The research behind The Attention Experiment

On this page

• Where my research comes from
• Noticing and naming
• Why phones hold our attention
• The body
• Young people, bans and restrictions
• Taking a break
• AI and thinking
• What we don't know yet
• My evidence to Parliament

What it foundWhen people named the emotion they were feeling, activity in the brain's reactive centre (the amygdala) went down, and activity in a part of the prefrontal cortex involved in regulating emotion went up.

Why it matters hereIt's one reason I believe naming "I'm in a Scroll Loop" matters. Putting what's happening into words brings it into conscious awareness, where choice becomes possible.

LimitsThe study was about naming emotions, not habits or phone use, and it was a small brain imaging study. Applying it to the Scroll Loop is my interpretation.

What it foundReviewing a large body of studies, the EEF rates metacognition and self-regulation as a high-impact, low-cost approach. It recommends explicitly teaching metacognitive strategies, teachers modelling their own thinking out loud, and building metacognitive talk into the classroom.

Why it matters hereNoticing is metacognition, choosing is self-regulation. In schools, a teacher saying "I'm in a Scroll Loop" and putting their phone down is modelling their thinking out loud.

LimitsThe EEF's evidence is about learning in the classroom, not about how we use our devices. I believe the same principles carry over, but that hasn't been tested directly.

What it foundIn a brain imaging study, personalised TikTok videos, chosen by the algorithm for each person, activated the default mode network (linked to daydreaming, memory and our sense of self) and a key part of the brain's reward system, more than videos that weren't personalised.

Why it matters hereIt helps explain the Loop: why time slips away when we scroll, and why it can feel like drifting into a dream.

LimitsA small study with a group of university students, watching videos in a scanner rather than scrolling in everyday life.

What it foundFacebook altered the news feeds of nearly 700,000 people. When people saw fewer positive posts, they went on to post fewer positive things and more negative ones, and the reverse.

Why it matters hereIt shows that the emotions in our feeds can spread to us, which matters when we're scrolling through frightening news.

LimitsThe effects were small, and the study was widely criticised for experimenting on people without their knowledge.

What it foundWhen researchers tracked people's actual phone use, people used their phones about twice as much as they estimated.

Why it matters hereSo much of our phone use happens without us noticing, which is exactly what the method works on.

LimitsA small study, and phone habits have changed a lot since 2015.

What it foundPeople in the UK are on track to spend nearly five years of their waking lives on their phones unintentionally, and four in ten say willpower alone isn't enough to change their phone behaviour.

Why it matters hereIt puts a number on the unconscious use that the method aims to reduce, and supports the idea that willpower isn't the answer.

LimitsA commercial survey based on what people report, rather than tracked data.

What it foundIn a study of 161 young adults, people who tended to distract themselves from uncomfortable body sensations, rather than staying with them, were more likely to use the internet in problematic ways. People more prone to worry and emotional reactivity were more at risk too.

Why it matters hereIt supports working through the body as well as the mind: noticing the breath, the tight shoulders, the pull, and staying with it for a moment rather than reaching for the phone.

LimitsAn early, small study that shows links rather than cause, with young adults in one country.

What it foundIn a survey of 2,299 children aged 11 to 17, nearly two in five said they had got around an online age check at least once, and half of those learned how from friends. Most harmful content children saw was met passively while scrolling, rather than searched for.

Why it matters hereIt suggests access restrictions alone won't be enough, and that the design of the feed itself matters.

LimitsBased on what children and parents report.

What it foundFamilies tried a daily time limit, an overnight curfew or removing social media apps. Teenagers reported better sleep, mood, focus and family time, alongside workarounds and a difficult first week. Lasting change was more likely when restrictions came with education and practical guidance, and felt like a shared norm.

Why it matters hereIt backs the idea that understanding and shared language need to go alongside any restrictions.

LimitsA qualitative study with a relatively small group, exploring experiences rather than measuring effects.

What it foundAround six in ten children aged 8 to 17 know that apps use algorithms to decide what to show them, rising with age. Among 13 to 17-year-olds who use social media and messaging apps, 72% say they help them feel closer to friends.

Why it matters hereKnowing algorithms exist isn't the same as understanding what they do to us, and phones matter to young people's friendships. Both shape how I work with schools.

LimitsBased on what children report.

What it foundYoung adults who took a one-week break from social media reported improvements in symptoms of depression, anxiety and poor sleep.

Why it matters hereEven short breaks can help, which is why the experiment starts with seven days.

LimitsPeople chose to take part, and the break was short, so we don't know how long the benefits last.

What it foundPeople writing essays with ChatGPT showed the lowest brain engagement, and most struggled to recall what they had just written. When later asked to write without it, their brains stayed less engaged.

Why it matters hereIt raises questions about what happens when we hand over our thinking, the pull I call the Prompt Loop.

LimitsA small study that hasn't yet been peer reviewed.

What it foundPeople who used AI tools more often tended to score lower on critical thinking. Handing thinking over to AI (cognitive offloading) was the key link, and 17 to 25-year-olds were the most affected.

Why it matters hereIt's why I teach noticing when AI is replacing our thinking rather than supporting it, especially with young people.

LimitsIt shows a link, not cause, and relies largely on questionnaires.

What it foundStudents supported by ChatGPT improved their essay scores in the short term, but engaged less in the thinking and self-regulation that deeper learning depends on. The researchers called this metacognitive laziness.

Why it matters hereIt connects AI use directly to metacognition, the skill at the heart of the method.

LimitsA lab-based study of one writing task, so it can't tell us about long-term learning.

What it foundReviewing many studies, the researchers found people tend to understand what they read better on paper than on screens.

Why it matters hereIt's worth knowing as more learning moves onto devices.

LimitsThe differences were modest and depend on the type of text and task.

What it foundIn research with young people across the UK, most worried AI would make people "intellectually lazy" and stop them thinking for themselves. They were clear that AI should complement human skills, not replace them. Some described chatbots as comforting, while worrying they could be damaging.

Why it matters hereYoung people are more aware of the risks than we often assume. My job is to give them a practical way to act on that awareness.

LimitsQualitative research, exploring views in depth rather than measuring how common they are.

What it foundNearly three in four US teenagers had used AI companions, and around half used them regularly.

Why it matters hereAI companions are already part of many young people's lives, often without them understanding how they're designed.

LimitsA survey of US teenagers, so UK figures may differ.