Pay Attention by David Pakman — Are Algorithms Rewiring Our Brains and Destroying Our Ability to Recognise Truth?

Written By James Taylor
Pay Attention: How Outrage Wins and Careful Thinking Loses Visibility

Why Attention and Truth Are Coming Apart

How Algorithms Shape What We Believe

You open your phone to check one thing. A few minutes later, you have watched an argument, absorbed a frightening claim and seen someone insist that everything you thought you knew was wrong. You remember the emotion. You may struggle to remember the evidence.

David Pakman’s Pay Attention: How the Algorithms and Media Wars Are Suppressing Truth and Rewiring Your Brain examines the forces behind that experience. Its central warning concerns a media environment where keeping people engaged can take priority over helping them understand.

The question deserves a careful answer. Algorithms can influence what reaches us, and digital habits can interfere with attention. However, proving that they permanently damage the brain or destroy our capacity to recognise truth requires much stronger evidence. The more immediate danger is that we keep exercising that capacity under conditions that reward something else.

What Pay Attention Is Warning About

Pakman is a political commentator and an online broadcaster. In his public explanation of the book, he describes the pressure involved in choosing titles, thumbnails and clips within systems that reward provocative material. He acknowledges participating in those decisions himself. His warning comes partly from operating inside the environment he criticises.

That perspective matters because the problem begins before a viewer encounters a false claim. A creator has already decided which subject to cover, which words to emphasise and what emotional promise will persuade someone to stop scrolling. By the time an audience starts assessing the content, competition for attention has helped shape its presentation.

Consider an imaginary interview about a complicated policy. One accurate title might promise an explanation of its costs and trade-offs. Another might suggest that a politician has been exposed or humiliated. Both can point towards the same recording, yet they prepare viewers for different experiences. The first invites evaluation. The second invites a verdict.

This example illustrates the book’s broader concern without assuming that every dramatic title is dishonest. Good journalism needs to attract readers. The danger appears when the emotional promise becomes stronger than the evidence, or when the search for a successful format repeatedly narrows what creators are willing to say.

The useful question is therefore what happens to information when visibility depends heavily on our reaction to it. A system can circulate accurate facts while encouraging a misleading understanding of their importance. It can show genuine disagreements while selecting the most aggressive representatives of each side. Distortion does not always require a fabricated sentence.

This article discusses that verified central premise and tests it against independent research. It does not reconstruct the book’s chapters or attribute unverified examples and recommendations to Pakman. The distinction allows the warning to be examined seriously without treating its strongest language as a scientific conclusion.

How a Feed Turns Behaviour Into Recommendations

A recommendation system has a practical task: select a small amount of material from a much larger pool. It uses available signals to predict what may interest a particular person. Those signals can include viewing behaviour, likes, shares and other interactions, although the details differ between platforms and change over time.

TikTok’s published explanation, for example, describes interactions and video information as recommendation inputs. It identifies completing a longer video as a potentially stronger signal than a shared location. That tells us something about the platform’s stated approach. It does not reveal every current ranking rule or establish how other services work.

The underlying difficulty is interpreting behaviour. Imagine watching an alarming video because you want to identify its mistake. Someone else watches because they believe it. A third watches because they find the speaker absurd. From a limited behavioural signal, those experiences can look more similar than they feel.

A person’s immediate response also differs from what they value afterwards. You can spend time following an argument and later wish you had read something useful instead. You can repeatedly inspect a claim because it worries you, while wanting fewer reminders of it. Captured attention is an imperfect measure of a worthwhile experience.

Research published in 2025 audited Twitter’s engagement-based ranking against a chronological baseline. It found amplification of emotionally charged and politically hostile material. Users’ stated preferences did not simply match the political tweets the algorithm selected. The study also identified trade-offs in an alternative based on stated preferences, including the possibility of reinforcing existing views.

The practical implication is modest but important: recommendation quality depends on what a system is trying to predict and how success is measured. More responses do not automatically mean more satisfaction, greater understanding or better information. A feed can perform well against one objective while disappointing the person using it.

Why Outrage Becomes a Successful Format

Political hostility is especially useful for illustrating the gap between engagement and understanding. A claim about an opponent can offer several rewards at once: an apparent explanation, a target for blame and an opportunity to signal loyalty. Sharing it may communicate membership of a group as much as belief in its details.

A 2021 study analysed more than 2.7 million posts from news accounts and members of the United States Congress. Posts referring to the political opposing group were shared or retweeted roughly twice as often as those referring to the same group. This was a pattern within the studied material, rather than a universal rule about every audience.

Another research project combined observational work with behavioural experiments to examine outrage expression. It found that positive social feedback increased the likelihood of later expressions of outrage, while the norms of a person’s network also influenced their behaviour. The outcome concerned what people expressed online; it should not be mistaken for a direct measurement of their private emotional lives.

Taken together, those findings suggest a feedback problem. Material receives attention, its author notices, and similar expression becomes more attractive. Other people see what appears successful and may adopt the format. A public conversation can become more confrontational through many ordinary adjustments, without everyone deciding consciously to make it worse.

Outrage itself is sometimes justified. Anger at corruption, cruelty or abuse can lead people to investigate and act. A demand for a calmer internet should never become a demand that victims describe serious harm politely enough to satisfy comfortable observers.

The relevant distinction is whether emotion helps people examine the underlying event or substitutes for that examination. An accusation can open an investigation. It can also close one prematurely by making uncertainty feel like disloyalty. The same emotional force can serve accountability or obstruct it, depending on the evidence and what follows.

For a reader, this creates a useful habit: identify the claim underneath the performance. What supposedly happened? Who can establish it? Which part is documented, and which part is someone’s interpretation? A dramatic delivery becomes less controlling when its factual content is written down in plain language.

Does “Rewiring the Brain” Mean Damage?

The phrase “rewiring your brain” sounds more precise than it usually is. Learning, practising a skill and forming habits all involve the brain. The existence of change does not establish injury. To assess a worrying claim, we need to know what was measured, how it changed and whether the effect lasted.

A study might measure performance on an attention task. Another might ask people how distracted they feel. A third might examine brain activity. These methods answer different questions. None automatically proves that ordinary scrolling permanently reduces intelligence, and one kind of result cannot simply be relabelled as another.

A 2025 systematic review and meta-analysis brought together 71 studies involving 98,299 participants. Greater short-form video use was associated with poorer cognitive outcomes, with attention and inhibitory control showing particularly strong associations among the measured domains. Inhibitory control concerns resisting an automatic or tempting response. The review also noted limited evidence for several other cognitive domains.

Those associations deserve attention, but they do not settle the direction of cause. People who already struggle to concentrate may gravitate towards short videos. Heavy viewing may also displace activities that support concentration. Different influences may operate together. Combining many studies improves the picture of an association without automatically resolving those questions.

The everyday concern is easier to state than the neurological claim. If checking a feed repeatedly interrupts reading, the reading has become harder in that setting. If a person spends their available evening moving between clips, they have less time for a sustained task. These are meaningful consequences even before permanent biological change has been demonstrated.

There is also a difference between capability and the conditions under which capability is used. Someone may still possess the ability to follow a long argument while rarely creating an opportunity to do so. That distinction matters because it changes the response: improving the environment may help, whereas a story of irreversible damage can encourage resignation.

The headline’s strongest question therefore remains only partly answered. Research supports concern about digital use and attention. It does not justify announcing that algorithms have universally destroyed the brain’s ability to judge reality. Precision makes the warning more useful because it identifies problems that can actually be examined.

Recognising Truth Requires More Than Seeing Information

Access to a claim is only the beginning of understanding it. A reader must identify what is being asserted, assess the evidence and place it in context. A short video can introduce an important fact, yet leave unanswered the questions that determine what the fact means.

Imagine a clip claiming that a public figure admitted wrongdoing. The decisive sentence may be genuine. Even so, the full exchange could show that the speaker was quoting an accusation, discussing a hypothetical situation or rejecting the premise. Checking the recording changes the task from reacting to a fragment to interpreting an event.

The problem is particularly acute when a post asks the viewer to make several decisions almost immediately. Is it authentic? Is it recent? Does it concern the place named in the caption? Does it prove the conclusion? Those are separate questions, although a persuasive presentation can make them feel like one obvious judgement.

Research led by Gordon Pennycook and colleagues offers an important distinction here. People’s assessments of headline accuracy did not neatly match their willingness to share headlines. Experiments, including a Twitter field intervention, found that directing attention towards accuracy could improve the quality of subsequently shared news. This supports an attentional explanation for some misinformation sharing.

The finding leaves room for other causes, including deliberate deception and strongly held partisan commitments. Its value is that it challenges an unnecessarily bleak assumption: every inaccurate share does not demonstrate that the person has lost the ability to recognise falsehood or ceased to care about truth.

Consider what sharing can accomplish socially. It can amuse friends, register disgust or show that someone belongs to a particular community. Accuracy may be important to the person in principle but receive little consideration during that moment. The feed has presented a social opportunity, and the user has responded to that opportunity.

A practical countermeasure is to change the immediate question. Before forwarding a claim, ask whether you would defend its factual accuracy if someone requested the evidence. That question is an editorial application of the research, rather than a guaranteed intervention. It reconnects the act of sharing with responsibility for what is shared.

When Familiarity Starts to Feel Like Evidence

Repeated exposure creates another difficulty. You see a claim in a video, encounter a paraphrase in a comment and later hear someone mention it in conversation. By the third encounter, the claim can feel established, even if all three versions ultimately depend on the same unsupported source.

Psychologists call the increased perceived credibility of repeated statements the illusory truth effect. Experiments by Lisa Fazio and colleagues showed that prior knowledge did not always prevent it. Participants sometimes relied on the ease of processing a repeated statement even when knowledge relevant to judging it was available.

This does not mean repetition can make everyone believe anything. It identifies a tendency in judgement, not an irresistible command. People still use knowledge, investigate claims and reject obvious mistakes. The important warning is that familiarity can contribute to credibility without providing new evidence.

An illustrative rumour shows the difference. Ten accounts might report that a celebrity has died, yet all ten could be repeating one unverified post. Counting the accounts makes the rumour appear corroborated. Tracing their information backwards reveals a single evidential thread. The quantity of repetition and the quantity of evidence are different.

This distinction applies to careful-looking material as well as obvious sensationalism. A precise figure, a formal graphic or a confident explanation can travel widely while preserving the same original error. Professional presentation tells us something about production. It does not settle whether the underlying claim was checked.

The useful response is source tracing. Find the earliest accessible report, document or recording that supposedly establishes the claim. Then check whether later accounts add independent evidence or merely refer to one another. Where the chain ends in uncertainty, repetition cannot repair it.

Readers should also distinguish an old event from a new development. A clip may be genuine and still misleading when presented as current. The factual question includes time: what happened then, what has changed since, and why is this material circulating now? Authentic content can acquire a false meaning through its framing.

The Algorithm Is Powerful, but It Is Not the Whole Explanation

One danger in discussing recommendation systems is making them responsible for every form of political division. People chose partisan publications, sought reassuring arguments and distrusted opponents before personalised feeds existed. Algorithms operate within those preferences and institutions; their influence needs to be established rather than assumed.

A major experiment during the 2020 United States election assigned consenting Facebook and Instagram users to chronological feeds instead of the default ranking. The change substantially reduced time spent on the platforms and altered what participants encountered. However, it did not significantly change measured political knowledge or several measures of polarisation during the three-month study.

The findings also complicate a familiar proposed solution. Chronological feeds increased exposure to political and untrustworthy content in that experiment. Removing one form of ranking did not automatically deliver a more reliable information environment. A feed’s results depend partly on the material available and the accounts supplying it.

This evidence limits sweeping claims about direct political transformation. It does not prove that every recommendation system is harmless, or establish what happens after years of use. Its immediate lesson is narrower: substantial changes in exposure and engagement need not produce corresponding short-term changes in measured political attitudes.

That distinction should guide our language. A study of which posts receive shares cannot alone establish how voters change their minds. An audit showing amplification cannot by itself establish permanent cognitive damage. Each piece of research illuminates a part of the problem, and connecting those parts requires care.

Personalisation also has real advantages. It can help someone discover a specialist explanation, an unfamiliar writer or a community otherwise difficult to find. The relevant choice is not whether recommendation should exist. It is how recommendations should be designed, assessed and placed under meaningful user control.

People have different reasons for using the same service. Someone seeking relaxation may welcome a stream of short entertainment. Someone researching a serious allegation needs reliable provenance and context. A design that serves the first purpose well can serve the second badly. Treating all engagement as equivalent hides that difference.

The Creator’s Dilemma: Making Careful Work Visible

Pakman’s account is particularly relevant to people publishing online. His public description of the book acknowledges that provocative framing can succeed where nuanced framing struggles. The broader dilemma is straightforward: careful work has little public value if nobody encounters it, yet attracting an audience can put pressure on its accuracy and tone.

Suppose a small publisher investigates a technical subject. The research involves documents, caveats and unresolved questions. The strongest factual conclusion may be less dramatic than the claim already circulating. A truthful headline must communicate why the explanation matters without pretending the uncertainty has disappeared.

An engaging opening can do that. It can present a concrete consequence, identify a real conflict or explain what a common claim gets wrong. It becomes misleading when it promises a certainty the article cannot supply, or uses a frightening implication that the evidence does not support.

This article’s own question illustrates the responsibility. “Destroying our ability to recognise truth” is a serious possibility to examine, not an established finding to announce. A publication that criticises sensational incentives while treating that phrase as proven would reproduce the behaviour it objects to.

There is a second pressure: an audience may reward a publisher for confirming its existing views. If the publisher later encounters inconvenient evidence, accuracy can threaten the relationship that has sustained the work. Independence from a corporate owner does not automatically mean independence from audience expectations.

One useful editorial standard is to make revision visible. Explain when a claim is corrected, preserve distinctions between reporting and interpretation, and show readers what evidence could change the conclusion. That approach asks the audience to value a method rather than permanent agreement with a personality.

Emotion still has a place. An account of suffering can communicate urgency; a vivid explanation can make an abstract problem understandable. The test is whether the emotional force remains proportionate to what is known. Good presentation should help the evidence reach people and remain comprehensible once it arrives.

Can Attention Improve When the Environment Changes?

A fatalistic account would leave readers with little to do beyond regretting their phone use. Experimental evidence offers a more constructive possibility. In a 2025 randomised study, researchers blocked mobile internet access on participants’ smartphones for two weeks while allowing calls and texts. Participants could still use the internet through other devices.

The study reported improvements in sustained attention, wellbeing and measured mental health outcomes. However, adherence was difficult: only around a quarter of the original participants met the study’s strict compliance criterion. Its results therefore support a specific intervention under studied conditions, rather than a promise that everyone will experience the same benefit.

The intervention changed an environment, not merely an intention. That is the useful principle for everyday application. Deciding to concentrate while leaving every interruption available asks a person to repeat the same choice continually. Removing some interruptions reduces the number of decisions required to protect a task.

For reading, that could mean placing the phone elsewhere and choosing a manageable period for one article or chapter. The purpose is to finish an argument before encountering another. For news, it might mean collecting stories at chosen times, then opening the full reports rather than treating their captions as sufficient.

These are practical suggestions, not a medical treatment or a claim about guaranteed neurological repair. Their value can be assessed through ordinary outcomes: whether a person completes the task, remembers its reasoning and feels less compelled to interrupt it. A useful change should help with something concrete.

For checking truth, a similarly modest routine can help. Establish the source, read beyond the headline and separate the factual claim from its interpretation. When the evidence is incomplete, postpone the verdict. You can continue following an event without announcing a conclusion each time another fragment appears.

That final step can be uncomfortable because public conversation often rewards instant positions. Yet uncertainty is sometimes the most accurate available response. Being able to say what remains unknown protects understanding more effectively than borrowing someone else’s confidence.

The Real Risk Is Letting Reaction Set the Standard

The strongest lesson from Pay Attention’s central premise is that the conditions of public thought matter. Information can be available while attention is continually directed towards its most provocative fragments. A person can care about accuracy while making sharing decisions organised around humour, anger or belonging.

The evidence does not establish that algorithms have made truth inaccessible or permanently destroyed our cognitive abilities. It does establish reasons to question the relationship between engagement, value and reliable judgement. Those distinctions give readers a firmer basis for responding than a general claim that everyone’s brain is broken.

There is a responsibility here for publishers and platforms as well as users. Publishers decide how strongly to frame a claim. Platforms decide which signals deserve emphasis. Readers decide when to pause, investigate or withhold a share. These responsibilities overlap, although no individual reader can redesign an entire information market.

A better standard would ask whether people leave an encounter able to explain what happened, what supports the account and what remains uncertain. That is a demanding aim because understanding often develops more slowly than reaction. It also provides a clearer measure of useful information than the intensity of a momentary response.

The next time a clip makes a claim feel undeniable, write down what it actually establishes. If the evidence supports less than the performance suggested, you have identified the gap. The ability to notice that gap is still worth protecting—and still available to exercise.

Sources

James Taylor
Previous

Dead Beat by Leigh Bardugo Explained: Dark Magic, Secret Societies and the Mysteries of Yale
Next

The Ship of Theseus: If Everything About You Changes, Are You Still You?