the professional thinks, adjusts, and reformulates while working. Reflection-on-action occurs afterward, when they review what was done in order to learn from it.
“A doctor who, while examining a patient, notices that something does not fit the initial diagnosis and adjusts her line of questioning on the fly is practicing reflection-in-action. She does not stop acting in order to think; she thinks while acting. After the consultation, when she reviews the case and asks herself ‘why was my first diagnosis wrong, and what signals should I have caught earlier?’, she is practicing reflection-on-action.”
5.2 Problem Setting vs. Problem Solving
Schön's most relevant contribution to Intention Architecture is his distinction between problem solving and problem setting. Most professional training teaches you to solve problems that are already formulated. But in real practice, the hardest part is not solving the problem—it is deciding which problem it is.
“A consultant who arrives at a company with problems of ‘low productivity’ does not have a formulated problem; he has a messy situation. Is the problem the low productivity, or the poor leadership that causes it? Is it the leadership, or the badly designed incentives that permit it? Is it the incentives, or the organizational culture that normalizes them? Problem setting—deciding which problem to solve—matters more than problem solving, because it determines the entire direction of the work that follows.”
Schön argues that the situations of real practice are “swampy lowlands”: messy, ambiguous, loaded with contradictory values, and lacking any clear formulation. The competent professional does not wait for someone to hand them a clean problem; they construct the problem out of the mud of the situation. And that construction—that problem setting—is a creative and epistemic act that requires all the tools we have studied: examining assumptions, questioning frames, distinguishing ends from means.
5.3 Implications for Intention Architecture
Schön's reflective practice has profound implications for how we design interactions with intelligent systems. If the user is standing in a swampy lowland and needs to set the problem before solving it, then a system that simply executes requests is operating on the wrong model. The system should function as a reflective colleague: someone who helps construct the problem, not merely solve it.
“A user writes: ‘I have an unmotivated team. Give me five strategies to motivate them.’ A reflective system would not deliver the five strategies immediately. First it would practice problem setting: ‘How do you know they are unmotivated—what specific behaviors are you observing? Is it the whole team or particular individuals? Is this recent, or has it always been this way? Did something change in the environment that coincides with the decline?’ Each question helps the user construct the real problem before jumping to solutions.”
6. Wicked Problems: When the Formulation Is the Problem
Horst Rittel and Melvin Webber, professors of design and public policy at Berkeley, published an article in 1973 that introduced a concept challenging everything we have studied so far: wicked problems. These are problems that, by their very nature, resist any definitive formulation.
6.1 The Characteristics of a Wicked Problem
Rittel and Webber identified ten characteristics of wicked problems. The most relevant for Intention Architecture are:
There is no definitive formulation. A wicked problem cannot be fully defined before it is solved. Each attempt to formulate it reveals new aspects that change the formulation. Problem and solution co-develop.
“‘How do we reduce poverty?’ sounds like a clear question, but each attempt to formulate it more precisely opens new dimensions: poverty of income or of opportunity? Absolute or relative? In the short term or structurally? In which cultural context? Each definition implies different assumptions, different values, and different solutions. The problem is genuinely indeterminate.”
Every solution is a one-shot operation. There is no trial and error: each attempted solution changes the situation irreversibly. You cannot undo a public policy, a launched product, or a corporate reorganization and return to the starting point.
There is no criterion of “right” or “wrong.” Solutions to wicked problems are not true or false but better or worse, and the evaluation depends on the values of whoever is judging. What is a brilliant solution for one group is a catastrophe for another.
Every problem is essentially unique. However similar they may look on the surface, wicked problems are not replicable. What worked in one context can fail in another that appears identical.
6.2 Wicked Problems in Everyday Practice
Although Rittel and Webber wrote about urban planning and public policy, wicked problems abound in everyday professional and personal life. Any problem that involves people, values, uncertainty, and irreversible consequences has elements of “wickedness.”
“‘How should our company approach artificial intelligence?’ is a wicked problem. It has no definitive formulation (it depends on the industry, the team, the competition, the regulation). Every action is irreversible (the talent you hire, the technology you adopt, the culture you create). There is no universal ‘right.’ And it is essentially unique: what works for one company will not work for another, however similar they appear.”
6.3 Implications for Intention Architecture
Wicked problems challenge the premise that every problem can be correctly formulated if enough rigor is applied. Sometimes the “correct” formulation does not exist—there are only formulations that are more or less useful, more or less transparent about their limitations. For Intention Architecture, this means:
Accept indeterminacy. Not everything can be resolved with a better formulation. Some problems require action under ambiguity, and the best contribution a system can make is to make that ambiguity explicit rather than hide it behind an answer that projects false certainty.
Make the conflicting values visible. When a problem is wicked, different formulations are not objectively “better” or “worse”; they reflect different values. The system must make these values visible so that the user chooses consciously, not by default.
Iterate with humility. In wicked problems, the plan is not “formulate correctly and then execute.” It is “formulate provisionally, act, observe, reformulate, repeat.” Pólya's Phase 4—looking back—becomes the principal phase.
7. Practical Application: Ends, Means, and Frames in Action
Let us bring together the concepts of this topic in an integrated scenario that shows how they all operate simultaneously.
7.1 Case Study: The Redesign Nobody Asked For
Santiago, a product director at an e-learning platform, asks an AI system: “I need to redesign our platform's interface. Users complain that it's hard to use. I want something modern, like Notion or Linear.”
Let us analyze the request through each lens of this topic:
Ends-means confusion: Santiago presents a means (redesigning the interface) as though it were the end. If we apply the ladder: why the redesign? So that it is easier to use. Why? So that users complete the courses. Why? So that the platform generates retention and revenue. The real end is retention, not a redesign. And perhaps retention does not depend on the interface at all, but on the content, the notification system, or the user's perception of progress.
Framing error: The problem is framed as a visual design problem (“something modern, like Notion”). But “hard to use” could be a problem of information architecture, of workflow, of onboarding, or of misaligned expectations. The visual frame closes off all of these alternatives.
Pólya — Phase 1 absent: Santiago has not completed his understanding of the problem. What does “hard to use” mean specifically? Where do users get stuck? How many complain versus how many simply leave in silence? Is there usage data confirming the problem, or only anecdotes from support?
Schön — Problem setting absent: Santiago arrived with the problem already formulated: “the interface is bad.” He did not construct the problem out of the situation; he adopted the most obvious formulation (user complaints) without examining whether it is the most productive one.
Rittel — Wicked elements: What makes an educational interface “good”? It depends on who uses the platform, for what kind of content, in what context. What is “modern” for a 25-year-old user in tech can be intimidating for a 55-year-old professional learning a new trade. The problem has wicked elements because it involves diverse users with conflicting values and expectations.
“A system with good intention architecture would respond: ‘Before redesigning, I want to make sure we're solving the right problem. You tell me users complain about difficulty. Do you have usage data showing where specifically they get stuck? Do the complaints come from all user segments or from one particular group? I also notice you mention Notion and Linear as references—those are productivity tools, not learning tools. Are we sure that a productivity aesthetic is what an educational platform needs? Perhaps the problem is not how the platform looks but how it feels to use it for learning. Can we explore that distinction before committing resources to a redesign?’”
7.2 Derived Design Principles
1. Separate the end from the means before executing. Faced with any request, ask “why?” at least once. If the answer is another means, keep climbing.
2. Offer explicit reframings. When the user's frame appears narrow, symptom-based, or solution-driven, present at least one alternative perspective.
3. Apply Pólya's Phase 1 before Phase 3. Understand the problem before solving it. Verify the data, the constraints, and the criteria for success.
4. Practice problem setting, not just problem solving. When the situation is ambiguous, help the user construct the problem instead of accepting the first formulation.
5. Detect wicked problems and be transparent. If the problem has characteristics of “wickedness,” do not pretend it has a clean solution. Make the conflicting values visible and accompany the iteration.
6. Activate Phase 4 periodically. After producing an answer, prompt verification: “Does this solve your real problem, or do we need to reformulate?”
8. Glossary of Key Terms
| Term | Definition |
| Frame | The perspective, vocabulary, and assumptions that shape how a problem is presented, determining which facts are considered relevant and which solutions are conceivable. |
| Means-ends ladder | An elicitation technique that navigates vertically between levels of intention: climbing with “why?” (from means to end) and descending with “how?” (from end to means). |
| Framing error | A formulation of a problem that predetermines and limits the conceivable solutions by presenting one perspective as though it were the only possible one. |
| End | The result one seeks to achieve; the real objective behind the request. What is genuinely wanted. |
| Means | The instrument, tool, or action chosen to reach an end. There can be multiple means for the same end. |
| Problem setting | In Schön, the act of constructing and defining the problem out of an ambiguous situation, prior to any attempt at a solution. |
| Wicked problem | A problem that resists definitive formulation, lacks a universal criterion of solution, involves conflicting values, and whose consequences are irreversible (Rittel and Webber). |
| Reflective practice | In Schön, the capacity of the professional to reflect during the action itself, adjusting their understanding and formulation of the problem in real time. |
| Reframing | The technique of presenting a problem from a different perspective, revealing dimensions and solutions invisible from the original frame. |
| Premature solution | The phenomenon by which the person presenting a problem has already chosen the solution and presents it as though it were the problem itself. |
9. Final Reflection and Self-Assessment
Before moving on to the next topic (Applied Logic and Systems Thinking), verify that you can apply these concepts.
10. Bibliography and Recommended Readings
Essential Readings
• Pólya, G. (1945). How to Solve It. Princeton University Press.
• Schön, D. (1983). The Reflective Practitioner. Basic Books.
• Rittel, H. & Webber, M. (1973). “Dilemmas in a General Theory of Planning.” Policy Sciences, 4(2), 155–169.
Complementary Readings
• Dorst, K. (2015). Frame Innovation: Create New Thinking by Design. MIT Press.
• Tversky, A. & Kahneman, D. (1981). “The Framing of Decisions and the Psychology of Choice.” Science, 211(4481), 453–458.
• Levitt, T. (1960). “Marketing Myopia.” Harvard Business Review, 38(4), 45–56.
• Dörner, D. (1996). The Logic of Failure. Basic Books.
Next Topic
The next book in this series addresses Applied Logic and the Scientific Method: how to move from a correct formulation of the problem to valid reasoning about its solutions. You will study the rules of logical inference, the most common fallacies in everyday reasoning, and how the scientific method provides a disciplined framework for advancing from problem to evidence. You will connect epistemology and problem formulation with the rigor of reasoning that separates legitimate conclusions from those that merely look like it.