Police algorithm said Lina was at 'medium' risk. Then she was killed

22 mmarian 12 4/20/2025, 8:53:20 AM bbc.co.uk ↗

Comments (12)

LatteLazy · 10d ago
Isn’t this how risk works: you assess 1000 people to be at medium risk, 1 dies. Because medium risk means 1/1000? Kind of like lottery tickets all being a low chance of winning the jackpot, yet someone one usually does…

I don’t mean to talk down one travesty. But this is just stats meeting reality/human (mis)understanding of stats.

echoangle · 10d ago
Yes, it’s like people complaining that there was a 90% chance of rain and then it didn’t rain.

Of course in reality you either get killed by your partner or you don’t, every judgement except „negligible“ and „extreme risk“ will be wrong if you need absolute certainty. And if you want to have a false negative rate of literally 0, you need to return „extreme risk“ for every single case.

hcfman · 10d ago
Similar thing happened to me friend of mine about 20 years ago but without algirithms in New Zealand. He had been warning for more than a year that he former girlfriend was dangerous and he wanted custody.

Then she suffocated their 6 year old daughter with a pillow and burnt the house down.

SOLAR_FIELDS · 10d ago
Who designed the algorithm and who implemented it, and where is the accountability for those two actors? Ultimately they are the ones at fault here, not the “algorithm”

The actual headline is a flawed narrative. It should read “police software written by X and deployed under Y’s purview caused death”. The story should be about that, not the falsehood that some nebulous “algorithm” was responsible