Interesting survey results showing most public-policy researchers use AI tools in research (despite few receiving any training), but in a limited way and despite having concerns. Most wanted better guidelines from professional bodies and universities: policyviz.com/2026/07/01/a...

School breaks in Sweden lead to fewer assaults (children aren't concentrated together at school) but more residential burglaries (homes are empty while people are on holiday). Which is a useful reminder to analyse different crime types *separately*! link.springer.com/article/10.1...

People in London support police using AI for a wide variety of tasks including translation, identifying high-risk repeat offenders and prioritising deployment. They do feel it’s important to have a human in the loop, though. www.police-foundation.org.uk/news/new-res...

The way I've gotten this across in talks to US audiences is to explain that by the time the Battle of Gettysburg was fought... ...the London Underground had already been open for six months.

Fantastic @reducingcrime.com podcast discussion between my colleague Shane Johnson and @jerryratcliffe.net on some of the many ways all types of crime are changing due to technology. Essential listening if you’re interested in crime of any sort.

Opinions from towns beginning with other letters are presumably also available! 😁

Feeling very sorry for store security guards, PCSOs and officers... anyone in a stab vest and kit today.

Requiring professionals to report possible child abuse to police (🇫🇮) was associated with – more incidents recorded by police – no increase in prosecutions – a big increase in investigation times suggesting the rule simply overwhelmed police capacity doi.org/10.1007/s106...

Lots of examples here of good everyday uses of AI tools. I use various AI tools for quite a lot of work tasks, but what they’re capable of has moved so fast that it’s very useful to hear what other people are using them for.

England’s latest sunset of the year (more or less). Although the sun is setting behind hills that are in Scotland! 📍 Allonby, Cumbria

In light of today's rail crash at Bedford (on the trains I use to commute), it's worth remembering how safe train travel has become in Britain: from 1981 to 2000 over 100 people were killed in train collisions, but in the past 20 years that number is only 2. www.railforums.co.uk/threads/repo...

Ethnic disparities in police stops/traffic tickets (🇺🇸) substantially reduced (and for some groups disappeared) when comparing stops/tickets to demographics of people caught by automated red-light cameras rather than comparing to residential population. www.crimrxiv.com/pub/2zpvatvy...

Journals often say it's hard to get peer reviewers for papers. I hear many stories of really good papers struggling to get published as a result. But I also see papers get published despite quite weak methods (example in next post). Something's definitely wrong, but what is it?

This dataviz is just amazing visquill.com/gallery/uk-a1

we seem to imagine that infrastructure can only ever subtract from natural beauty, despite so many examples to the contrary

I've been seeing this map all week across social media, but quite a few things about it are wrong (no Southwark, two Bexleys) that suggest it's been AI generated without checking the details. So I created a version based on the actual census data:

Domestic abuse victims who’re themselves police officers/staff (🏴󠁧󠁢󠁥󠁮󠁧󠁿) report receiving worse support from police than other victims, eg because investigators expect them to be more resilient than other victims, so they’re denied proper support.

Senior detectives leading major investigations (🏴󠁧󠁢󠁥󠁮󠁧󠁿) find organisational causes of stress (eg lack of resources, excessive internal bureaucracy) more stressful than outside causes (eg traumatic events, court delays, poor work–life balance).

At the point where people are setting fire to vehicles, those people are either rioters or arsonists, not protestors.

My UCL Crime Science colleague Jyoti Belur has set herself quite the challenge to answer this question in her professorial inaugural lecture this evening!

The struggle is real

In a detail which will surprise no one, the first of the "patriots" to plead guilty to offences connected to the rioting in Southampton on Tuesday night has a criminal record as long as your arm, including.... possessing a knife.

Men's professional football matches in London led to real but quite small increases in police use of force in the local area. The typical effect of a match was that there were 0.77 more uses of force in a borough on a match day than would be expected otherwise. doi.org/10.1111/obes...

Some observations: - Rates of contact with policing differ by age, sex and ethnicity - Some of that will be a function of institutional and individual officer decision making - Much will reflect wider inequalities in society - It's difficult to identify the split

Most street violence occurs in a small number of hotspots. New research shows hotspots (🇺🇸) are connected by (the same offenders using) the same weapons across hotspots, with nearby hotspots and those with similar ethnic makeup being more connected.

Interesting framework on how police (🇬🇧) are likely to use different AI tools in future, and the different ethical and reputational issues that creates. Good to see the Met having this debate at least somewhat in public.

Routine arming of police (🇦🇺) didn’t lead to any change in the rate at which police shot people. This study doesn’t look at other potential consequences such as officer suicides using their firearm, which has been a consistent problem in Australia.

Police body-worn video recordings (🇺🇸) were associated with faster investigation times and (for some crime types, e.g. domestic abuse) less chance of prosecutors discontinuing a case, but little change in the likelihood of conviction. doi.org/10.1111/1745...

Social media posts claiming that Dorset Police said 44% of sex offences are carried out by asylum seekers living in asylum hotels are false. We could find no evidence this is true. Dorset Police told us the figure is “incorrect” and was not issued by them.

Useful chapter on applying routine activity theory to cybercrime – likely to make a good addition to the ‘further reading’ if you’re teaching routine activities to students.

[stares pointedly at almost every academic seminar speaker]

This review finds the evidence on the effectiveness of offender-based interventions (eg using cognitive behavioural therapy) to prevent repeat domestic abuse is weak. Lots of places are spending money on this, so we really need better evidence. doi.org/10.1007/s108...

Police (🏴󠁧󠁢󠁥󠁮󠁧󠁿) are much more likely to strip search Black and Asian people brought into custody, even after the offence a person is arrested for, whether they use drugs/self harm and how co-operative they were are taken into account.

Lovely example for teaching of how weak statistical designs can produce misleading results, and how stronger designs can fix that.

Gunshot detection technology (🇺🇸) led to a large increase in police workload in responding to calls for shots fired (which tripled, partly due to false-positive alerts from the tech), but led to no change in the likelihood of solving any crimes.

The City of Ottawa posted the raw speed data from before and after Doug Ford banned speed cameras and the effect is so obvious it's not even necessary put a line marking when that happened. This chart shows the percentage of drivers going 15km/h or more over the posted limit.

❓Gen AI could well annihilate scientific publishing by overwhelming journals with submissions. Many journals are in any case struggling as reviewers get ever harder to find. But I’ve no idea what comes next. How will we publish in future and what should we be doing to prepare?

Banning police officers from lying to juvenile suspects during interviews (🇺🇸) led to no change in the likelihood of crimes being solved. (Lying to any suspects in interviews should be banned anyway, of course!)

Police reports written by officers (🇺🇸) using an AI assistant (Axon Draft One) were rated by experienced supervisors as less accurate in a blind test versus reports written without an AI tool.

Really fascinating and detailed maps on how the population of different bits of Europe has changed over the past 60 years: rural depopulation and the move from east to west being the big stories. correctiv.org/en/europe/20...

I touched on this in this piece: www.ft.com/content/4d00... Violent crime has declined. Demand has shifted to stuff that happens behind closed doors. It’s really hard and time-consuming to investigate.

You also see this just by looking at the British Retail Consortium's 'most shoplifted items' list, where every year it's not 'things people desperately need to live' it is 'things you need to make illegal drugs'!

A really useful framework for police leaders to think through the issues associated with adopting a new technology, written by my colleagues Shane Johnson and Kate Bowers. More detail in this paper: doi.org/10.1111/1467...

Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work. Admin is work.

Example 2️⃣ I use purrr ALL the time to pull in a long list of csvs (or other files), often binding them together in the same step.

Apparently today is the day that a lot of people are learning that the law absolutely allows police officers (and, indeed, members of the public) to kick people in the head when faced with a potentially lethal threat such as a person actively stabbing people.

Good explanation of why events that obviously seem to be terrorism aren't always immediately declared as being terrorist incidents. Whether something is a terrorist incident depends largely on the offender's motivation, which it can take time to gather evidence of.

Allowing bars to open later in a city (in 🏴󠁧󠁢󠁳󠁣󠁴󠁿) increased alcohol-related late-night crime and ambulance calls, while allowing only nightclubs that met quality standards to open later in another city led to no change. ➡️ Venue management matters. bmjpublichealth.bmj.com/content/4/2/...

@scottkeay.bsky.social 1h 41min!? 😳 This had better be good.