Reading "Factfulness": 4 Instincts, 4 Corrections from a Formulation Lab

A recap of Hans Rosling's Factfulness and its 4 instincts (gap/negativity/linear/fear) — plus a formulation scientist's take on why "better and worse coexist" deserves a permanent spot in the lab.

Before opening Hans Rosling's Factfulness, I took the 13-question test at the front of the book. It said, "If you can answer all of them correctly, your picture of the world is probably accurate."

I got 3 right. The other 10 were all wrong.

Worse than random — blind guessing gives you a 33% hit rate; I scored below that.

It doesn't feel good. Every question in the book is essentially saying: most of what you think you know about the world is wrong.

But after finishing the book, I realized that being "called out" is far more valuable than reading something that makes you feel good. A formulation scientist makes judgments under uncertainty every day — and if your underlying intuitions about what the world actually looks like are wrong, then the foundation you stand on when reading data, forecasting stability, or evaluating process routes may be built on a pile of illusions.

This post is what I want to write down after reading Factfulness: the 4 instincts (my notes already had the first 4, but #4 I only had the categories of harm, never finished — this post completes it), the corrections for each, and a final section — as a formulation scientist, applying these 4 instincts to my daily decisions, I found them worth far more than "knowledge."

1. What the 13-Question Test Actually Tests

I don't remember all 13, but a few left an impression:

"What percentage of 1-year-old children in the world today have been vaccinated?" I picked 60%. The answer is 80%.
"Where do most of the world's poorest people live?" I picked "South Asia/Africa" without thinking. The answer is sub-Saharan Africa — but the share is lower than most people's gut impression.
"In the last 20 years, how has the proportion of the world living in extreme poverty changed?" I picked "basically unchanged." The answer is nearly halved (from ~29% in 1997 to ~9% in 2017).

On the second question, I had no information — I was just guessing based on "map + color" — which is exactly Rosling's "gap instinct" at work: Africa and South Asia carry a "backward country" label in my head, but the moment the data is laid out, that "label-driven judgment" collapses.

More importantly: behind every question, the author put real data from Gapminder (income per capita, child survival, energy use, vaccine coverage). Rosling was a public health researcher — none of his conclusions are moral lecturing; they are charts that can be directly verified.

A scientist's favorite way to be convinced: give me the data, give me the method, let me reproduce it myself.

2. 4 Instincts, 4 Corrections from a Formulation Lab

I'll first finish the 4 instincts from the book (note #4 was only the harm categories — I'll complete it), then describe how I deal with them in the BASF formulation lab.

2.1 Avoid the Gap Instinct

The gap instinct misleads us into treating smooth transitions as polarization, coexistence as divergence, and "common ground with differences" as outright contradiction.

The simplest correction Rosling offers: replace two-level categories with four-level ones.

Level 1: $1–2 per day
Level 2: $2–8 per day
Level 3: $8–32 per day
Level 4: more than $32 per day

Once the 1965–2017 global GDP data is laid out, most people are "bunched" in the middle of Levels 2 and 3 — there is no "developed vs. developing" chasm. Most so-called "developing countries" actually entered Level 3 back in the 1980s.

Rosling used 1965–2017 GDP and fertility data to falsify this binary intuition. But there's a key caveat he keeps flagging: watch the average trap — once you only look at the mean, you fall into another illusion.

When comparing averages: be careful of "the average," but pay even more attention to the actual distribution. Per-capita GDP may look dramatically different, but the vast majority of the world's people cluster tightly around the median — not the polarized picture the average suggests.
When comparing extremes: if you only look at the highest and lowest, the world looks more and more split; but if you plot the actual distribution of the middle 80%, the "middle ground" is far thicker than you imagined.
Look up, not down: many misperceptions come from "looking from the top down" — looking at the aggregate first, then drawing conclusions. The right move is to look at the distribution first, then conclude.

2.2 Avoid the Negativity Instinct

"The world is getting worse" is a major misperception — that is the negativity instinct.

It comes from three mechanisms we're not aware of:

We're wired to react more strongly to bad news (an evolutionary legacy — bad things could kill you).
The past gets memory-polished; the "good old days" filter makes us forget the past's bad parts.
News naturally selects the negative — "gradual progress is not news."

Better and worse coexist; good news is not news; gradual progress is not news; more bad news does not mean more bad things; beware the overly romanticized past.

Once you understand this, the way you read news can be made concrete with three habits:

When you see negative news, force yourself to ask "what's the corresponding positive data?" — Gapminder and Our World in Data have the counter-curve for almost every negative story.
When you see a claim that "the last 5/10 years got better/worse," first check the 50–100 year curve — most indicators rise steadily on a century scale; what the news reports is just the recent wobble.
Beware the "everything was better when we were kids" nostalgia — that's a polished memory illusion.

2.3 Avoid the Straight Line Instinct

The straight-line instinct comes from our ancestors' hunter-gatherer survival pressure — when the grass moved, run the other way; a useful simplification. Applied to today's world, that instinct gets things wrong constantly.

Some counter-examples from Rosling:

The spread of Ebola is exponential — it looks like nothing at first, then a jump in orders of magnitude tomorrow. Linear forecasting gets destroyed by exponentials.
World population growth has recently slowed; the total number of children in the future is essentially flat — not a runaway climb, but an approach to a plateau.
Female education years vs. fertility — not linear, but an S-curve.

More specifically, nature and human society have at least 4 curve shapes:

S-curve: starts at 0, accelerates, approaches a ceiling (e.g., tech adoption, female education rate)
Slide curve: fast, then slower, then flat (e.g., a new drug's sales curve, fast early expansion then plateau)
Hump curve: rises then falls (e.g., early adopters in innovation diffusion, market lifecycle of a formulated product)
Exponential curve: slow, then extremely fast (e.g., epidemic spread, compound interest)

Before predicting, first look at what shape the curve actually is — don't pretend everything is a straight line.

2.4 The Fear Instinct (I never finished this section before)

The fear instinct drives us to overreact to three categories of risk:

1. Physical harm: violence and destruction from people, animals, sharp objects, natural environment
2. Entrapment: loss of control, loss of freedom, being held down
3. Contagion: infection or poisoning by invisible substances

Rosling's correction is direct — don't be misled by the visibility of risk; calibrate with data:

The more visible a risk, the deeper the memory — but not necessarily the higher the actual frequency.
The share of deaths from violence (war, homicide, terror) in the modern world is declining, but people's fear of violence rises because of the news.
Far more people die from "invisible dangers" (chronic disease, air pollution) than from "visible dangers" (car crashes, violence) — but the former has almost no news value.

Concrete habit: when you encounter something "frightening," first ask three questions —

What is the probability of this event (per year / per 10,000 people)?
How does its scale of harm compare on a historical scale?
Where does it rank among all causes of death/harm? (Only if it's in the top 5 is it worth genuinely worrying about.)

This is the most counter-intuitive of the 4 instincts, because it means much of what you fear is wrong — and the more you keep "scrolling the news for horrors," the more your picture of the world drifts away from the real risk distribution.

3. How These 4 Instincts Show Up in the Formulation Lab

That's the book. Now comes the part I most identify with as a formulation scientist — these 4 instincts show up in the formulation lab almost every day, in subtler forms.

3.1 Gap = "This Formulation Succeeded vs. Failed"

The biggest "gap" trap in the formulation lab is sorting formulations into "works vs. doesn't work."

But the real distribution is nothing like that — most formulations are "basically usable, but a couple of metrics fall short." By one criterion it's a failure; swap the criterion and it might be a local win.

A safer framework is to replace "success/fail" with a multi-metric distribution:

Every formulation experiment should list at least 6 orthogonal metrics (thermal stability, foam, rheology, storage stability, pH drift, sensory evaluation) — don't just look at pass/fail.
The actual distribution across these 6 metrics is usually uneven — some hit target, some don't, some exceed expectations. The whole picture isn't "a binary split" — it's a multidimensional map.
Real interpretation is: which metrics exceeded expectations (unexpected wins), which fell short (need correction), which were exactly as expected (consolidate the baseline). Each of these three situations carries different value for formulation development.

A specific habit I've now built: for any failed formulation experiment, force myself to list 3 metrics I didn't pre-specify, and re-examine the data. Most of the time, you find the "binary split" crushed some signal — and it crushed the signal itself, not just the conclusion.

3.2 Negativity = "Failures Spread Farther Than Successes"

The formulation lab has a counter-intuitive phenomenon: "this formulation failed" is more worth reporting than "this formulation succeeded."

The reason is a mirror of the negativity instinct — success gets attributed to "of course it worked," but failure gets dissected carefully. So the truly valuable knowledge is often hidden in the negative experience.

I have a rule for my weekly lab report: if a week contains only one "why this formulation is good," it's a press release; if it contains three "why this formulation is bad," that's actual technical accumulation.

That's why Gapminder-style tools can be used in reverse in the formulation lab — instead of looking at the "successful formulation" timeline, look at why the old formulation that got replaced was replaced. What got replaced exposes the real process constraints.

3.3 Straight Line = "The Shape of the Scale-Up Curve Matters More Than the Scaling Factor"

Scale-up (lab → pilot → plant) is the engineering problem a formulation scientist faces every day, and the step most easily misled by straight-line thinking.

The most common failures I've seen:

Stirring speed scaled from 200 rpm to 1500 rpm: you think it's a 7.5× scale-up, but it's nonlinear — shear rate is an S-curve or logarithmic function for some processes.
Dosing time scaled from 30 min to 8 h: you think it's a 16× extension, but the underlying reaction/crystallization kinetics may be exponentially sensitive.

My new default move: for any "linear scale-up" assumption, force myself to draw 2 curves — one is the real curve (data points + fit), the other is the straight-line prediction (dashed). Once the two lines diverge, the curve is not linear — reassessment is mandatory.

Of the 4 instincts, this is the most direct and most valuable tool for me — it can immediately stop a "just scale linearly" proposal in a process review meeting.

3.4 Fear = "The Most Dangerous Formulation Risk Is Usually the One That Doesn't Look Dangerous"

The classic fear-instinct traps in the formulation lab:

Known acute toxicity: everyone is alert, every SOP covers it. What really causes incidents and recalls is chronic, low-dose, invisible risk — e.g., degradation products of a raw material under long-term storage, side reactions of an additive under humidity variation, cumulative effects of a solvent residue.
"Strange but rare" phenomena get reported loudly: e.g., a color anomaly in one batch — but on a long horizon, the correlation between color anomalies and performance issues is weak.
"Controllable but invisible" processes: e.g., cumulative effects of a trace oxidation step — these are actually the real source of most batch-to-batch variation.

Concrete habit: for any process change, first ask three questions —

Is this change captured by any measurable metric (color, odor, pH, rheology, post-storage drift…)?
If no orthogonal measurable metric exists, it's an "invisible danger" — prioritize evaluation, don't defer it.
Is its impact cumulative or one-shot? (Cumulative is far more dangerous than one-shot, because the fear instinct is sensitive to single events and insensitive to cumulative ones.)

This one, after I read the fear-instinct chapter, immediately corrected my SOP priority: I used to put "visible acute risk" at the top; now I put "invisible cumulative risk" on equal footing with acute risk.

4. The 4 Instincts, Bundled into a Daily Prompt

After reading, I left myself a very simple prompt (stuck on the first page of my lab notebook):

Factfulness 4 Prompts (once daily, for any data judgment):

1. Gap instinct? — Are you looking at the real distribution instead of the average?
2. Negativity instinct? — Better and worse coexist; check the 50-year curve.
3. Straight line instinct? — Draw 2 curves; assume the curve is one of: S / slide / hump / exponential.
4. Fear instinct? — Check whether the "invisible danger" matters more.

In any judgment, once you've asked these 4 questions — a conclusion that seemed intuitively solid often collapses.

That is the most counter-mainstream thing about Factfulness: it doesn't teach you new knowledge — it teaches you to make fewer mistakes.

References

Hans Rosling, Factfulness: Ten Reasons We're Wrong About the World—and Why Things Are Better Than You Think (2018)
Gapminder.org (the data visualization site Rosling founded; all 13 questions and global trend data are available there)
Our World in Data (long-horizon global data, free and open; corresponds to the "50-year curve" section)