Cognitive Psychology

Learning Objectives

Upon completing this book, you will be able to:

1. Explain how the human mind processes information and how this shapes the intentions we express.

2. Distinguish between automatic processing (System 1) and deliberate processing (System 2), and recognize when each dominates your decisions.

3. Identify the limitations of working memory and design communications that respect those limitations.

4. Recognize how selective attention filters out critical information and how to compensate for those omissions.

5. Analyze your own and others' mental models to detect distortions, gaps, and unexamined assumptions.

6. Apply these principles to the design of interactions with intelligent systems, formulating clearer and more complete intentions.

1. Introduction: What is Cognitive Psychology?

Cognitive psychology is the scientific study of internal mental processes: how we perceive the world, how we think, how we remember, and how we make decisions. Unlike behaviorism, which for decades limited itself to studying only observable behavior, cognitive psychology treats the mind as an information processing system that receives inputs from the environment, encodes them through attention and perception, stores them in memory, and retrieves them to guide behavior.

“Think of something as simple as looking for your keys. Your brain receives visual information from the environment (perception), filters out irrelevant objects (attention), checks where you last left them (memory), and decides where to look first (decision). All this happens in seconds, without you noticing the process. Cognitive psychology studies exactly that invisible process.”

This discipline is fundamental to the Architecture of Intention because every intention begins as a cognitive event: a goal is represented, alternatives are evaluated, and a course of action is selected. If you don't understand how this internal process works, you can't design systems that correctly interpret what people really want.

1.1 Historical Context

To understand cognitive psychology, it's helpful to know what it reacted against. During the first half of the 20th century, behaviorism dominated academic psychology. Figures like John Watson and B.F. Skinner argued that science could only study what was observable: stimuli and responses. The mind was a "black box" that wasn't worth opening.

“Imagine a dog salivating when it hears a bell. For the behaviorist, that's all that matters: bell → saliva. They're not interested in what happens inside the dog's brain. For the cognitive psychologist, what's interesting is exactly what happens inside: how did it associate the sound with food? What internal representation was formed? Why a bell and not any other sound?”

The cognitive revolution of the 1950s and 1960s changed this. Three developments converged: Claude Shannon's information theory showed that information could be quantified and processed; Noam Chomsky's generative linguistics demonstrated that language had an internal structure that behaviorism couldn't explain; and the first computers offered a powerful metaphor for understanding the mind: the mind as an information processor.

“Chomsky demonstrated something simple but devastating: a three-year-old can construct sentences they've never heard before, like 'the green cat flew to the moon.' If language were just imitation of stimuli, as behaviorism said, this would be impossible. The child must have internal rules—a mental grammar—that allow them to generate new sentences. The black box did have something inside it.”

This computational metaphor has important limitations that we'll explore later, but its contribution was decisive: it legitimized the scientific study of internal processes and provided a vocabulary to describe them: encoding, storage, retrieval, parallel processing, attentional bottlenecks.

“It's as if before the cognitive revolution, psychologists studied an airport by only looking at it from the outside: they see planes arriving and departing, but refuse to enter the control tower. Cognitive psychology decided to enter the tower and understand how flights are coordinated.”

2. Dual Process Theory: System 1 and System 2

Daniel Kahneman, Nobel Prize in Economics in 2002, articulated in his book Thinking, Fast and Slow the most influential distinction in contemporary decision psychology: the dual process theory. This theory proposes that human thought operates through two systems with radically different characteristics.

“Imagine you're walking down the street and you hear a loud horn. Instantly, you jump towards the sidewalk. You didn't think, you didn't evaluate options, you didn't calculate the speed of the vehicle. That was System 1. Now imagine you're choosing a health insurance plan by comparing three plans with different deductibles, copays, and coverage. That requires System 2. Same brain, two radically different ways of operating.”

2.1 System 1: Fast, Automatic, Intuitive

System 1 operates automatically, quickly, with no apparent effort, and without a sense of voluntary control. It's the system that allows you to read these words without conscious effort, that makes you turn your head when you hear your name, that generates an instant feeling of dislike when you see a disturbing image.

Basic characteristics of System 1:

Speed: It operates in milliseconds. Emotional evaluation of a situation occurs before you have time to think consciously about it.

“When you see someone's face for the first time, in less than a second you already have an impression of whether that person seems trustworthy or not. You didn't choose to evaluate; it simply happened.”

Automaticity: You can't choose not to activate it. If you know how to read Spanish, you can't look at these letters without reading them. You can't look at a face without evaluating its emotional expression.

“Try to look at the word RED written in blue without experiencing a conflict. You can't avoid reading the word, even if you just want to see the color. That's automaticity in action—the famous Stroop effect.”

Associativity: It works by association and pattern recognition. When you see a situation, System 1 automatically searches your past experience for something similar and generates a response based on that pattern.

“An experienced doctor who looks at a patient entering the room and 'feels' that something is wrong before doing any examination is using pattern recognition. Thousands of previous cases feed that instant intuition.”

Narrative coherence: System 1 builds the best possible story with the available information. It doesn't evaluate the quality of the evidence; it works with what it has. Kahneman calls it WYSIATI: What You See Is All There Is (what you see is all there is).

“If you read that 'John is intelligent, hardworking, and envious,' you immediately form a coherent impression: someone capable but with a flaw. If the same words appear in reverse order—'envious, hardworking, intelligent'—the impression changes. System 1 gives more weight to what you read first and builds the story from there.”

2.2 System 2: Slow, Deliberate, Analytical

System 2 is in charge of conscious, deliberate, and effortful thinking. It's the system you activate when you calculate 17 × 24, when you look for a specific person in a crowd, when you fill out a tax form, or when you consciously compare two complex options.

Basic characteristics of System 2:

Relative slowness: It requires time. You can't accelerate deliberate thinking without sacrificing precision.

“Try to solve 347 ÷ 13 mentally. You feel how the process is slow, sequential, and demands that you keep numbers in your head. If someone interrupts you, you lose the thread. That fragility is the footprint of System 2 in action.”

Conscious effort: It consumes limited cognitive resources. After a prolonged period of intense concentration, you feel mentally exhausted. This phenomenon is called ego depletion or decision fatigue.

“That's why judges grant more conditional freedoms after lunch than just before. It's not cruelty; it's decision fatigue. After hours of making complex decisions, the brain seeks the option that requires less effort—which usually is the status quo.”

Sequentiality: Unlike System 1, which processes multiple inputs in parallel, System 2 operates serially: it can only attend to one demanding task at a time.

“That's why you can't maintain a complex phone conversation while making a left turn in a busy intersection. Both tasks compete for the same limited resource. One of them will degrade.”

Sense of agency: When System 2 is active, you have the subjective experience of choosing, reasoning, deciding.

2.3 The Interaction Between Both Systems

The relationship between the two systems is not one of competition but of division of labor. System 1 continuously generates impressions, intuitions, intentions, and feelings. When everything works smoothly, System 2 adopts these suggestions with minimal or no modification. That is: most of what you think and do originates in System 1, but System 2 has the final word when it's activated.

“It's like an autopilot and a human pilot in an airplane. The autopilot (System 1) handles the flight most of the time. The human pilot (System 2) takes control only when there's turbulence, an emergency, or a complex landing. The problem is that the human pilot sometimes falls asleep when they should be awake.”

The problem is that System 2 is lazy. Activating it has a real energy cost, and the brain is optimized to conserve energy. This means that, in practice, most of our daily decisions are the product of System 1, including many that deeply affect us.

“When you choose what to have for dinner, you probably don't do a nutritional analysis. You choose what 'you feel like,' which is System 1's suggestion based on mood, habit, and what you've seen recently. And that's okay for dinner—but the same mechanism operates when you choose an investment, a supplier, or how you formulate a request to an AI system.”

3. Working Memory: The Limited Space of Consciousness

Working memory is the mental space where conscious thought takes place. It's where you actively manipulate information: where you hold a phone number while looking for a pen, where you hold the elements of an argument while building it, where you compare options before choosing.

“Imagine you're cooking a new recipe. You need to remember simultaneously the ingredients you've already added, the oven temperature, the time what's frying has been cooking, and the next step in the recipe. When someone interrupts you with a question, you feel how part of that information “falls” out of your mind. That fragile space that overflowed is your working memory.”

George Miller, in his classic 1956 article titled “The Magical Number Seven, Plus or Minus Two”, established that the capacity of short-term memory is approximately seven elements (±2). Later research has refined this figure: for ungrouped information, the real limit is closer to four elements.

“Four elements sounds like very little, and it is. Think about trying to compare five mobile phone plans that vary in price, data, calls, validity, and coverage. They're five options with five variables each: 25 pieces of data that your working memory simply can't hold at the same time. That's why you end up choosing the one that “feels right” instead of the objectively best one—the System 1 rescues the System 2 when it gets saturated.”

3.1 Implications for Formulating Intentions

This limitation has profound consequences. When a person makes a request to an AI system, they are working with limited working memory. This means that:

They can't consider all relevant variables at once. A manager who asks for a market analysis may forget to mention budget constraints, demographic segments, or deadlines, not out of negligence but due to cognitive limitation.

They tend to simplify complex problems. Faced with a problem with multiple dimensions, the mind reduces the variables to those that fit in working memory, discarding potentially critical information.

“When someone says “I need the product to be cheaper,” they're compressing a multidimensional problem (material costs, margins, value perception, competition, volume) into a single variable that their working memory can handle. The real intention might be “I need to be more competitive,” which admits very different solutions.”

The last information received dominates. The recency effect makes the most recent information have a disproportionate weight in formulating the intention.

“If a boss has just come out of a meeting where a customer complained about the service, their next request to their team will likely be dominated by that complaint—although statistically 95% of customers are satisfied. The recent complaint occupies all the working memory and biases the intention.”

3.2 Chunking: Expanding the Limits

Chunking is the cognitive strategy that allows partially overcoming the limitations of working memory. It consists of grouping individual elements into larger meaningful units. The phone number 5551234567 is difficult to retain as 10 separate digits, but easy as three groups: 555-123-4567.

Experts in any domain are experts, to a large extent, because they have highly developed chunking schemes. A chess master doesn't see 32 pieces on a board; they see strategic configurations that work as units. An experienced programmer doesn't read code line by line; they recognize complete design patterns.

“If you've never seen a musical score, you'll see hundreds of black dots distributed on lines. An experienced musician sees complete musical phrases—chords, progressions, rhythmic patterns—each of which occupies a single “space” in their working memory. The same information, but compressed by experience.”

4. Attention: The Entrance Gate to Consciousness

Attention is the mechanism that selects which information from the environment will be consciously processed and which will be ignored. It is, simultaneously, our greatest ability and our greatest limitation: it allows us to focus on what is relevant, but blinds us to everything else.

“At this moment, you are reading these words and are probably not aware of the pressure of the chair against your body, nor the ambient sound of the room, nor the temperature of your feet. All that information was there, but your attention filtered it out until this sentence brought it to the forefront. That is selective attention.”

4.1 Selective Attention and Inattentional Blindness

The most famous experiment on selective attention is the invisible gorilla, conducted by Daniel Simons and Christopher Chabris in 1999. Participants had to count the passes of the ball between players wearing white shirts while ignoring those wearing black shirts. During the video, a person dressed as a gorilla walked slowly through the center of the scene, beat their chest, and walked away. Approximately half of the participants did not see the gorilla.

“Did they not see a gorilla that was right in front of them for nine seconds? Yes. And the most revealing thing: when they were shown the video again, they could not believe they had not seen it. It is not a vision problem; it is that focused attention on counting passes literally made the gorilla invisible to their consciousness.”

This phenomenon, called inattentional blindness, demonstrates something profound: we do not see the world as it is. We see what our attention selects. And our attention is directed by our objectives, expectations, and emotional state at the moment.

4.2 Implications for the Architecture of Intention

When a user makes a request to an AI system, the information they include is determined not only by what they know, but by what they happened to notice. Critical variables may be absent from the request simply because the user's attention was focused on another aspect of the problem.

“A restaurant owner asks an AI: 'Help me improve my Google reviews.' Their attention is fixed on the negative reviews they just read. But the real problem might be the quality of service on Friday nights, when they have little staff. The reviews are the symptom that captured their attention; the cause is outside their focus.”

A robust architecture of intention must build mechanisms to bring to the surface what the user may have overlooked. This is not correcting the user; it is compensating for a universal cognitive limitation. The right question, asked at the right time, can transform a vague intention into a precise instruction.

5. Mental Models: The Internal Maps of Reality

A mental model is a simplified internal representation of how something works in the world. We don't reason from raw data; we reason from these models. When you think about how the economy works, how your boss will react to a proposal, or how an AI system will behave, you're consulting a mental model.

“If someone asks you ‘what if I raise the prices of my product?’, your instant response depends on your mental model of how the market works. Maybe you think ‘I'll sell less’ (a simple supply and demand model). But a more complete model would consider: what if the higher price generates a perception of superior quality? What if competitors also raise their prices? What if the product is a necessity with no substitutes? Your mental model determines which possibilities you even consider.”

5.1 Characteristics of Mental Models

They are necessary simplifications. The world is infinitely complex; to function in it, we need simplified versions. The map is not the territory, but without maps, we can't navigate.

“The metro map of your city doesn't reflect the real distances between stations or the depth of the tunnels. But it works perfectly for its purpose: to take you from one point to another. Mental models operate in the same way: they are false in the strict sense, but useful.”

They are implicit. Most people are not aware of the mental models they use. They operate in the cognitive background, determining which variables are considered relevant, which causal relationships are assumed, and which outcomes are anticipated.

“Many people operate with the mental model that ‘more hours worked = more productivity.’ They never articulate it explicitly, but they make decisions based on it: they stay late, distrust those who leave early, measure commitment by presence. The model is invisible but determines behavior.”

They are resistant to change. Once formed, a mental model tends to persist even in the face of contradictory evidence. This connects with the confirmation bias, which we will study in the next topic.

They vary between people. Two people looking at the same situation can have radically different mental models, resulting in different interpretations, intentions, and requests.

“When faced with the phrase ‘sales dropped 15%’, the marketing director thinks ‘we need better advertising’, the product director thinks ‘the product needs improvement’, and the financial director thinks ‘we need to cut costs.’ Same data, three different mental models, three completely different intentions.”

5.2 Mental Models and Intention

When a user formulates an intention for an AI system, they are projecting from their mental model. That model may contain gaps (unconsidered variables), distortions (incorrect causal relationships), and outdated assumptions (information that was true in the past but is no longer).

“An entrepreneur who asks an AI to ‘create a Facebook Ads campaign for my business’ may be operating with a mental model from 2018 when Facebook was the dominant channel. In 2026, their audience might be on TikTok, YouTube Shorts, or platforms that didn't even exist when they formed their model. Their real intention is ‘get customers’, but their outdated model translates it into a specific channel.”

The intention architect needs tools to diagnose users' mental models and design interactions that refine them without being paternalistic or intrusive. The clinical question is not ‘What do you want?’ but ‘What do you assume is true about the situation?’

6. WYSIATI: The Illusion of Complete Information

Kahneman coined the acronym WYSIATI (What You See Is All There Is) to describe one of the most dangerous tendencies of System 1: building the most coherent story possible with the available information, without considering what information might be missing.

“You're presented with a candidate for a position: ‘He's a graduate of a prestigious university, speaks three languages, and led a team of 20 people.’ You've probably already formed a positive impression. But they didn't tell you that he was fired from his last two jobs, that his team had the highest turnover in the company, or that the three letters of reference declined to participate. Your System 1 built a perfect story with three pieces of data—and ignored the fact that twenty were missing.”

WYSIATI explains why people can feel great confidence in judgments based on minimal evidence. If the story that System 1 builds is internally coherent, the mind accepts it as true. The quantity and quality of the evidence matter much less than the narrative coherence.

“An investor reads that a company will ‘triple its revenue next year’ based on an enthusiastic article. The story is coherent and exciting. They don't ask: Who wrote the article? Do they have a conflict of interest? What does the competition say? What's the real track record of predictions from that source? WYSIATI makes coherence feel like evidence.”

6.1 WYSIATI in Interaction with AI

This principle has direct implications for how users interact with AI systems. When a user writes a prompt, they include the information they have available in their consciousness at that moment. They don't include what they don't know, what they forgot, or what they didn't consider relevant. And, critically, they don't feel like anything is missing, because their System 1 has already built a coherent story with what they had.

“A user writes: ‘Draft an email to reject the supplier's offer.’ It seems complete. But they didn't mention that this supplier is also their only supplier of another critical component, that there's a personal relationship with the owner, or that the contract has a penalty clause. They didn't omit it deliberately; their attention was on the specific offer, and WYSIATI took care of the rest.”

An AI system that simply executes what's requested without questioning the user's premises is amplifying WYSIATI on a large scale. The architecture of intention must include mechanisms that break this illusion of completeness: checklists, probing questions, presentation of implicit assumptions.

7. Practical Application: Cognitive Psychology in Action

Everything mentioned so far is not abstract theory. It is the operational reality of every human interaction with technology, with other people, and with oneself. Let's see how it manifests in concrete scenarios.

7.1 Case Study: The Incomplete Request

An entrepreneur asks an AI system: “Write a business plan for my startup.”

Let's analyze which cognitive processes are operating:

System 1 dominant: The request was formulated quickly, probably the first formulation that came to mind. There was no deliberation about what information to include.

Saturated working memory: The entrepreneur has their business idea in mind but cannot articulate the target market, competition, finances, team, growth strategy, and legal restrictions simultaneously.

Selective attention: They are probably focused on the aspect of the business that excites them the most (the product) while ignoring less exciting but equally critical aspects (financial projections, risk analysis).

Limited mental model: Their model of what a business plan is may be based on a generic template they saw once, not on what they really need for their specific context.

WYSIATI: They don't feel that any information is missing. Their request seems complete and reasonable to them.

“Let's compare with a user who has internalized these concepts: “Write a business plan for a healthy food delivery app in Mexico City, targeting professionals aged 25–40, with an initial investment of $500K USD, focused on differentiation by certified nutritional quality. I need 3-year projections, a competition analysis with Rappi and UberEats, and a section on regulatory risks.” The difference is not intelligence; it's that the second user activated System 2 to complete what System 1 would have omitted.”

7.2 Derived Design Principles

From cognitive psychology, we can derive concrete principles for designing interactions with intelligent systems:

1. Don't assume System 2 activation. Design for users who formulate quickly and with little effort.

2. Respect working memory limits. Present information in small, meaningful groups.

3. Compensate for selective attention. Actively ask for what might be missing.

4. Diagnose mental models. Before executing, verify the user's assumptions.

5. Break the WYSIATI illusion. Explicitly present what the user didn't consider.

“Applying principle 3, a well-designed system would respond to “Write a business plan for my startup” not with a generic plan, but with: “I can help you. To make a really useful plan, I need to know: what is your target market? What is your budget? Who is your direct competition? What is your differential advantage?” Each question compensates for a likely cognitive omission.”

8. Glossary of Key Terms

9. Final Reflection and Self-Assessment

Before moving on to the next topic (Basic Neuroscience), take a moment to evaluate your understanding. It's not about memorizing definitions, but about checking that you can apply the concepts.

10. Bibliography and Recommended Readings

Essential Readings

• Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux. [Thinking Fast, Thinking Slow. Editorial Debate]

• Gigerenzer, G. (2007). Gut Feelings: The Intelligence of the Unconscious. Viking. [Instinctive Decisions. Editorial Ariel]

• Ariely, D. (2008). Predictably Irrational. HarperCollins. [The Traps of Desire. Editorial Ariel]

Complementary Readings

• Miller, G.A. (1956). “The Magical Number Seven, Plus or Minus Two.” Psychological Review, 63(2), 81–97.

• Simons, D.J. & Chabris, C.F. (1999). “Gorillas in Our Midst.” Perception, 28(9), 1059–1074.

• Sweller, J. (1988). “Cognitive Load During Problem Solving.” Cognitive Science, 12(2), 257–285.

• Baddeley, A.D. & Hitch, G. (1974). Working Memory. Psychology of Learning and Motivation, 8, 47–89.

Next Topic

The next book in this series addresses Basic Neuroscience: the biological basis of everything we have discussed here. You will understand how the brain physically produces decisions, intentions, and judgments, including Damasio's somatic marker hypothesis and Libet's experiments on the timing of conscious intention.