The Modern Human
Shop Modern Picks
The Evidence

What we're learning about life in a digital world

We distinguish association from correlation from causation, and we say when the evidence is thin. Most of it is thinner than the headlines suggest.

Psychology·Technology·Human behaviour
A passenger looking out of a window while others use their phones
Worth knowing this week

Total screen time remains a weak predictor of almost everything.

Large-dataset analyses repeatedly find associations between hours and wellbeing that are real but tiny. That isn't a reason to stop paying attention — it's a reason to measure something more informative: what the time replaces, and how it leaves you.

Screen time ≠ the whole story

What
content
Why
motivation
What it replaces
sleep · connection · movement · focus

The evidence is increasingly pointing toward context over raw hours.

How strong is the evidence?

Read the label first
●●●

Experimental

Can test causal effects under controlled conditions.

●●○

Correlational

Shows relationships, but not necessarily cause.

●○○

Association

A relationship exists, but many explanations may remain.

●○○

Mixed evidence

Studies point in different directions or effects vary substantially.

01

Attention Research

What it costs to be interrupted, and what actually restores you afterwards.

●●●Experimental

Interruptions cost more in resumption than in duration

Work on task-switching and attention residue suggests the expensive part of an interruption is reconstructing context, not the interruption itself.

What the evidence says

Laboratory studies consistently show performance decrements after switching, with larger effects when the prior task was left unfinished. Field estimates of 'time to full resumption' vary widely and should be treated as illustrative rather than precise.

Why it matters

Protecting a block of uninterrupted time is usually cheaper than recovering from a fragmented one.

●○○Association

Recovery requires low input, not merely low effort

Attention restoration appears to depend on soft fascination and reduced demand, which scrolling does not provide despite feeling passive.

What the evidence says

Nature exposure studies show restoration effects on directed-attention tasks, though methodological quality varies and publication bias is plausible.

Why it matters

If a break leaves you flatter than it found you, it may have been low effort rather than restorative.

●●●Experimental

Expecting interruption is itself disruptive

Anticipating a notification can degrade focus even when no notification arrives.

What the evidence says

Studies manipulating phone presence and expected contact show measurable effects on sustained-attention tasks, though effects are modest and some replications are weaker.

InterruptionContext lostReorientationFocus
The interruption itself may be brief. Rebuilding context is often the expensive part.
Low effort
  • scrolling
  • streaming
  • passive input
Low input
  • walking
  • quiet
  • nature
  • daydreaming
Low effort and psychological restoration are not necessarily the same thing.
A desk with a phone and notebook in natural light
02

Social Media

The behaviour tells us more than the timer does.

How, not how long.

Passive consumptionActive connection
Context and behaviour often tell us more than minutes alone.
●○○Mixed evidence

How you use a platform predicts more than how long

Directed, reciprocal interaction tends to be associated with better wellbeing outcomes than passive consumption of others' content.

What the evidence says

The active/passive distinction is well replicated in cross-sectional work but weaker in longitudinal and experience-sampling studies. Effect sizes are generally small. Direction of causation is not established.

Why it matters

Noticing whether you are consuming or connecting is a more useful question than counting minutes.

Teenagers studying together in an everyday setting
03

Kids + Teens

Aggregate hours have become the most measured and least informative number in the field.

●●○Correlational

Total screen time is a weak predictor of teen wellbeing

Large dataset analyses find associations that are real but very small — comparable in size to factors like wearing glasses or eating potatoes.

What the evidence says

Specification-curve analyses across several large datasets find average associations near r = -0.04. This does not mean nothing matters; it means aggregate hours is the wrong unit of measurement. Content, context and displacement of sleep appear more informative.

04

Sleep + Screens

The mechanism is less exotic than the headlines, and more actionable.

DeviceLater bedtimeLess sleep
Light exposureMelatonin effect
The heavier path reflects the better-evidenced mechanism: displacement and arousal, rather than screen colour temperature.
●●●Experimental

Sleep displacement is better evidenced than blue light

The clearest mechanism linking devices and poor sleep is simply time and arousal, not screen colour temperature.

What the evidence says

Melatonin suppression from device-level light exposure is measurable but modest in ecological conditions. The larger effects come from delayed bedtime and cognitive/emotional arousal from content.

Why it matters

Moving the device out of the bedroom addresses timing and arousal — the parts with the stronger evidence.

05

Digital Connection

For most people, online contact sits alongside offline life rather than replacing it.

●○○Mixed evidence

Digital contact supplements more often than it substitutes

For most people online communication adds to offline contact; substitution effects appear concentrated in specific groups.

What the evidence says

Displacement hypotheses have weak overall support, but subgroup findings — particularly among heavy passive users and socially anxious individuals — suggest the picture is not uniform.

People spending time together
06

Gaming

Why someone plays predicts more than how much they play.

●●○Correlational

Motivation predicts outcomes better than hours played

Play driven by enjoyment and social connection looks different from play driven by escape or obligation.

What the evidence says

Objective play-time data from platform holders shows near-zero or slightly positive associations with wellbeing. Problematic use appears better predicted by need frustration elsewhere in life.

A note on measurement

Where objective platform data exists, it tends to complicate the simplest stories rather than confirm them.

07

AI + Human Behaviour

A young literature, moving faster than its own evidence base.

Early evidence is not a conclusion.

●○○Association

AI companionship: early, thin and worth watching

Short-term studies report reductions in self-reported loneliness. Long-term and developmental evidence is essentially absent.

What the evidence says

Sample sizes are small, follow-up windows short, and most studies rely on self-report from self-selected users. Strong claims in either direction are currently unsupported.

08

Work + Technology

Availability norms, rather than hours, appear to do much of the damage.

●●○Correlational

Always-available norms predict burnout better than workload

Perceived expectation of responsiveness outside hours is associated with exhaustion independent of actual hours worked.

What the evidence says

Telepressure research finds consistent associations with recovery failure and sleep quality. Interventions targeting norms show more promise than interventions targeting individuals.

Why it matters

Team norms about after-hours responsiveness are more changeable than individual willpower.

A laptop glowing in a dim room after hours
The bigger picture

Hours are easy to measure.Human experience isn't.

What we're watching

Some areas are moving faster than the evidence.

These are the questions we're following closely and reporting on carefully — where the research is genuinely developing too quickly for confident conclusions.

01

AI + companionship

Early evidence. Long-term effects largely unknown.

02

Child development + AI

Rapidly evolving with limited longitudinal evidence.

03

Digital social norms

How expectations around availability, responsiveness and connection change human behaviour.

Not every interesting finding is a conclusion. Sometimes the responsible answer is: we don't know yet.

The bigger picture

What technology replacesmay matter more than how long we use it.

The Evidence Desk

Curious, not alarmist.

The Modern Human follows emerging research across psychology, technology and human behaviour — separating what we know from what we're still figuring out.

Read the journal →Explore the ecosystem →
Share

Worth passing on?

Send this to someone who might want it too.