THE GREAT ATTENTION EXPERIMENT: 5 - THEY DIDN'T INVENT HUMAN PSYCHOLOGY. THEY LEARNED TO USE IT.
You unlock your phone.
There is no urgent message waiting.
Nobody has called.
You do not need directions.
You are not checking the weather.
You are not paying a bill, answering an email or looking for information.
For a moment, you cannot even remember why you picked it up.
Yet your thumb already knows what to do.
An application opens.
The screen refreshes.
Something appears.
Perhaps it is interesting.
Perhaps it is not.
You move to the next thing.
Then the next.
Several minutes later, you realise you have been pulled into an activity you never consciously decided to begin.
It is tempting to describe that experience as a failure of discipline.
Perhaps sometimes it is.
But that explanation ignores how much effort has gone into making the behaviour easy, repeatable and increasingly automatic.
The people who built modern digital products did not invent curiosity.
They did not invent boredom.
They did not invent our need for social approval.
They did not invent our attraction to novelty, our tendency to repeat rewarded behaviour or our preference for the easiest available path.
Those characteristics existed long before the internet.
They helped human beings learn.
Belong.
Explore.
Avoid danger.
Build relationships.
Conserve energy.
Adapt to changing environments.
Technology companies did not create human psychology.
They learned how to design around it.
Then they learned how to measure which designs worked.
Then they tested those designs across millions and eventually billions of people.
That progression matters.
Because there is a profound difference between creating a product people enjoy using and deliberately engineering the conditions under which using it becomes increasingly automatic.
Persuasion Did Not Begin With Technology
Human beings have always tried to influence one another.
Parents influence children.
Teachers influence pupils.
Governments influence citizens.
Religions influence believers.
Advertisers influence consumers.
Friends influence friends.
Architecture influences movement.
Language influences interpretation.
A staircase encourages one route.
A locked door prevents another.
A supermarket places certain products at eye level.
A road sign prompts a driver to slow down.
Design has always shaped behaviour.
Technology made the process more systematic.
A printed sign remains the same regardless of who reads it.
A digital interface can change.
It can observe what someone does.
It can compare their response with the responses of thousands of other people.
It can alter the wording, timing, colour, position or sequence of an experience.
It can identify which version produces the desired behaviour most reliably.
Then it can repeat that version at enormous scale.
This did not initially emerge as a project to capture attention.
Researchers recognised that computers could help people learn, exercise, save money, take medication, reduce energy use and adopt healthier habits.
If technology could make beneficial behaviour easier, that represented a genuine social opportunity.
The field became known as persuasive technology.
Computers Became Tools for Behaviour Change
In the late 1990s, researchers at Stanford began studying how computers could be designed to influence what people thought and did.
The field was sometimes called captology, derived from the phrase computers as persuasive technologies.
BJ Fogg, who founded Stanford's Persuasive Technology Lab, later developed a practical model for understanding how behaviour occurs.
The Fogg Behaviour Model proposes that three elements must converge at the same moment:
- motivation;
- ability;
- a prompt.
A person must possess sufficient motivation.
They must be capable of performing the behaviour.
Something must prompt them at the relevant moment.
If one of those elements is missing, the intended behaviour is less likely to occur.
The model is not a theory of social-media addiction.
It is not inherently manipulative.
It can explain why someone remembers to take medication.
Why they complete an exercise session.
Why they donate to a charity.
Why they reply to a message.
Why they open an application.
The ethics depend upon the behaviour being encouraged, the methods being used and whether the outcome serves the person being influenced.
But consider how useful the model becomes to a company that benefits when a person performs one particular action more frequently.
Open the application.
Watch the video.
Send the message.
Post the photograph.
Click the notification.
Make the purchase.
Return tomorrow.
The problem becomes easier to break down.
Is the person sufficiently motivated?
Is the action easy enough?
Can we prompt them at the right moment?
That is not mind control.
It is something more ordinary.
Behavioural engineering.
Motivation, Ability and Prompt
Imagine a person has downloaded a social application but rarely uses it.
The company could try to increase motivation.
It might show interesting content.
Emphasise social connection.
Display positive feedback.
Remind the person that friends are active.
Create fear that something is being missed.
Promise entertainment.
Offer recognition.
The company could also increase ability.
Reduce the number of steps.
Keep the person logged in.
Preload the content.
Allow one-tap reactions.
Remember preferences.
Make posting simpler.
Begin the next video automatically.
The easier the action becomes, the less motivation is required to perform it.
Finally, the company can introduce a prompt.
A sound.
A vibration.
A red badge.
A message.
A reminder.
A recommendation.
An email announcing that someone has responded.
The behaviour occurs when the three elements meet.
A motivated person encounters an easy action at the moment they are prompted.
Good interface design often works precisely because it reduces unnecessary effort.
Nobody wants to enter their password repeatedly.
Nobody wants to complete fifteen steps to send a photograph.
Nobody wants a video to buffer constantly.
Ease is not the enemy.
But removing friction changes behaviour.
The fewer moments at which someone must stop, think or decide, the easier it becomes to continue without conscious intention.
Friction Is Not Always a Problem
Technology design often treats friction as something to eliminate.
Usually, this makes sense.
A complicated payment system causes abandoned purchases.
A confusing form discourages applications.
An inaccessible website excludes users.
A slow page wastes time.
A difficult interface creates frustration.
Removing those barriers can improve people's lives.
But friction also performs a psychological function.
It creates a moment of choice.
Think about the difference between opening a packet of biscuits and finding one already placed in your hand every few minutes.
The biscuit itself has not changed.
The effort separating one from the next has.
That effort creates a boundary.
You must decide to reach into the packet again.
Digital systems can remove equivalent boundaries.
The next video begins without being selected.
The next post appears with a movement of the thumb.
Payment details are remembered.
A message can be reacted to instantly.
A feed has no final page.
Each individual action feels insignificant.
The user is spared effort.
The company also avoids creating a moment in which the user might decide to stop.
YouTube describes Autoplay as a feature that makes deciding what to watch next easier by automatically playing a related video when the current one ends. The company now switches Autoplay off by default for users aged thirteen to seventeen and allows users or parents to control the setting.
That safeguard is important.
So is the design principle beneath the original feature.
A natural stopping point existed.
The programme ended.
Another decision would ordinarily have been required.
Autoplay made that decision unnecessary.
Convenience and continued consumption became the same feature.
The Prompt Does Not Need to Be Answered to Work
Notifications are often discussed as though their only effect occurs when someone opens them.
Research suggests the prompt itself can carry an attentional cost.
In a 2015 experiment, researchers found that receiving a mobile-phone notification significantly disrupted performance on an attention-demanding task even when participants did not interact with the device.
The sound or vibration does not need to produce a conscious response to interrupt the cognitive process already underway.
It introduces a question.
Who was that?
What happened?
Do I need to check?
Can it wait?
Even when the answer is yes, part of the mind has already moved.
A later study comparing adolescents, young adults and middle-aged adults found that notification sounds affected mathematical task performance, with the strongest effects appearing among the adolescent participants.
These findings do not prove that every notification is harmful.
Some are valuable.
A message may be urgent.
A calendar reminder may prevent a missed appointment.
A banking alert may reveal fraud.
A school notification may matter to a parent.
Prompts exist because information sometimes deserves attention.
The design question is whether the notification serves the person receiving it or primarily serves the product sending it.
A message telling you that your child needs collecting serves an obvious human purpose.
A vague alert designed to make you reopen an application may serve another.
Both arrive through the same device.
The nervous system must treat each as potentially important before it knows which is which.
Human Beings Are Sensitive to New Information
Novelty has always mattered.
A movement in the bushes might signal danger.
A change in someone's expression might reveal anger.
A new sound might indicate an opportunity or threat.
Information about other people could affect belonging, reputation and survival.
Paying attention to change was adaptive.
A predictable environment can be safely ignored.
An unpredictable environment demands monitoring.
Modern digital feeds concentrate novelty.
Every movement reveals something different.
A face.
A headline.
A joke.
A conflict.
A message.
A product.
A tragedy.
A celebration.
A piece of information.
A threat.
A reward.
The content does not need to be consistently good.
It needs to remain sufficiently unpredictable that the next item might be.
This is where discussions of digital behaviour often collapse into one word.
Dopamine.
The Dopamine Explanation Is Usually Too Simple
You may have heard that every notification, like or video gives the brain a dopamine hit.
The phrase sounds scientific.
It is often used as though dopamine were a small dose of pleasure released whenever someone enjoys something.
Human neurobiology is more complicated.
Dopamine is involved in movement, motivation, learning, attention and the way organisms update expectations about rewards.
Classic research by Wolfram Schultz and colleagues showed that dopamine neurons responded not simply to rewards themselves, but to differences between expected and received rewards.
When a reward was better than expected, activity increased.
When an expected reward failed to arrive, activity decreased.
As learning occurred, the response shifted towards the cue predicting the reward.
This is known as reward prediction error.
The brain learns from the gap between expectation and outcome.
Dopamine should therefore not be described simply as a pleasure chemical.
Nor can we responsibly claim that every swipe, like or notification produces a specific measurable dose.
Most studies of ordinary social-media behaviour do not directly measure dopamine release.
The popular explanation often runs far ahead of the evidence.
But the underlying principle of reward learning remains relevant.
Behaviour changes according to what follows it.
The reward does not need to occur every time.
Uncertainty can make the next outcome especially informative.
Social Feedback Is a Reward
Human beings care what other people think.
Again, technology did not invent this.
Reputation mattered in small communities long before written history.
Acceptance could determine access to cooperation, protection, relationships and resources.
Rejection carried consequences.
Praise felt good.
Humiliation hurt.
A smile communicated approval.
Silence communicated uncertainty.
Social media transformed those ambiguous signals into visible measurements.
Likes.
Shares.
Comments.
Views.
Followers.
Reactions.
Approval became countable.
A person no longer had to wonder vaguely whether others appreciated something they had created.
The platform could display a number.
That number could be compared with yesterday.
With another post.
With another person.
With an expectation.
With what someone else received.
Researchers studying adolescent responses to simulated Instagram posts found that participants were more likely to like images that had already received many likes than images showing few likes.
Viewing highly liked images was also associated with greater activity in brain regions implicated in reward processing, social cognition, imitation and attention. The effect occurred for both neutral images and images depicting risky behaviour.
The study does not prove that likes control adolescents.
The sample was limited.
The simulated environment could not reproduce every aspect of real social-media use.
Brain activation cannot be translated directly into a simple story about behaviour.
But the findings demonstrate something important.
Visible social endorsement changes how content is perceived.
The number does not merely report popularity.
It can contribute to it.
The Reward Changes Future Behaviour
The most persuasive evidence does not come from metaphors about slot machines.
It comes from observing what people actually do after receiving social rewards.
A 2021 study analysed more than one million posts from over four thousand users across several social platforms.
The researchers used computational models derived from reinforcement-learning theory.
They found that posting behaviour changed in ways consistent with reward learning.
People adjusted the timing of future posts according to the rate of social rewards they had previously received.
An additional controlled online experiment found that manipulating the rate of social rewards causally influenced posting behaviour in the direction predicted by the model.
The study was careful about its limits.
It did not prove that social-media use was addictive.
It did not claim that reward seeking explained every reason people posted.
People also communicate, express identity, maintain relationships, exchange information and participate in communities.
The research did show that social feedback contributes to behaviour through learning mechanisms that existed long before social media.
People received more approval.
Their behaviour adjusted.
The system did not create the underlying learning process.
It placed that process inside a measurable loop.
Uncertainty Keeps the Loop Open
Imagine posting a photograph and knowing exactly what will happen.
The same three people will respond.
At the same time.
With the same words.
The experience would quickly become predictable.
Social feedback is rarely like that.
You do not know who has seen the post.
Who will respond.
How quickly.
Whether the number will be higher or lower than last time.
Whether someone important will notice.
Whether approval will continue arriving.
The uncertainty can keep the interaction psychologically open.
Perhaps one more check will reveal something new.
Perhaps the number has changed.
Perhaps a response has arrived.
This is commonly compared with variable reinforcement schedules used in behavioural psychology and gambling.
The comparison can be useful.
It can also be overstated.
A slot machine is explicitly designed around monetary wagers and randomised outcomes.
A social network contains real people, relationships, information and communication.
A like from a friend can possess genuine emotional meaning.
The products are not identical.
The more defensible claim is narrower.
Unpredictable rewards can maintain repeated checking because the outcome remains unresolved.
The person is not returning solely for a guaranteed reward.
They are returning to discover whether a reward exists.
Habit Design Became an Industry Method
The principles were not confined to academic laboratories.
They were translated into practical frameworks for product designers.
Nir Eyal's Hook Model, developed from patterns he observed in the video-gaming and online-advertising industries, described four stages through which habit-forming technologies could encourage repeated use:
- trigger;
- action;
- variable reward;
- investment.
The external trigger brings the person back.
The action is made as easy as possible.
A variable reward provides an uncertain outcome.
The user then contributes something that increases the value of returning, such as time, data, content, effort, social connections or reputation.
Eyal argued that repeated cycles could eventually connect product use to internal triggers.
The person would no longer need to receive an email or notification.
A feeling could become the prompt.
Boredom.
Loneliness.
Uncertainty.
Discomfort.
Fear of missing out.
The product becomes the learned response.
That is a profound commercial achievement.
Advertising is expensive.
A habit brings the user back without requiring the company to purchase their attention again.
The person's own emotional state begins prompting the behaviour.
This does not mean companies can mechanically install habits in everyone.
People differ.
Contexts differ.
Products fail.
Users leave.
The model is a design framework, not a guaranteed formula.
Its significance lies in the objective it states openly.
Repeated product use can be deliberately designed.
From External Trigger to Internal Reflex
At the beginning, a person may need reminding.
An email tells them someone has joined.
A message announces that a photograph has received a response.
A notification says new content is available.
The prompt creates the action.
Over time, the external reminder may become unnecessary.
The person reaches for the device during a pause in conversation.
While waiting for a kettle to boil.
At a traffic light.
In a queue.
During a difficult piece of work.
After receiving an awkward email.
Before getting out of bed.
The environment no longer needs to interrupt them.
They interrupt themselves.
This is the point at which the relationship becomes difficult to recognise.
Nothing appears to be forcing the behaviour.
The application has not opened itself.
The person chose to open it.
But choices can be learned.
A repeated response becomes easier.
The cue becomes less visible.
The behaviour begins before conscious intention catches up.
The fact that a behaviour is voluntary does not mean it has not been shaped.
Investment Makes Leaving More Expensive
A useful product becomes more valuable as someone uses it.
A music service learns what they enjoy.
A photo library stores years of memories.
A social network contains friends and professional contacts.
A messaging platform becomes the place where family conversations happen.
A game contains progress, achievements and possessions.
A creative platform stores work and audiences.
This investment is not inherently exploitative.
It is often evidence of genuine utility.
The product improves because the person has contributed something.
But investment also creates attachment.
Leaving means losing more than access to an application.
It may mean losing history.
Status.
Connections.
Convenience.
An audience.
A record of identity.
Years of accumulated effort.
Eyal's description of the investment stage explicitly includes contributions of time, data, effort, social capital and money that improve the service and increase the likelihood of another cycle of use.
The more someone invests, the less attractive an alternative may feel.
Habit and switching cost reinforce one another.
The product is easier to return to because it already contains part of the user's life.
Social Proof Was Turned Into Interface
People often look to others when deciding what is safe, desirable or true.
A crowded restaurant appears more appealing than an empty one.
A long queue suggests something may be worth waiting for.
A recommendation from a trusted person carries more weight than an anonymous advert.
Digital platforms made social proof visible everywhere.
View counts.
Follower counts.
Ratings.
Trending lists.
Popular searches.
Most-liked comments.
The number of people currently watching.
These signals can be useful.
They help people navigate abundance.
When millions of options exist, evidence of what others found valuable can reduce the effort of choosing.
The same signals can create conformity.
An idea appears credible because it is popular.
A product seems desirable because others bought it.
A person appears authoritative because many people follow them.
Popularity becomes a shortcut for judgement.
The system does not need to tell someone what to think.
It can show what everybody else appears to think.
The Product Learns Which Prompt Works
Early persuasion relied heavily upon general assumptions.
Red attracts attention.
Scarcity creates urgency.
People respond to social proof.
Modern digital systems can test those assumptions continuously.
Does this wording produce more opens?
Does that timing increase returns?
Does a photograph outperform text?
Does showing the sender's name increase the response rate?
Does the red badge work better with a number?
Do people return more frequently when told that something is waiting?
Does removing one step increase completion?
The company does not need one theory of human behaviour to be universally correct.
It can test variations and retain whichever one changes behaviour most effectively within a particular population.
This is where the business model from the previous essay becomes essential.
A persuasive feature may generate more engagement.
More engagement may produce more advertising opportunities, data, transactions or retention.
The commercial system rewards the design that works.
At scale, the product can become increasingly persuasive through thousands of small improvements.
No designer needs a complete understanding of the human mind.
The metrics reveal which arrangement produced the behaviour.
The System Does Not Need to Understand Why
Suppose a notification sent at 7.15 pm produces more openings than one sent at 6.45 pm.
The company may not know why.
Perhaps dinner has finished.
Perhaps people are tired.
Perhaps they are more likely to feel lonely.
Perhaps the later time coincides with television viewing.
The reason may not matter.
The result can still be used.
This is a significant shift from traditional psychology.
A researcher tries to understand the mechanism.
A commercial system may need only to improve the outcome.
If one version reliably increases behaviour, it can be selected even when the psychological explanation remains uncertain.
The platform learns which button, prompt, image or sequence works.
It may not understand the human experience beneath the number.
Optimisation can proceed without empathy.
When Design Becomes Manipulation
Not every attempt to influence behaviour is manipulation.
A satnav prompts a driver to turn.
A fitness application encourages movement.
A language application reminds someone to practise.
A bank warns a customer before they become overdrawn.
A medication application helps someone follow a treatment plan.
These systems use psychological principles.
They may rely upon prompts, rewards, progress indicators and reduced friction.
The question is not whether behaviour is being influenced.
The question is whether the influence supports the person's goals.
Persuasion becomes more ethically troubling when:
- the real objective is hidden;
- the design exploits a known vulnerability;
- stopping is intentionally made difficult;
- consent is confused or obstructed;
- the short-term behaviour benefits the company while imposing a foreseeable cost upon the user;
- the person would probably reject the influence if they understood how it operated;
- children or vulnerable users are exposed without meaningful protection.
The Federal Trade Commission describes dark patterns as design practices capable of obscuring, subverting or impairing consumer choice.
Its 2022 report documented tactics including disguised advertising, obstructive cancellation processes, hidden terms and interfaces designed to steer people towards surrendering more personal information.
The FTC also noted that digital companies can collect data, conduct experiments and identify which patterns are most effective.
The difference between helpful design and manipulation cannot be reduced to one visual feature.
The same button can serve different purposes.
The ethical test lies in alignment.
Whose desired behaviour is being made easier?
Who benefits?
Who bears the cost?
Would the person still agree if the design's purpose were fully visible?
Ethical Concerns Were Not Discovered Yesterday
It would be inaccurate to suggest that nobody anticipated the ethical risks of persuasive technology.
Stanford's own Behaviour Design Lab records that ethical questions were being taught alongside persuasive technology from the late 1990s.
Fogg published work on the ethics of persuasive technology in 1998.
The lab organised an ethics panel in 1999.
Further work followed in 2002.
In 2006, the lab produced material warning regulators about problematic uses of persuasive systems and persuasion profiles.
This does not prove that every technology executive knew what modern platforms would become.
Nor does it prove that academic warnings predicted every present-day harm.
It establishes that the ethical tension was visible early.
The ability to influence behaviour through computers was recognised as powerful.
The possibility of misuse was recognised too.
The industry cannot credibly claim that persuasive design arrived with no accompanying questions about responsibility.
The questions existed.
The commercial incentives answered many of them in practice.
The Same Psychology Can Be Used for Good
A fair investigation must resist turning every behavioural technique into evidence of wrongdoing.
Prompts help people remember medication.
Progress bars help them finish difficult tasks.
Reduced friction makes services accessible.
Social encouragement can support exercise.
Streaks can help someone maintain a beneficial routine.
Personalisation can improve education.
Rewards can reinforce rehabilitation, saving and healthy behaviour.
Stanford's lab now emphasises behaviour design aimed at helping people achieve positive changes they already want, including health-related behaviours.
The techniques are not moral agents.
A hammer can build a home or damage one.
The fact that psychological knowledge can be misused does not make ignorance preferable.
The ethical question is whether the technology strengthens or weakens the user's agency.
Does it help someone do what they have consciously decided matters?
Or does it teach them to act before deciding?
Habit Is Not the Same as Addiction
The language surrounding digital products often becomes careless.
People say everyone is addicted to their phone.
Every platform is compared with a drug.
Every notification becomes a dopamine hit.
Every repeated behaviour is treated as a clinical disorder.
That exaggeration weakens the argument.
Habits are automatic behaviours triggered by familiar contexts.
Many are harmless.
Some are useful.
Addiction is a more serious concept involving impaired control, continued behaviour despite harm and clinically significant disruption.
Not everyone who checks social media frequently is addicted.
Not every child who enjoys a game has a disorder.
High use alone does not establish pathology.
The 2021 reward-learning study explicitly stated that its findings did not determine whether intense social-media use was maladaptive or addictive.
We do not need to diagnose billions of people to recognise a problem.
A behaviour can be commercially engineered, difficult to regulate and costly to attention without meeting the threshold of addiction.
That is the more important point.
The system does not need to make everyone clinically dependent.
It needs to increase average use slightly across an enormous population.
A few additional minutes per person can create millions of additional hours of engagement.
The commercial objective can succeed long before a doctor would diagnose anything.
Individual Vulnerability Is Uneven
The same design does not affect everyone equally.
Some people can ignore notifications easily.
Others experience intense pressure to respond.
Some enjoy social feedback without relying upon it.
Others connect approval to self-worth.
Some use short-form video deliberately and stop when intended.
Others lose far more time than they expected.
Age matters.
Temperament matters.
Mental health matters.
Social circumstances matter.
The purpose of the product matters.
The type of content matters.
The surrounding family and cultural environment matter.
Research into adolescents' habitual social-media checking has found associations with developmental changes in sensitivity to social rewards and punishments, although such findings do not establish that checking alone caused the observed differences.
This variability creates a difficult ethical problem.
A feature that feels harmless to the average adult may affect a vulnerable child differently.
A system optimised for aggregate engagement can conceal the experiences of smaller groups carrying greater costs.
The average metric rises.
The individual harm disappears inside it.
The User Is Not Powerless
Recognising design influence does not mean denying human agency.
People disable notifications.
Delete applications.
Set limits.
Choose subscription services.
Use feeds deliberately.
Teach children how recommendation systems work.
Leave devices outside bedrooms.
Reject prompts.
Build alternative habits.
Persuasive design influences behaviour.
It does not determine it completely.
The danger lies at either extreme.
One extreme claims technology controls people entirely.
The other insists every action is a free and unaffected personal choice.
Neither reflects reality.
Human beings make choices inside environments.
The environment changes which choices are easy, visible, immediate and rewarding.
Agency remains.
The cost of exercising it changes.
A person can stop watching when another video begins automatically.
They must notice the transition and choose to interrupt it.
A person can ignore a notification.
They must absorb the initial distraction.
A person can leave a platform containing years of relationships and memories.
The option exists.
It is not costless.
Responsibility Cannot Rest Entirely With the User
Technology companies often provide controls.
Notification settings.
Time reminders.
Parental supervision.
Muted words.
Hidden like counts.
Recommendation feedback.
Autoplay switches.
These tools matter.
YouTube now disables Autoplay by default for users aged thirteen to seventeen.
Such changes demonstrate that products are not fixed.
Design choices can be reversed.
Defaults can change.
Friction can be reintroduced.
But responsibility cannot end with making a protective setting available somewhere inside a menu.
Defaults matter because most people do not redesign every product they use.
The company builds the initial environment.
The user enters it.
If the most commercially advantageous setting is enabled automatically while the more protective option requires knowledge and effort to find, the architecture still favours the company.
A fair system should not require every user to become an expert in behavioural science before they can use an ordinary communication tool safely.
The Most Successful Influence Becomes Invisible
A clumsy attempt at persuasion is easy to resist.
An advert that shouts too loudly becomes irritating.
A notification sent too often is disabled.
A reward that feels artificial loses meaning.
Effective influence blends into the experience.
The product feels intuitive.
The behaviour feels self-chosen.
The prompt arrives at the right moment.
The action requires almost no effort.
The reward feels socially or personally meaningful.
The person contributes enough to make returning more valuable.
Eventually, the product stops feeling like something they use.
It becomes part of how they respond to life.
Bored?
Open it.
Lonely?
Open it.
Uncertain?
Open it.
Avoiding something difficult?
Open it.
Waiting?
Open it.
The company no longer needs to capture attention from outside.
The habit delivers it from within.
The Moral Question Is Not Whether It Works
Persuasive design works imperfectly.
Not everyone responds.
Not every technique succeeds.
Not every platform creates strong habits.
But enough of it works often enough to support some of the largest businesses in history.
That is no longer the interesting question.
The important question is what obligation follows from possessing that capability.
If a company can learn which prompts interrupt people most effectively, should every effective prompt be used?
If social approval increases engagement, should every interaction be quantified publicly?
If removing a stopping point increases consumption, should the stopping point disappear?
If a child is more sensitive to social reward, should the same persuasive architecture apply?
If a product begins connecting boredom or loneliness with automatic use, does the company have a duty to weaken that association even when it reduces engagement?
These are not anti-technology questions.
They are design questions.
Business questions.
Moral questions.
They concern what should happen when the most profitable behaviour is not necessarily the most beneficial one.
From Understanding Behaviour to Exploiting It
The word exploit needs careful use.
To exploit something can simply mean to make effective use of it.
Engineers exploit physical principles.
Doctors use biological knowledge.
Teachers use an understanding of memory.
But the word also describes taking unfair advantage of someone or something.
The distinction depends upon purpose, transparency and consequence.
A language application using reminders to help someone meet a goal they consciously chose is using psychology.
A platform discovering that anxiety-producing prompts increase returns and continuing to use them despite foreseeable harm would be exploiting psychology.
The mechanism may be similar.
The alignment is not.
This is why evidence of intent cannot be reduced to the existence of a persuasive feature.
A notification is not proof of misconduct.
A reward is not proof of manipulation.
Personalisation is not proof of exploitation.
The case must examine what the company knew, what outcome it optimised, which alternatives were available and what happened when user welfare conflicted with commercial performance.
That evidence will become increasingly important as this series progresses.
They Learned Faster Than We Did
There is another imbalance we have not yet fully considered.
Individuals learn from experience.
Perhaps someone notices that notifications interrupt their work.
They switch them off.
Perhaps a parent recognises that Autoplay extends viewing.
They disable it.
Perhaps a teenager realises that visible like counts affect their mood.
They adjust how they use the platform.
Human learning takes time.
The product is learning too.
It observes millions of responses.
It tests alternatives.
It updates.
It identifies patterns across populations.
It improves faster than any individual can study it.
The person may understand one technique just as the design changes.
The red badge becomes a personalised message.
The general recommendation becomes a prediction.
The external prompt becomes an internal habit.
The next essay will examine that acceleration directly.
Because the most powerful development in the attention economy is not that companies learned to use human psychology.
It is that machines can now learn which version of that psychology works best for each person.
Final Thought
Technology companies did not invent the desire to belong.
They quantified approval.
They did not invent curiosity.
They built feeds capable of supplying endless novelty.
They did not invent learning through rewards.
They placed social feedback inside measurable loops.
They did not invent boredom.
They made relief available within seconds.
They did not invent habit.
They developed frameworks for building products that become habitual.
They did not invent human weakness.
They learned which designs required the least conscious resistance.
None of this proves that every designer intended harm.
Many persuasive technologies help people live healthier, safer and more capable lives.
The same psychological principles can strengthen human agency.
But once those principles became connected to a business model that profited from repeated engagement, the objective changed.
Understanding human behaviour was no longer only a way to make technology useful.
It became a way to make leaving less likely.
The attention economy did not need to redesign the human mind.
Evolution had already provided curiosity, uncertainty, social need, reward learning and a preference for ease.
The industry simply learned how those systems worked.
Then it built products around them.
Then it measured the results.
The question is not whether the psychology is real.
The evidence shows that it is.
The question is whether companies should be allowed to use that knowledge without accepting responsibility for what their designs teach billions of human beings to do automatically.
Sources & Research Gaps
Principal Sources
BJ Fogg and the Fogg Behaviour Model
Fogg's model proposes that motivation, ability and a prompt must converge for a behaviour to occur.
The model is widely used within behaviour design and persuasive technology. It is not inherently manipulative and can be applied to beneficial or harmful objectives.
Stanford Behaviour Design Lab, Ethics of Persuasive Technology
Stanford records that ethical concerns accompanied the study of persuasive technology from its early development.
The lab taught ethics from the late 1990s, published work on the subject and warned policymakers about possible harmful applications.
Nir Eyal, The Hook Model
Eyal's published explanation of the Hook Model identifies four components of habit-forming technology:
- trigger;
- action;
- variable reward;
- investment.
His work is used here as direct evidence of how behavioural principles were translated into commercial product-design methods, rather than as independent scientific proof that every product using the model successfully creates habits.
Schultz, Dayan and Montague, Reward Prediction Error
The 1997 Science paper provided foundational evidence that dopamine-neuron activity can encode differences between expected and received rewards.
The research is used to correct simplistic descriptions of dopamine as merely a pleasure chemical.
Lindström and Colleagues, Reward Learning and Social-Media Engagement
The researchers analysed more than one million posts from over four thousand users across several social-media platforms.
They found posting patterns consistent with reinforcement-learning principles and experimentally demonstrated that changing the rate of social rewards influenced future posting behaviour.
The study did not establish social-media addiction or directly measure dopamine.
Sherman and Colleagues, The Power of the Like in Adolescence
This study used a simulated Instagram environment and functional magnetic resonance imaging.
Adolescents were more likely to endorse images already displaying many likes, and highly liked images were associated with activity in regions connected to reward processing, social cognition, imitation and attention.
Stothart, Mitchum and Yehnert, The Attentional Cost of Notifications
The study found that receiving mobile-phone notifications could disrupt performance on an attention-demanding task even when participants did not interact with the device.
Whiting and Murdock, Notification Alerts Across Age Groups
The researchers examined the effects of notification sounds among adolescents, young adults and middle-aged adults.
The strongest performance effects were observed among adolescents.
YouTube Autoplay Documentation
YouTube describes Autoplay as reducing the need to decide what to watch next by automatically beginning a related video.
Autoplay is now disabled by default for users aged thirteen to seventeen and can be controlled by users or parents.
Federal Trade Commission, Bringing Dark Patterns to Light
The FTC documented interface practices capable of obscuring or impairing consumer choice, including disguised adverts, obstructive cancellation systems, hidden terms and privacy interfaces designed to steer users towards greater data disclosure.
The regulator also highlighted the ability of companies to test and optimise such designs at scale.
Research Gaps and Limitations
Persuasive design is an umbrella term covering many techniques and purposes.
It should not be treated as synonymous with manipulation.
The Fogg Behaviour Model provides a practical framework for analysing behaviour. It does not prove that any specific digital feature caused a particular user action.
Nir Eyal's Hook Model is an industry design framework rather than a controlled scientific theory of all product use.
The term variable reward is frequently used loosely. Social feedback is uncertain, but social platforms are not identical to gambling machines.
Comparisons with slot machines may illuminate some reinforcement mechanisms while obscuring the social, informational and relational value present in digital platforms.
Dopamine is involved in multiple functions, including learning, movement, attention and motivation.
It is inaccurate to claim that every notification, swipe or like produces a discrete, equivalent dopamine hit.
Most studies of ordinary social-media behaviour do not directly measure dopamine release.
Brain-imaging studies identify patterns of activity associated with particular tasks. They do not reveal a person's complete mental state or prove that one feature controls behaviour.
The adolescent Instagram study used a simulated environment and a limited sample. Its findings should not be generalised to every adolescent or every platform.
The reward-learning study focused primarily on posting and social feedback. It did not examine every form of social-media consumption, recommendation or checking.
Notification studies often take place under controlled conditions and may not capture how people adapt to alerts in everyday life.
Product features change frequently.
Defaults, safeguards and controls differ by age, platform, country, account type and device.
Habitual use is not equivalent to addiction.
High engagement does not automatically establish psychological harm or impaired control.
Further research is needed into:
- which persuasive-design techniques produce durable behaviour change;
- how effects differ by age, temperament, neurodivergence and mental-health status;
- the interaction between notifications, recommendations and self-initiated checking;
- whether removing visible social metrics changes wellbeing or behaviour;
- how internal emotional triggers become associated with product use;
- how often companies test for long-term user benefit rather than short-term engagement;
- whether protective friction can be introduced without reducing genuine usefulness;
- which safeguards are most effective for children and adolescents;
- how artificial intelligence personalises persuasive design;
- how to distinguish ethical persuasion from manipulation through measurable standards;
- the extent to which users understand the behavioural objectives built into digital products.