The Bigger Picture: Can We Make Harm Unprofitable?
A thought experiment about capitalism, corporate structures, human behavior—and unintended consequences
This website began with a question about type 2 diabetes.
That investigation led us from biology to food, from food to corporate incentives, and from corporate incentives to healthcare economics.
But eventually another question appeared.
What if the pattern we were looking at wasn't really about diabetes?
What if diabetes was simply one particularly visible example of something much larger—one of several observations about the diabetes ecosystem that point toward the same underlying pattern?
This is a thought experiment about that possibility.
It is not an argument against corporations.
It is not an argument against capitalism.
And it is certainly not an argument that people running corporations deliberately set out to harm anybody.
The question is more interesting than that.
Could perfectly rational decisions made by perfectly reasonable people produce harmful outcomes because the economic system rewards the behaviors that produce them?
And if so:
Can we make harm unprofitable?
Start with diabetes
The central Diabetes Is Profitable argument is that the biological mechanisms behind type 2 diabetes are well understood—but biology alone may not tell us very much about why so many people are developing the disease.
Human beings did not suddenly change biologically.
Our environment did.
Food became extraordinarily available, inexpensive, convenient and increasingly optimized for consumption.
None of this requires anyone in the food industry to want people to develop diabetes.
Quite the opposite.
Imagine a perfectly decent executive running a food company.
The company competes with other food companies.
It needs customers.
It needs revenue.
If people buy more of its products, that's generally good for the company.
If a competitor discovers a product that consumers find more appealing, our hypothetical executive needs to respond.
The company that gets progressively better at selling food tends to outperform the company that gets progressively worse at it.
That's capitalism doing something it does remarkably well.
It optimizes.
The interesting question is what happens when the thing being optimized isn't perfectly aligned with long-term human welfare.
Then we noticed something
Once we started looking at diabetes this way, the same basic pattern seemed to appear elsewhere.
Not identically.
Not necessarily for the same reasons.
And these examples do not prove that a single mechanism explains all of them.
But the similarities are striking enough to justify a thought experiment.
News: when fear captures attention
Imagine a woman walking alone through a park. A man approaches from the opposite direction. Perhaps he is seventy years old. Perhaps he is simply taking the same afternoon walk she is.
How dangerous is that encounter?
And how did the woman develop her perception of the danger?
Why danger captures attention
Human beings naturally pay attention to threatening information. Evolution offers an obvious explanation. An ancestor who occasionally mistook a harmless rustle for a predator wasted a little energy. An ancestor who repeatedly mistook a predator for a harmless rustle may not have survived to reproduce.
Our minds therefore appear to give possible danger special priority. Across seventeen countries, researchers measuring people's physiological responses to television news found a broadly shared tendency to react more strongly to negative news than to positive news—although individuals varied substantially.¹
There is nothing irrational about having a danger-detection system. The problem begins when an ancient survival mechanism enters a modern marketplace in which attention can be measured, sold and optimized.
The interesting part is what happens when attention to danger becomes commercially valuable.
When one negative word has economic value
A news organization needs viewers, readers, listeners or clicks. We now have unusually strong experimental evidence that negative language can attract more of them.
A large study published in Nature Human Behaviour examined more than 105,000 headline variations generating approximately 370 million impressions. For an average-length headline, adding just one negative word increased the click-through rate by about 2.3%. Adding a positive word reduced it.²
Two-point-three percent may sound small until we remember what was changed. The article did not become more alarming. The entire headline was not rewritten around catastrophe. A single negative word was added.
At the scale of hundreds of millions of impressions, even that small advantage is commercially meaningful. It gives publishers a measurable reason to select one word rather than another—and repeated optimization can gradually change the emotional character of the news people receive.
When the whole story is negative
The effect becomes much larger when negativity characterizes the article rather than one isolated word.
A 2024 analysis of 95,282 news articles and approximately 579 million Facebook and Twitter posts found that negative articles were shared about 1.91 times as often as non-negative articles overall—roughly 91% more sharing. The estimated advantage varied by outlet and platform, but the direction was remarkably consistent.³
These findings fit together. One negative word can improve a headline's performance. A story whose overall emotional orientation is negative can receive dramatically more social distribution.
This does not show that journalists gather in a room and decide to make society frightened. It shows something more consequential for this argument:
Negativity has measurable economic value in an attention-based system.
Every story can be true—and the picture can still be false
Now add another peculiar feature of mass communication.
In a country containing hundreds of millions of people, extraordinarily unusual things happen every day. Imagine an event with only a one-in-a-million annual probability for any one person. Across 350 million people, we would still expect roughly 350 instances—potentially almost one horrifying story every day.
Every report could be accurate. Every victim could be real. Nothing would need to be fabricated.
The distortion does not require misinformation. It can arise from selecting the rare, vivid and frightening event every day while giving little attention to the millions of ordinary encounters in which nothing happened.
The harmless encounters supply the denominator. The news supplies almost exclusively the numerator.
In simple terms: The news shows us the rare times something bad happens. It almost never shows us the millions of times the same kind of situation happens and everyone is safe.
A society can become safer while believing it is becoming more dangerous
Crime offers a striking real-world illustration. FBI data analyzed by Pew Research Center show that the U.S. violent-crime rate fell 49% between 1993 and 2022. Yet in 23 of 27 Gallup surveys conducted over that period, at least 60% of Americans said crime nationally had increased from the previous year.⁴
United States, 1993–2022
The majority response in 23 of 27 Gallup surveys
That discrepancy does not prove that news exposure alone caused the public belief. Political messaging, personal experience, social media and many other influences matter. But it does show that lived statistical reality and perceived danger can move in opposite directions.
Research on local crime news points in the same direction. In a 2024 Pew survey, people who consumed local crime news often were more likely to say they were concerned about becoming a victim. Women reported fear more often than men—49% compared with 34%.⁵ Correlation is not proof of causation, but the pattern is consistent with a substantial body of research showing that repeated exposure to threatening information can influence perceived risk.
Possibility is not probability
A frightening event being possible does not tell us how probable it is. That distinction matters because human intuition is poor at comparing a vivid, memorable danger with an ordinary risk encountered thousands of times.
Driving is a useful comparison. It feels familiar and controllable, yet motor-vehicle crashes killed 39,345 people in the United States in 2024.⁶ The risk of walking past a particular older man in a busy public park cannot be reduced to one honest national statistic: the denominator is not measured, the context varies, and the identity of an offender cannot be inferred from a passerby's age or sex.
What the available evidence can tell us is that violence against women is much more often committed by someone the victim knows than by a stranger. Bureau of Justice Statistics data on female murder victims in 2021 found that, where the relationship was known, 76% were killed by someone they knew; 34% were killed by an intimate partner, and 12% by a stranger.⁷ Historical federal data on rape and sexual assault similarly found that most offenders were known to the victim.
None of this makes an unfamiliar person risk-free. It changes the question from “Could something happen?”—to which the answer is nearly always yes—to “How large is this risk compared with the other risks I routinely accept?”
Calibrated caution
Sonia offers a useful human example. She travels internationally, takes road trips alone, sleeps in her car in carefully chosen restaurant parking areas and stays by herself in state parks. Some of her friends regard this as extraordinarily brave. Others regard it as reckless.
But Sonia is not oblivious to risk. She thinks carefully about where she stops. She favors orderly, well-used places, avoids rundown or isolated areas, pays attention to her surroundings and leaves when something does not feel right.
There is a difference between minimizing risk and organizing your life around fear.
Sonia practices the first without surrendering to the second.
Her story is an illustration, not scientific evidence. Nor is she alone. Significant online communities of women share information about solo road trips, car camping and independent travel. Their existence raises an interesting question: how can some women understand the risks, take sensible precautions and still experience freedom, while others experience a level of fear that may be far greater than the statistical danger?
Different information environments
A key difference is likely to be the information environment people inhabit.
Sonia rarely watches sensationalized news broadcasts. She does not spend her days clicking alarming headlines or consuming a continuous stream of crime stories. She reads books, follows subjects and hobbies that interest her, and uses general overview-oriented sources to remain aware of what is happening in the world.
That does not prove that her media habits caused her relative lack of fear. It does, however, illustrate a conclusion supported by the broader evidence: people do not estimate risk from crime tables alone. They estimate it partly from the examples repeatedly made available to their minds.
If the information environment repeatedly presents the exceptional tragedy as though it represents ordinary life, exaggerated fear is a predictable result.
The physical and social cost of exaggerated fear
Fear changes behavior. A woman who believes ordinary public spaces are extremely dangerous may walk less, travel less, explore less and surrender opportunities she might otherwise enjoy. Chronic fear can also affect sleep, stress and wellbeing.
The effects extend to children. Independent travel and outdoor freedom have declined dramatically across several affluent countries. In England, the share of 10- and 11-year-olds allowed to travel home from school without an adult fell from 94% in 1970 to 54% in 1990 and 47% in 1998. Australian research found a similarly sharp decline in children's independent travel home from school between 1991 and 2012.⁸ Changing streets, traffic, family schedules and social expectations all contributed; media fear cannot be assigned sole responsibility. But an exaggerated perception of stranger danger is part of the environment in which parents make these decisions.
Safety precautions can be rational. Fear can be protective. But fear that is grossly disproportionate to actual risk has costs of its own.
When reporting harm creates additional harm
We need journalism to report crime, investigate institutional failures and warn people about genuine dangers. Concealing harm would not make society safer.
The structural problem arises when the commercial reward for frightening material leads rare dangers to be selected, repeated and emphasized far beyond their prevalence in everyday life.
Corporate structures need not intend the fear. They need only measure that negative material performs better and respond rationally to that signal.
The revenue is captured by the organization. The cost of the resulting fear is distributed across society.
The questions we should be asking
How can people distinguish a danger that is vividly imaginable from one that is genuinely probable?
How can news organizations report real tragedies without allowing exceptional events to define people's perception of ordinary life?
What would journalism look like if accuracy included not only whether each story was true, but whether the overall picture was proportionate?
If disproportionate fear causes real harm—and producing that fear is profitable—how do we make that harm unprofitable?
Social media: when engagement becomes the objective
The same conflict now reaches beyond social platforms. A related examination asks whether artificial intelligence is different—or the same old corporate story.
Now consider social media.
Human beings want social connection.
We care what other people think about us.
We compare ourselves with others.
We seek approval.
We respond to novelty.
Again, none of those characteristics was invented by a technology company.
They are part of being human.
But now introduce a business whose economic success depends partly upon keeping people engaged.
Every additional minute of attention can have economic value.
So systems can optimize.
Which notification brings someone back?
Which content keeps them scrolling?
Which recommendation produces another click?
Which material generates another interaction?
Modern algorithmic systems can measure these behaviors at enormous scale. They do not necessarily need a theory of human psychology to discover which choices increase engagement.
And there is legitimate reason to be concerned about whether maximizing engagement and maximizing wellbeing always point in the same direction.
The U.S. Surgeon General's advisory on social media and youth mental health concluded that we cannot currently say social media is sufficiently safe for children and adolescents. It also reported evidence that young people spending more than three hours per day on social media face roughly twice the risk of certain mental-health problems, while explicitly recognizing that social media can also provide important benefits.⁹
That distinction matters.
This isn't an argument that social media is bad.
It isn't even proof that social media causes every harm associated with heavy use.
It is an observation about alignment:
Maximum engagement and maximum human wellbeing are not necessarily the same objective.
And if the economic system rewards one of those far more directly than the other, we should probably pay attention.
Addiction: when the best customer may be the most vulnerable customer
Consider gambling.
Millions of people can visit a casino, spend some money, enjoy themselves and leave.
There is nothing inherently mysterious about that.
But some people respond very differently.
Some human beings are particularly susceptible to compulsive behavior and addiction.
Now the economic question becomes uncomfortable.
What happens when the customers who have the greatest difficulty regulating their consumption can also become some of the most economically valuable customers?
This isn't merely hypothetical.
The World Health Organization reports estimates suggesting that people gambling at harmful levels generate around 60% of gambling losses—that is, gambling revenue. A systematic review of gambling consumption and harm likewise found substantial concentration of spending among people experiencing gambling problems, although the extent varies significantly by type of gambling.¹⁰
Research into electronic gaming machines adds another interesting dimension. A systematic review of experimental research found that design features such as "near misses" can motivate continued play.¹¹
The point is not that gambling businesses are uniquely evil.
Nor is the argument that gambling, alcohol or other potentially addictive activities should necessarily disappear.
The underlying human susceptibility is not something corporations invented.
The structural question is narrower:
What happens when an economic system can become more profitable as vulnerable customers consume more?
The same question can potentially be asked about alcohol, nicotine and other products or behaviors capable of becoming addictive.
Again, we don't need to establish malicious intent.
The incentive conflict exists whether anyone intended to create it or not.
Healthcare: when treating the problem creates economic activity
Healthcare is different from the previous examples, but it raises another version of the same puzzle.
We want doctors.
We want hospitals.
We want medicines.
We want researchers developing better treatments.
Modern medicine has produced extraordinary improvements in human life.
But healthcare also creates economic activity when people become sick.
Tests generate revenue.
Procedures generate revenue.
Medications generate revenue.
Ongoing treatment generates revenue.
Preventing someone from becoming sick in the first place can be enormously valuable to that person and to society.
But where does the economic reward for that prevention appear?
This isn't merely a philosophical concern.
For years, healthcare policymakers themselves have explicitly discussed the problem of payment systems that reward the volume of services rather than the value of outcomes.
The Centers for Medicare & Medicaid Services has described traditional fee-for-service arrangements as rewarding providers according to how many patients they see, tests they perform or procedures they deliver. CMS has been developing value-based payment systems specifically to better align financial incentives with quality, outcomes and population health.¹²
That is fascinating in the context of this thought experiment because it demonstrates something important:
Incentive structures can be redesigned.
The problem is not necessarily healthcare companies or healthcare professionals.
The problem may be that different payment structures reward different behaviors.
And so another alignment question emerges:
Can keeping someone healthy become at least as economically compelling as treating them after they become sick?
Five different domains. One curious pattern.
We now have examples involving:
Food.
Healthcare.
News.
Social media.
Addiction.
Dating platforms provide another concrete case: a service intended to relieve loneliness may benefit when users continue searching.
These are very different parts of society.
Yet something similar seems to keep appearing.
A normal human characteristic or vulnerability exists.
Appetite.
Attention to danger.
Need for social approval.
Susceptibility to reward and addiction.
Illness and biological vulnerability.
A commercial system interacts with that characteristic.
Competition rewards organizations that become progressively better at producing an economically valuable outcome.
And sometimes the economically valuable outcome and the long-term interests of the human being aren't quite the same thing.
That doesn't prove a grand theory.
But it seems sufficient to justify asking whether there is one.
Imagine you're an alien
Many years ago, I encountered a science-fiction story whose title and author I can no longer remember.
I barely remember the story.
But one idea from it stayed with me for decades.
Aliens arrived on Earth and observed our civilization.
They concluded that humans weren't actually the dominant intelligent entities on the planet.
Corporations were.
I don't remember precisely how the author developed the idea, and I don't want to pretend that I do.
But the concept is fascinating as a thought experiment.
Look at a corporation from sufficiently far away.
It acquires resources.
It processes information.
It competes with other corporations.
It adapts.
It grows.
It develops mechanisms for protecting itself.
It retains institutional memory.
Its individual human components come and go while the organization itself persists.
Executives leave.
Employees retire.
Shareholders sell.
Generations die.
The corporation continues.
Corporations aren't literally biological organisms, of course.
Nor does this analogy require us to imagine that corporations are conscious.
But viewed as systems, they behave in some surprisingly organism-like ways.
And perhaps most importantly:
They undergo something resembling selection.
Organizations that succeed tend to survive.
Organizations that fail tend to disappear.
Successful strategies are copied.
Unsuccessful strategies are abandoned.
Which raises an intriguing possibility.
Maybe asking corporations to behave better misses part of the problem.
Perhaps we should be asking:
What behaviors are we selecting for?
Capitalism may be working exactly as designed
This is where it would be very easy to misunderstand the argument.
This isn't an argument against capitalism.
Quite the opposite.
Capitalism has arguably been the most extraordinarily productive economic system humans have yet developed.
It harnesses competition.
It rewards innovation.
It discovers efficiencies.
It reallocates resources.
And above all:
It optimizes.
That is an extraordinary capability.
But powerful optimization systems have an interesting characteristic.
You need to be very careful about what you ask them to optimize.
If increased consumption produces greater economic success, businesses will become extraordinarily good at increasing consumption.
If attention produces greater economic success, businesses will become extraordinarily good at capturing attention.
If engagement produces greater economic success, businesses will become extraordinarily good at generating engagement.
If treatment is rewarded more directly than prevention, enormous resources may naturally accumulate around treatment.
None of this requires bad people.
It requires only a powerful optimization mechanism responding rationally to the incentives we've given it.
Perhaps capitalism's potential weakness, therefore, isn't that it doesn't work.
It's that it works extremely well.
Don't fight the optimizer. Change what winning means.
Suppose we discover that some corporate behavior produces substantial social harm.
One response is to ask corporations to behave differently.
Be more responsible.
Put society first.
Accept lower profits.
Sometimes that may work.
But there is an obvious structural problem.
If one corporation voluntarily accepts a significant competitive disadvantage and another doesn't, which one does the economic environment reward?
Eventually we are back to selection.
So perhaps the more durable solution isn't asking corporations to stop behaving economically.
Perhaps it is changing the economic environment in which they compete.
Make the socially desirable behavior economically advantageous.
Make harmful behavior economically disadvantageous.
Then unleash capitalism on the problem.
Let companies compete.
Let entrepreneurs innovate.
Let investors search for opportunities.
Let businesses relentlessly optimize.
But change what constitutes winning.
Healthcare already gives us a limited real-world example of the principle. Value-based payment programs attempt to reward quality and outcomes rather than simply rewarding greater quantities of medical services. Whether particular programs succeed is an empirical question, but the underlying idea is exactly the one we're considering:
change the incentive and allow rational economic behavior to respond.
Can we make harm unprofitable?
That may be the bigger question hiding behind our investigation of diabetes.
Not:
How do we stop corporations making profits?
Not:
How do we persuade executives to become better people?
Not even:
How do we make corporations responsible for human welfare?
Something simpler.
Can we make harm unprofitable?
Can we design economic structures so that producing significant harm ceases to be a competitive advantage?
Can we arrange the incentives so that the economically rational decision and the socially desirable decision increasingly point in the same direction?
Can we harness capitalism's extraordinary ability to optimize rather than fight against it?
We don't pretend to have the answer.
That's why this is a thought experiment.
And we deliberately haven't tried to produce a catalogue of every industry or social problem to which the idea might apply.
That would rather defeat the purpose.
If the pattern is real, perhaps you will recognize it somewhere we haven't mentioned.
Perhaps you won't.
Perhaps there are important reasons why these examples aren't as similar as they first appear.
That's worth thinking about too.
Because this isn't intended to tell you what to think.
It is an invitation to look at familiar problems from a slightly different direction.
Instead of asking:
Who is behaving badly?
Perhaps sometimes we should ask:
What behavior is the system rewarding?
And then one final question:
Can we make harm unprofitable?
Evidence and further reading
1. Biological negativity bias. Soroka S, Fournier P, Nir L. Cross-national evidence of a negativity bias in psychophysiological reactions to news. Proceedings of the National Academy of Sciences. 2019. Participants across 17 countries showed a general tendency toward stronger physiological responses to negative news, with meaningful individual variation.
2. Negative words and clicks. Robertson CE, Pröllochs N, Schwarzenegger K, et al. Negativity drives online news consumption. Nature Human Behaviour. 2023. More than 105,000 randomized headline variations and approximately 370 million impressions showed a 2.3% click-through increase from one additional negative word in an average-length headline.
3. Negative articles and sharing. Negative online news articles are shared more to social media. Scientific Reports. 2024. Analysis of 95,282 articles and approximately 579 million posts estimated that negative articles received about 1.91 times the social sharing of non-negative articles overall.
4. Crime and national perception. Pew Research Center analysis of FBI crime statistics and Gallup public-opinion data. The U.S. violent-crime rate fell 49% between 1993 and 2022, while majorities in 23 of 27 Gallup surveys said national crime had risen.
5. Local crime news and fear. Pew Research Center. Americans' Experiences With Local Crime News. 2024. Frequent consumers of local crime news reported greater concern about victimization; women reported fear more often than men.
6. Road risk. National Highway Traffic Safety Administration. 2024 Traffic Fatalities Data. NHTSA estimated 39,345 U.S. traffic deaths in 2024.
7. Victim-offender relationships. Bureau of Justice Statistics, Female Murder Victims and Victim-Offender Relationship, 2021, together with federal analyses of rape and sexual assault. These sources show that violence against women is more often committed by a known person than by a stranger.
8. Children's independent mobility. Hillman M, Adams J, Whitelegg J. One False Move: A Study of Children's Independent Mobility, and subsequent British and Australian research. Multiple studies document large generational declines in children's independent travel and outdoor freedom.
9. Social media and youth mental health. U.S. Surgeon General. Social Media and Youth Mental Health. The advisory summarizes evidence of both benefits and potential harms and concludes that social media cannot currently be considered sufficiently safe for children and adolescents.
10. Gambling harm and revenue concentration. World Health Organization, Gambling fact sheet; and systematic research examining concentration of gambling expenditure among people experiencing gambling problems.
11. Gambling-machine design. Barton KR et al. The Effect of Losses Disguised as Wins and Near Misses in Electronic Gaming Machines: A Systematic Review. Journal of Gambling Studies. 2017.
12. Healthcare incentives. Centers for Medicare & Medicaid Services materials on fee-for-service and value-based payment programs, including CMS's explicit distinction between paying for volume and paying for quality/value.