Tag Archives: news

Home Affairs Committee report on West Midlands Police, fabricated AI evidence, and special treatment for Maccabi fans

Posted on by
Reply

There are some episodes that feel representative of the times, and this is one of them.

West Midlands police generated evidence with AI that contained a data-driven fabrication that was simply false. They submitted that fabricated evidence to back their policy. They denied using AI but were caught out in an elementary error. Somehow the WMP Chief Constable Craig Guildford believed he should keep his job. He eventually ‘retired’, without reflecting on his role in this, instead citing a “media frenzy”. He has been backed by a frantically deflecting MP Ayoub Khan. Other senior police officers have been exposed as misinforming policy makers. How did this institutional antisemitism come about? What happened in the following months?

In this particular case – are there other cases? – West Midlands Police wanted to stop Maccabi Tel Aviv fans from going to their team’s match with Aston Villa in Birmingham last Autumn. They carried out a match assessment and categorisation and focused the evidence collection to fit to that decision. They failed to collect full evidence. They were inclined to treat Israeli fans differently and worse than fans of other teams.

Local independent MP Ayoub Khan is very involved in attempting to ostracise Israelis, and the lengths he goes to in this have compromised him. Most people in the know say that some Maccabi Tel Aviv fans behave badly but comparatively not particularly badly. But because they are Israeli and Jewish, they receive more hostility than football fans in general. Khan’s petition to first have the match cancelled, and then to ban the Maccabi fans cited their violence, but in fact it was Birmingham cleric Asrar Rashid who urged his audience to “show them no mercy” and MP Iqbal Mohamed, who is obsessed with people who accept Israel’s existence (Zionists), who told us to think of Maccabi fans as “terrorists”. Maccabi fans seem to have their normal share of louts, to be policed in normal ways. Superintendent Jack Hadley and Deputy Constable Mike O’Hara implied that the threat was from Maccabi fans but as the police later disclosed, it was the Maccabi fans who were under threat – “We do have specific information about residents wanting to ‘arm themselves’”. Who inspired those locals? We know that even now Ayoub Khan MP has no reservations about fomenting communal grievance and division, insisting that the Chief Constable was a victim, trying to convince us that the updated evidence was somehow extorted, and clinging to discredited misrepresentations of Maccabi fans as worse than other fans. He has abandoned his principles to his hatred of Israelis.

But look how normal football violence is, even today. UEFA fines clubs and associations quite frequently. We police them. Israel’s fans are not the worse ranked of the world’s fans, though they’re solidly in the bottom half. Maccabi fans are normal, and the threat they posed was exaggerated.

Yet the campaign against Israeli participation in culture and sport is treated as if it deserves special status and special indulgence. WMP gathered many charges against the Maccabi fans, citing a bunch of stuff from a match in Amsterdam that seems to have been made up. The scale of this is appalling. For example (Vikram Dodd in The Guardian),

“But HMIC spoke to Dutch police, who told its inquiry that several key claims relied on by WMP clashed with its experience of policing Maccabi fans during that game, which was marred by violence.

The Guardian understands that Dutch police said claims such as one that Maccabi fans had thrown Muslims into an Amsterdam river were incorrect. Indeed Dutch police said the only incident remotely like that involved a Maccabi fan being found in the water.

HMIC were also told that the presence of Maccabi fans and the trouble that followed required 1,200 Dutch officers, not 5,000 as WMP had claimed.

Furthermore, the inquiry was told that Maccabi fans did not specifically target Muslim communities in Amsterdam. While there were clashes, and Maccabi hooligans did strike locals, the Israeli fans did not go hunting for people to attack, multiple sources have said.

Dutch police told HMIC that the trouble was confined to central Amsterdam, with the city’s Muslim communities living outside that area.”

The HMIC aren’t saying that West Midlands Police were motivated by antisemitism, or capitulated pressure from Ayoub Khan. But this isn’t the point. It misses the significance of the generative AI fabrication that Craig Guildford submitted in evidence to MPs, about a match with West Ham in 2023, which he initially denied when questioned on December 1st:

“He was then asked by Paul Kohler MP: “Hold on, so you did an AI search, got something about West Ham and just whacked it into the…” Mr Guildford replied: “No, not at all. We do a very comprehensive assessment.””

He denied it again on January 6th, definitively saying “We do not use AI”. He had to go back on that this week.

“In his contrite letter he has admitted that Microsoft Co Pilot was used by officers gathering evidence about previous European ties involving Maccabi Tel Aviv. The tool had cited a match between the Israeli club and West Ham, which turned out to be fictitious. However, mention of the ‘fake’ game was included in a subsequent police report, included without any further checks.”

It’s another example of institutional antisemitism – in other words, ambient antisemitism without particularly willful antisemites. It is well known that GenAI reflectes the biases of the society or linguistic group that provides its training data. Reporting for the UN, Ashwini K.P. sets out the dangers for human rights of predictive policing (it’s fabrications rather than predictive policing that I’m talking about in this post, but the point she is making is that AI reproduces bias). As other authorities explain, if an AI is trained on biased information, “it may hallucinate patterns or features that reflect these biases“. Important to note that it’s inaccurate to refer to GenAI ‘hallucinations’, since they are designed into GenAI and are integral. They are rightly called ‘data-driven fabrications’.

I also can’t accept the argument that the Israeli fans were banned for their own good. When the far right rallied in London last year, the Met Commander Clair Haynes didn’t tell Muslims to stay away. She said she recognised the threat against them and that:

“There have been some suggestions that Muslim Londoners should change their behaviour this Saturday, including not coming into town. That is not our advice. Everyone should be able to feel safe travelling into and around London. Our officers are there to ensure that is the case and we’d urge anyone who is out on Saturday and feels concerned to speak to us.”

West Midlands Police didn’t extend the same principles to Maccabi fans under threat of pre-emptive attack. Instead they wound up getting them banned by Birmingham City Council Safety Advisory Group. This group included at least one Councillor, Waseem Zaffar, who has a prejudice against Israeli participation in anything at all. BBC Sport, reporting tensions over the Israel-Hamas war in a “predominantly Muslim area”, compounds this sense that you can’t be openly Israeli in a Muslim part of town. This kind of domineering hostility, if it exists, would be dangerous to indulge.

Ayoub Khan MP has (rightly) drawn attention to Islamophobia from far right latch-ons who claim to defend Jews but are actually using Jews as a pretext. But he witholds solidarity from Jews targeted by the weird and avid enthusiasm of the campaign he prominently supports, to ostracise Israelis. I don’t think Khan cares about racism, I think he’s a divisive sectarian. I’m probably better than him when it comes to Islamophobia.

The injustice here is that WMP and Khan made Maccabi fans out to be particularly monstrous. They overstated and fabricated evidence to justify the ban they wanted to impose. Going along with local Muslim majority concerns based on fabrications was the path of least resistance, according to His Majesty’s Inspector of Constabulary. I don’t think this quite gets it. In failing to check sources, including GenAI fabrications, for bias, they have discredited themselves. It’s really bad that Chief Constable Craig Guildford retired rather than resigning and helping the police service to learn from his errors. The fabrication of his officers and the sloppiness in checking is already disturbing. For me yielding to the vocally hostile majority (on this matter, not on all matters) in Birmingham is extremely ominous, and the fact it was officially overturned only reminds me how thin the membrane is between the rule of law and majoritarianism. The HMIC letter says:

“Because of the make-up of the local population and WMP’s stated inability to engage effectively with the local Jewish community, the force encountered a lack of balance in the range and strength of views it was hearing. It appears to me that the force didn’t recognise this imbalance or satisfactorily take steps to redress it.”

One good this is that this has been surfaced by an inquiry by His Majesty’s Inspectorate of Constabulary. I thought at the time that if I were Home Secretary Shabana Mahmood, I’d commission an investigation into GenAI use in our police services. I’d also take action to officially stop referring to AI ‘hallucinations’. Everyone should know by now that these fabrications are not a bug but a feature and history is a casuality.

*****

Time has passed and on February 24th The Register published a piece on the Home Affairs Committee’s report about ‘The Maccabi Tel Aviv ban‘. Understandably The Register focuses on GenAI use and reports that WMP has switched Copilot off,

“All of this lands at an awkward moment for policymakers. In a white paper published last month, the government set out plans to ramp up the use of AI across policing, including £115 million over the next three years for a new National Centre for AI in Policing known as Police.AI”

The report itself is a good reminder that community engagement helps towards robust decision-making. It highlights that WMP didn’t consult with the local Jewish community and therefore limited their own access to information, at the same time as favouring representations from the Muslim community (para 15) and politicians (para 34). They then went with their bias, unduly blaming Maccabi fans for unrest that locals were involved in (para 11). They have harmed community relations (para 21).

The Safety Advisory Group was vulnerable to political manipulation, and there is a recommendation that elected politicians no longer sit on them (page 23). However, I can’t find reference to that in the government response.

Three methods to withstand anti-politics

Posted on by
Reply

Following the strands of analysis of the wicked problem that is Donald Trump’s presidency, as far as I can tell his success boils down to ignorance. Ignorance about the democratic process and about the policies of the candidates. To call voters ignorant is generally considered insulting, so very few pundits are grasping that nettle. So rather than focus on the details of Democratic messaging that probably won’t even reach its target audience, what’s needed is to go upstream and address the root of the problem. Less and less do citizens benefit from serious encounters with contrasting views. Highly spreadable information without oversight, algorithmically filtered to boost emotional content, is now a major way people get their news. Propaganda based on misinformation has weakened democracy, perhaps to a critical point. I canvass a lot, and it’s clear that you can’t assume anybody knows anything at all. There are time limits on what you can do while canvassing so the doorstep won’t be the upstream intervention that democracy proponents are looking for.

While cleaning the house over the past few weeks I listened to two democratic engagement methods in action, and I would love to see them mainstream.

The first is ‘The Experiment‘, second episode of the BBC Radio 4 series ‘How would we know if democracy had died?‘. The centre right think tank Onward recently found that a quarter of young people believe democracy is a bad way of governing the country, around 61% favour a strong leader who isn’t tangled up in parliament and elections, and 41% were positive about military rule. Working with Public First, nine people from the East Midlands aged 18 – 35, selected for representativeness, are recruited to role play the cabinet of a UK governance responding to a scenario of public disorder at a time of strain on emergency services and risk of financial crisis. As the scenario unfolds, the cabinet are asked questions about their decisions – should the police use force at this point, should social media misinformation be restricted, should troops be deployed – and reach consensus or vote, sometimes with individuals casting a deciding vote. Updates are given after each round of discussion. The disorder escalates, supermarkets have to ration, schools close, the police ask to use rubber bullets and to detain without charge, emergency powers are considered, the opposition pressurise the government and the army is called. The crisis intensifies and the decisions become more difficult, as the scenario seeks each participants’ red lines about democracy. Then a charismatic army colonel diffuses a flashpoint, and the cabinet debate inviting him to join the government. I won’t give spoilers. The discussion is recorded and excerpted in this 40 minute episode, and the debrief discusses the Onward findings above.

The second method is Surrounded, broadcast by US company Jubilee Media. I can’t find it on their website – maybe it’s new – but there is an edited film from the two hour session on their YouTube. Before the US election I listened to US Secretary of State for Transport Pete Buttigieg (Democrat) try to convince undecided voters in the swing state of Michigan, where he lives. The format is Buttigieg at a table for two, surrounded by a circle of 25 articulate undecided voters who lean either to Trump, a third party or not voting at all. It’s not a free-for-all; Buttigieg makes claims (to paraphrase, Trump is in it for himself while Harris is in it for the public; law and order is better pursued through a litigator than a felon; and one other I forget) and individual participants take the chair opposite him to test those claims, with each claim given 20 minutes of discussion. Each participant has a red flag and if 11 are raised then a new participant takes the chair opposite Buttigieg to pursue their own line of argument. After the discussion, there is a vote. The result wasn’t included in the video, but as with Democrat arguments in general, when engaged with directly, they are often favourably received.

A frame from the video illustrating the format of Surrounded - two people in dialogue at a table surrounded by 24 other participants seated on a circle of chairs. A clock counts down.

The Surrounded method was a breath of fresh air to me. It avoided polarity and introduced diversity by recruiting undecided voters rather than Republican voters, so we become acquainted with different viewpoints. It takes the rich dialogue you only get in pairs or small groups and makes it go further by broadcasting it, which compared to political monologue or debate, is full of life. The questions are voters’ questions rather than focus groups’ or the pundits’. I also love what the method demands from participants by way of responsibility. The prospect of being replaced for low quality engagement is a clear motivation to bring their best arguments or otherwise take a listening role. This is underlined by continuous fact checking. I would modify the format a little for equity reasons – not everyone can race to the chair, and the diversity of the speakers could have been better given who was in the room. But the basics of this design are beautiful.

Buttigieg didn’t always succeed, incidentally, though he did significantly shift the voters. One ecologically-minded participant whose attire shouted Green insisted he address her dilemma between Harris and Jill Stein, and expressed her complete disillusionment with a two party system. He responded with what this smart voter surely already knew, that in Michigan a vote for anyone but Harris was a vote for Trump so voting Harris is what Michigan citizens who grasp the planetary polycrisis of bioversity, global heating and pollution must do. I wonder how she voted, in the end. This is a decision I have had to face in my own first-past-the-post system where overall vote share is irrelevant.

ClimateFresk is the third method, and different in that, as far as I know, it hasn’t been recorded for an audience. I paid £10 for and joined in person for three hours one evening after work. ClimateFresk teaches the causes and consequences of global heating in facilitated groups, followed by a whole-group discussion to connect knowledge, emotions and intention to take action on the climate crisis. There are other Fresks on different aspects of the polycrisis but this one is focused on climate change. Each group of six to eight gathers round a table covered by a 1x2m sheet of plain paper. A facilitator puts the Human Activity card at one end of the table, and hands out further sets of cards in rounds interspersed with discussion about how they relate to each other and where to put them on the table. One side of the card is an illustrated title. Often the illustration is a chart from a pivotal piece of research, with the group jointly interpreting these graphics. On the other side of the card is a little more information. The first set of cards is different kinds of human activities, then come sets on the kinds of consequences that scientists know about but which most of us can’t detect, and finally the consequences that we see and feel – conflict, drought, floods, fire. The aim is to have a rough flow of causes and consequences. Then we take pens and draw further inter-relationships including the kinds of feedback loops that lead to irreversible tipping points where cause and effect intensify each other. The result is the fresco-type artefact in the title. Then participants sit together and discuss how we feel. This is important because often the shorter your time engaging fully with climate damage, the more emotional upheaval you experience, and no responsible method would leave people to face that alone. And what helps with these emotions is to think about action, with emphasis on what individuals can influence rather than only our own personal consumer behaviour. That is the final part of the session.

Climate Fresk in action. Photo from above, participants lean over table arranging cards to illustrate causes and consequencies.

This method is very effective at visualising the causes and consequences and seeing what connects with what. Participants learning enough climate science and systems thinking to avoid falling into the trap of thinking there is nothing humans can do, and to highlight where the effort to address global heating needs to go. Tough on drought, tough on the causes of drought, sort of thing. You can clearly trace from the consequences back to the human activities. Obviously, this method depends on people choosing to be there, and to scale it needs trained facilitators. It’s growing steadily, and you can filter this calendar to find one you can join. Expect to pay, but not much.