Chapter 22 Exercises

How to use these exercises: Work through the parts in order. Part A builds recognition skills, Part B develops analysis, Part C applies concepts to your own domain, Part D requires synthesis across multiple ideas, Part E stretches into advanced territory, and Part M provides interleaved practice that mixes skills from all levels.

For self-study, aim to complete at least Parts A and B. For a course, your instructor will assign specific sections. For the Deep Dive path, do everything.

Part A: Pattern Recognition

These exercises develop the fundamental skill of recognizing map-territory confusions across domains.

A1. For each of the following scenarios, identify (i) the map, (ii) the territory the map is supposed to represent, (iii) what the map captures accurately, and (iv) what the map distorts or omits.

a) A university ranks its departments by average research citation count. A department of three researchers -- two with moderate citation counts and one superstar -- ranks higher than a department of twenty researchers who all produce solid, well-cited work.

b) A fitness tracker records a user's daily step count. The user walks 10,000 steps on a treadmill while watching television. The tracker records this as equivalent to 10,000 steps hiking in varied terrain.

c) A credit score summarizes a person's financial history into a single number. A recent immigrant with no debt, substantial savings, and a stable job receives a low score because she has no credit history.

d) A customer satisfaction survey asks patients to rate their hospital experience on a 1-5 scale. Patients who received painful but necessary treatments rate their experience low. Patients who received pleasant but unnecessary treatments rate their experience high.

e) A standardized test measures a student's reading comprehension through multiple-choice questions. A student who reads voraciously and engages deeply with complex texts scores lower than a student who has been trained in test-taking strategies but reads little outside of school.

f) A GDP measure shows a country's economy growing after a devastating hurricane, because the rebuilding effort generates economic activity.

g) A social media platform measures user engagement through time spent on the platform. A user who is deeply absorbed in a single long-form article for 20 minutes generates the same engagement score as a user who compulsively scrolls through inflammatory content for 20 minutes.

h) A body mass index (BMI) calculation classifies a muscular athlete with 8% body fat as "overweight" because muscle weighs more than fat and BMI does not distinguish between the two.

A2. Classify each of the following as primarily operating at Level 1 (using the map consciously), Level 2 (forgetting the map is a map), or Level 3 (defending the map against the territory) of map-territory confusion.

a) A physicist uses Newtonian mechanics to calculate the trajectory of a baseball, knowing that relativistic effects are negligible at this speed.

b) A manager insists that employee performance can be fully captured by quarterly metrics, and dismisses concerns about unmeasured contributions like mentoring and team morale.

c) A doctor orders a biopsy for every incidental finding on a CT scan, saying "if it shows up on the scan, it might be dangerous."

d) A financial regulator continues to use the Gaussian copula as the standard risk model in 2006, dismissing critics who point out its treatment of tail risk, because changing the standard would disrupt the entire regulatory framework.

e) A linguist studies the Sapir-Whorf hypothesis while acknowledging that language influences but does not determine thought.

f) A political pundit insists that GDP growth proves the economy is healthy, even as wages stagnate, inequality increases, and environmental degradation accelerates.

g) A cartographer adds a note to a Mercator projection explaining that the map distorts relative sizes, especially near the poles.

h) A social scientist uses survey data to study human behavior, openly discussing the limitations of self-reported data in her methodology section.

A3. For each of the following pairs, identify which is the map and which is the territory. Be careful -- some of these are tricky.

a) A weather forecast and the weather itself.

b) A company's organizational chart and the actual pattern of influence and communication within the company.

c) A patient's medical record and the patient's actual health.

d) A country's constitution and the actual distribution of power in the country.

e) A recipe and the dish it produces.

f) A person's resume and the person's actual capabilities.

g) A photograph and the scene it depicts.

h) A musical score and a performance of the music.

A4. Alan Watts used the analogy of "confusing the menu for the meal." Generate your own analogies for the map-territory confusion using the following template: "It is like confusing the _ for the ___."

Create at least five analogies, each from a different domain (e.g., education, medicine, business, art, relationships). For each, explain briefly what the map captures and what it misses.

A5. Identify three maps you use daily -- representations, models, metrics, categories, or labels -- that you have been treating at Level 2 (forgetting they are maps). For each:

a) What is the map? b) What territory does it represent? c) What does the map capture? d) What does the map omit? e) When might the omission matter?

Part B: Analysis

These exercises require deeper analysis of map-territory patterns.

B1. The Map Audit. Choose one of the following domains and conduct a "map audit" -- an inventory of the maps (models, metrics, categories, frameworks) the domain relies on:

• Healthcare (diagnoses, test results, treatment protocols, insurance codes)
• Education (grades, test scores, degree classifications, learning objectives)
• Criminal justice (crime statistics, recidivism rates, risk assessment scores)
• Climate science (temperature records, climate models, emissions inventories)
• Social media (engagement metrics, follower counts, algorithmic feeds)
• Urban planning (zoning maps, census data, traffic models)

For your chosen domain:

a) List at least five key maps the domain uses.

b) For each map, identify what aspect of the territory it captures and what it omits.

c) Identify at least two cases where the domain operates at Level 2 or Level 3 of map-territory confusion -- where the map is being treated as the territory or defended against the territory.

d) Propose at least one change in practice that would help the domain maintain Level 1 awareness.

B2. The Copula Diagnosis. The Gaussian copula failed because it systematically underestimated the probability of extreme events. This is a common failure mode for mathematical models: they work well in normal conditions and fail catastrophically in extreme ones.

a) Identify three other models, metrics, or maps (from any domain) that work well in normal conditions but fail in extreme ones.

b) For each, explain what "extreme conditions" look like in that domain.

c) For each, explain what the model assumes about the territory that makes it fail under extreme conditions.

d) Is there a general principle here? Why do so many models fail at the extremes?

B3. The Language Map. The chapter discusses how language shapes perception through the Sapir-Whorf hypothesis.

a) Identify a technical term from your own field that non-specialists often misunderstand. Explain what the term maps (for specialists) and what non-specialists think it maps.

b) Identify a concept in your experience that you find difficult to express in words. What aspect of the territory does language fail to capture in this case?

c) If you speak or have studied a second language, identify a word or expression in that language that maps a territory your primary language does not map well. What does this untranslatable word tell you about the limitations of your primary language's map?

d) How might the language used within a profession create Level 2 map-territory confusions for practitioners? Give a specific example where professional jargon might cause experts to mistake a linguistic category for a natural kind.

B4. The Overdiagnosis Dilemma. Medical imaging illustrates a paradox: a better map can lead to worse outcomes.

a) Identify two other domains where more detailed information leads to worse decisions. (Hint: consider information overload, analysis paralysis, and premature optimization.)

b) Explain the structural mechanism that causes "more information = worse outcomes" in each case.

c) Is there an optimal level of map detail? How would you determine it? What principles guide the decision about how detailed a map should be?

d) Connect this to the concept of redundancy vs. efficiency from Chapter 17. How does the overdiagnosis problem relate to the efficiency-fragility tradeoff?

B5. Box's Dictum Extended. George Box said "all models are wrong, but some are useful." Extend this dictum to each of the following domains:

a) "All diagnoses are _, but some are ." b) "All maps are , but some are _." c) "All words are , but some are ." d) "All financial models are _, but some are ." e) "All categories are __, but some are _____."

For each, explain what makes the extension valid and what it illuminates about map-territory relations in that domain.

Part C: Application to Your Domain

These exercises connect map-territory concepts to your own field of study or work.

C1. Identify the single most important map (model, metric, theory, or framework) in your field. Then answer:

a) What territory does it represent? b) What features does it capture? Why are these the features it captures (what purpose was the map designed for)? c) What features does it omit? Are these omissions well-understood by practitioners, or have they been forgotten? d) At what level of map-territory confusion does your field typically operate with this map? e) Has the territory ever "fought back" -- have there been cases where reality contradicted the map, and how did your field respond?

C2. Design a "map warning label" for the most important model in your field. Like a food nutrition label or a pharmaceutical side-effects warning, this label should include:

a) What the model is designed for (its intended use) b) What the model assumes about the territory c) Known limitations and distortions d) Conditions under which the model should NOT be used e) What to do when the territory contradicts the model

C3. Identify a case in your field where someone at Level 3 (defending the map against the territory) caused harm. Describe:

a) What map was being defended b) What territory data contradicted the map c) How the defender responded to the contradicting data d) What the consequences were e) How the situation was eventually resolved (if it was)

Part D: Synthesis

These exercises require integration across multiple concepts from this chapter and previous chapters.

D1. The Unified Map-Territory Framework. The chapter argues that overfitting (Ch. 14), Goodhart's Law (Ch. 15), and legibility (Ch. 16) are all special cases of map-territory confusion. Construct a table with the following columns: Pattern | What is the map? | What is the territory? | What is the specific confusion? | What is the consequence? Fill in rows for overfitting, Goodhart's Law, legibility, the Gaussian copula, overdiagnosis, and linguistic relativity.

D2. Map-Territory Confusion and Cascading Failures. Using the cascading failure framework from Chapter 18, trace how a map-territory confusion can trigger a cascade. Choose one of the following starting points:

a) A financial model (like the Gaussian copula) underestimates risk. b) A medical imaging protocol detects an incidentaloma. c) A standardized test defines educational quality for a school district.

For your chosen scenario, map out at least five steps in the cascade, showing how each step follows logically from the previous one and how the cascade could have been interrupted by restoring Level 1 awareness at any point.

D3. Korzybski Meets Goodhart. Construct an argument that Goodhart's Law (Ch. 15) is a special case of the map-territory relation. Then construct a counter-argument that they are distinct phenomena. Which argument do you find more convincing, and why?

D4. The Map-Territory Paradox of Communication. This textbook is itself a map. It maps patterns across domains onto words on pages. Apply the map-territory framework to the book you are reading:

a) What territory does this book attempt to map? b) What does it capture well? c) What does it necessarily omit? d) At what level of map-territory confusion might a reader of this book operate, and how could the book design itself to prevent Level 2 or Level 3 confusion? e) Is there an irony in using a book (a map) to teach the lesson that maps are not the territory? How does the book handle this self-referential problem?

Part E: Advanced Extensions

These exercises push into territory beyond the chapter's explicit coverage.

E1. The Limits of Cartography. If "all knowledge is cartography," is there anything that is NOT a map? Consider:

a) Direct sensory experience (seeing red, feeling pain) -- is this a map or the territory? Make arguments both ways. b) Mathematical truths (2+2=4) -- is this a map of something, or is it the territory itself? c) Consciousness (the fact that there is "something it is like" to be you) -- is your experience of yourself a map, or is it the territory?

For each, explain how your answer affects the scope of Korzybski's claim. If some knowledge is NOT cartography, does that undermine the threshold concept, or does it define its boundaries?

E2. The Ethics of Map-Making. Maps are not neutral. They emphasize certain features and obscure others, and these choices have consequences.

a) Who decides which features a map includes? What power dynamics are embedded in this decision? b) Give three examples of maps that serve the interests of their makers at the expense of others (e.g., colonial maps, credit scoring systems, diagnostic categories). c) Is it possible to make a "fair" map? What would fairness mean in the context of cartography? d) How does the concept of "map-territory confusion" relate to the concept of ideology? Could an ideology be defined as a map that has been elevated to Level 3?

E3. Artificial Intelligence as Cartography. Machine learning models are maps trained on data. Apply the map-territory framework to AI:

a) What is the training data? (Hint: it is a map of a map.) b) What is the territory the model is supposed to represent? c) How many layers of mapping separate the model from the territory? d) Where is the model most likely to fail -- where will the map diverge from the territory? e) How does this analysis connect to the overfitting discussion in Chapter 14?

Part M: Interleaved Practice

These exercises mix map-territory concepts with ideas from earlier chapters to build cross-domain fluency.

M1. A hospital implements a patient satisfaction survey (a map of care quality). The survey becomes a Goodhart target (Ch. 15) as doctors optimize for patient happiness rather than health outcomes. The survey also functions as a legibility project (Ch. 16), making complex care quality visible to administrators. When a doctor argues that the survey does not capture real care quality, administrators defend the survey (Level 3 confusion). Eventually, the overemphasis on survey scores leads to iatrogenic harm (Ch. 19) as doctors prescribe unnecessary painkillers to boost satisfaction scores.

Trace this scenario step by step, labeling each step with the relevant concept: map-territory confusion (specifying the level), Goodhart's Law, legibility, iatrogenesis, and cascading failure. At which point(s) could the cascade have been interrupted?

M2. Consider the cobra effect (Ch. 21) as a map-territory confusion. The incentive designer creates a map of desired behavior (kill cobras → collect bounty → fewer cobras) and deploys it. The territory (the actual ecology of responses) diverges from the map (the designer's model). Using the three-level framework from this chapter, classify the following stages of the Delhi cobra bounty:

a) The initial design of the bounty. b) The period when the bounty appeared to be working. c) The moment the British administrators learned about cobra farming but initially dismissed the reports. d) The final cancellation of the bounty.

M3. Design a "map-territory awareness curriculum" for one of the following professions:

• Financial analysts
• Medical residents
• Urban planners
• Data scientists
• Policymakers

Your curriculum should include: a) Three key maps the profession relies on b) Common Level 2 confusions specific to the profession c) At least one historical Level 3 failure in the profession d) Practical exercises for maintaining Level 1 awareness e) A "red team" exercise where practitioners actively try to find the limits of their most important map

M4. Write a brief (200-word) dialogue between Korzybski and one of the following figures, in which they discover they have been talking about the same insight in different language:

a) George Box ("All models are wrong, but some are useful") b) James C. Scott (author of Seeing Like a State) c) Alan Watts ("Don't confuse the menu for the meal") d) David X. Li (creator of the Gaussian copula)

M5. Review your Pattern Library. You now have entries for (among others): feedback loops, gradient descent, overfitting, Goodhart's Law, legibility, cascading failures, iatrogenesis, cobra effects, and the map-territory relation. Draw a connection diagram showing how the map-territory relation connects to at least five of these earlier patterns. For each connection, write one sentence explaining the relationship.