It may be making us less tolerant, more racist, more easily persuaded, and more fixed in our mindsets. But is the internet also contributing to less violence? The global homicide rate fell 14% between 2015 and 2024. Canada’s homicide rate dropped 16% in 2025. In the United States, the FBI reported a 9.3% decline in the violent-crime rate that year. These improvements are uneven, and homicide statistics cannot describe every form of danger. But they raise a question worth asking in a world where so much of what we see online feels hostile: could digital life be contributing to greater physical safety?
We spend considerable time discussing what the internet is doing to our attention, relationships and politics. We should also consider what happens when an enormous share of human interaction moves into a space where people are physically separated. Some of the conflict that once required proximity can now happen through words. Some of the isolation that once went uninterrupted can now be broken by a message. Some of the people we once knew only through stereotypes can now explain their lives to us directly.
My hypothesis is that this transformation may be contributing to reductions in some forms of offline violence. Its effect could be significant, even if its most visible expression sometimes looks like an endless argument. Establishing how significant will require evidence we don’t yet have, but we can make inferences about empirical evidence. Like: is the Internet bad for discourse, or showing us who we really are? Are we both much better and more annoying than we thought? Now that the every day thoughts of every day people, their interests, habits,behaviour and obsessions are now visible to us continuously, do we find less to fear about each other? Aren’t most strangers with whom we follow and interact online fairly decent people? Don’t, ironically, our most irritating interactions happen with people we know?
The scale of our investment in online interaction is extraordinary, and can lead to suppositions it’s based on a fascination and curiosity about other people. The International Telecommunication Union estimated that six billion people used the internet in 2025. DataReportal counted 5.66 billion social media user identities in October that year. WhatsApp reported handling more than 100 billion messages and two billion calls on a regular day. Radicati projected worldwide email traffic of 392.5 billion messages sent and received daily in 2026.
It is a steady drumbeat of humanity, of things said and almost as often things left unsaid. It describes a world in which connection with and knowledge of strangers is available at a scale and durations previous generations could barely imagine.Those figures overlap, and a substantial portion of digital traffic is commercial, automated or passively consumed. They cannot be added into a total of meaningful human conversations. But that hardly seems to matter.
A 2024 meta-analysis covering 88 independent samples and 9,385 participants found a small effect consistent with digital contact between groups reducing prejudice. That gives some support to the idea that online connection can expand our understanding of one another. A stranger can become someone whose circumstances, fears and motivations we recognize.
The more interesting possibility is that some conflict now remains verbal because it takes place online. Distance can interrupt the immediate progression from insult to confrontation. A person can stop replying, leave a group or return to a conversation later. Even an unpleasant exchange may have a different outcome when the participants cannot physically reach each other.
There is some evidence that this even changes criminal opportunity. Economist Ilaria Masiero’s study of broadband diffusion across U.S. states between 2000 and 2012 found a reduction in crime consistent with people spending more leisure time at home. The effect was not statistically significant for more severe offences, so it cannot establish that broadband reduces murder. It does suggest that changing where people spend their time can change the circumstances in which crime occurs.
This should not be confused with the claim that expressing anger automatically releases it. A 2024 meta-analysis of 154 studies did not support venting as a general anger-management strategy.But it might increase both empathy and reduce anger sufficiently to moderate impulses and behaviour. Rehearsing a grievance can be very different from processing it with someone who helps you regain perspective or sharing it with someone. The question is whether digital interaction changes what people do with their anger, including their opportunities to act on it. This is a possible w okanaruon for why we seem angrier but manifestations of anger are dropping.
The decline in known serial killers offers another illustration of how the conditions for violence have changed. Researchers point to DNA analysis, surveillance cameras, phones, digital records and changing public behaviour. Repeated offending becomes harder when investigators can connect evidence and reconstruct movements. That does not demonstrate a decline in violent intentions. It shows how technology can restrict the opportunity to act repeatedly without detection.
Aggressive enforcement should not be treated as synonymous with crime prevention. A 2024 systematic review covering 56 studies found that aggressive order-maintenance strategies did not produce statistically significant crime reductions. The strongest effects came from community and problem-solving approaches addressing conditions in particular places. The researchers also identified concerns about harmful consequences of aggressive enforcement, including racial disparities in arrests. Source: Braga, Schnell and Welsh, 2024. (crimejusticelab.org)
Some criminal-justice interventions can make subsequent outcomes worse. A study of nonviolent misdemeanour cases in Suffolk County, Massachusetts, found that declining prosecution reduced the likelihood of a new criminal complaint over the following two years by 53% among defendants whose treatment depended on the prosecutor assigned. This is evidence that prosecution can be counterproductive for some low-level cases; it does not establish the same effect for every offence or defendant. Source: Agan, Doleac and Harvey, “Misdemeanor Prosecution”. (doi.org)
Deterrence requires a separate analysis. The National Institute of Justice concludes that increasing punishment severity does little to deter crime, while the perceived likelihood of being caught matters substantially more. Visible presence and opportunities for early intervention therefore cannot simply be used to justify harsher punishment. Source: “Five Things About Deterrence”. (nij.ojp.gov)
Neighbourhood policing offers another approach. Japan’s kōban model places small police posts within communities, supporting regular contact and cooperation with residents. A JICA evaluation of kōban-style community policing in metropolitan São Paulo, using data from 2006–2019, estimated reductions of 9.2% in vehicle robbery, 8.5% in other robbery and 3.4% in intentional injury compared with areas without the intervention. It found no statistically significant effect on murder. These results support the potential of sustained local presence and community cooperation, while leaving their individual contributions unresolved. Source: JICA Annual Evaluation Report 2023. (jica.go.jp)
Investment in economic security and treatment also has measurable preventive effects. Research on young people with disabilities found that withdrawing income support at age 18 increased subsequent criminal charges by approximately 20% over two decades. Separately, a study of five U.S. states expanding Medicaid coverage of psychological treatment estimated a 7.8% reduction in the index crime rate relative to its pre-expansion mean. These findings establish concrete reasons to include income support and accessible mental-health care in public-safety policy. They also challenge the assumption that greater reliance on enforcement is the natural response to crime. Sources: income-support study; psychological-treatment study. (nber.org, onlinelibrary.wiley.com) Any explanation must also respect the timeline. North American crime began falling in the early 1990s, before mass social media adoption but just as the Internet started to take off and email made everything so much smaller.
Meanwhile, our exposure to other people’s experiences, challenges and day to day lives, has expanded enormously. We see a lot more how different peoples exist. A single feed can bring together joy, celebration, family, wars, work, achievement, security and tiny daily challenges to overcome which end up looking remarkably similar from all around the world. The algorithm is self referential. Negativity finds fear. The resulting experience can feel relentless even where local risk is falling.
The distinction appears in the data. The Council on Criminal Justice found that an average of about 69% of Americans believed national crime had increased annually during 2005–2024, despite generally declining crime. Yet Gallup found that 73% of adults worldwide felt safe walking alone at night in 2024, its highest recorded result. Perceptions of society and experiences of one’s own neighbourhood can move differently.
None of this makes online cruelty harmless, and it is an enormous, shapeshifting threat, now being used for profit and politics. Research by Karsten Müller and Carlo Schwarz found evidence that social media helped propagate anti-refugee violence in Germany. Digital threats can become physical threats. Harassment, coercion and fraud can inflict lasting damage without an offender ever entering the victim’s room.
But reducing physical assault remains a meaningful benefit. Avoiding injury or death matters even when a conflict remains emotionally painful. We need to assess both outcomes, including whether harm has been prevented, displaced or amplified. And we need to be more realistic about the ranges of human behaviour and plan for them instead of being shocked by the inevitable.
There may also be a longer process of collective learning underway. Recognizing provocation, checking a claim, leaving a destructive exchange and understanding an unfamiliar perspective are skills people can develop. We should investigate whether those skills are becoming more widespread and whether they carry into offline behaviour. Time spent using a platform alone does not prove that sophistication is increasing.
We are conducting an immense, ongoing experiment in human proximity without physical proximity. Its consequences may take decades to understand. The internet’s ugliness is visible every day; a confrontation that never becomes an assault is much harder to see. Understanding the full effect of digital life requires finding ways to measure both.
Sources
Statistics Canada: police-reported crime, 2025
WhatsApp: daily messaging and calling
Radicati: email traffic estimates and forecasts
Masiero: Broadband Internet and Crime
Digital intergroup contact meta-analysis
Anger-management meta-analysis
Northeastern: decline in serial killers
NIJ: Crime and Policing Revisited
