AI is revolutionizing policing but at what cost?

DW (English)

AI is bringing a tech revolution to policing. But AI policing software can also reinforce existing biases and threaten our right to privacy. DW lays out the pros and cons of the technology.

Transcript

00:00Using AI to fight crime? Sounds great, right?

00:03Facial recognition could help catch criminals faster

00:06and predictive policing might even stop crimes

00:09before they actually happen.

00:10But it's not that simple.

00:12More security or a threat to privacy.

00:15Let's weigh the pros and cons of AI in policing.

00:19A recent study found that 75% of European citizens

00:22support the use of AI by police and military,

00:25for example, in surveillance.

00:27That's surprising, given how many innocent people

00:30have been harmed by faulty AI.

00:32Now, don't get me wrong,

00:33AI can sift through large amounts of data quickly,

00:37like databases of wanted persons or crime statistics.

00:40It can also draw conclusions faster than any police officer.

00:43But AI does make mistakes and can be misused,

00:47for example, in facial recognition systems.

00:5175% of Argentina's capital, Buenos Aires,

00:54is under video surveillance.

00:56The city rolled out a massive facial recognition program

00:59in 2019.

01:01Within months, the government claimed

01:02nearly 1,700 wanted criminals had been caught.

01:06But dozens of errors were also made,

01:08leading to unjustified police checks and even arrests.

01:12One resident, Guillermo Ibarrola,

01:14was wrongfully detained for six days.

01:16Data protection activists sued the city,

01:19which led to the system being shut down in 2022.

01:23And it's been in limbo ever since.

01:25Activists and city representatives

01:27are still debating over a legal framework.

01:29Because there are more concerns.

01:31The investigation found data not just on criminals,

01:34but also politicians, activists, and journalists.

01:37Were police using the system to track people illegally?

01:40An even bigger concern with facial recognition

01:43is that it can be used for ethnic profiling.

01:45China, for example, has used this tech

01:47to monitor and detain the Muslim Uyghur minority.

01:50And facial recognition also has a general flaw.

01:54It doesn't work equally well for everyone.

01:56Studies show it's less accurate for people of colour,

01:59women, and non-binary individuals.

02:01So there's a lot of work to be done

02:03before these systems can work without bias.

02:07Predictive policing.

02:10But what if crimes could be prevented

02:12before they're even committed?

02:14That's the idea behind predictive policing.

02:16With AI, large data sets can be analysed

02:19to spot patterns and trends humans might miss.

02:21In theory, this could make police work more efficient

02:24and reduce human error in decision-making.

02:27But accuracy and fairness of these models

02:29depend purely on the quality and diversity

02:32of the data that they are trained on.

02:34So the risk of reinforcing existing biases is high.

02:37When AI is trained on bias historical crime data,

02:41it can reinforce those biases.

02:43Over-policed minority neighbourhoods

02:45may appear to have higher crime rates.

02:47And as a result, predictive policing tools

02:49could then unfairly target these communities,

02:52only increasing inequality.

02:54That said, predictive models are already being used

02:57in certain fields.

02:58For instance, they help assess risks at large events

03:01such as football matches.

03:02This allows police to focus on areas

03:04where issues are most likely to occur.

03:07For example, fights.

03:09AI and police.

03:10Can it work?

03:11AI can save police officers time.

03:14For example, by creating case logs in the future.

03:16It can ensure that they're in the right place

03:18at the right time.

03:20And it could even lead to fairer decisions

03:22by removing human prejudices from the equation.

03:25But to get there, some obstacles need to be overcome.

03:28First, databases must be truly representative

03:31and diverse to ensure they treat everyone fairly.

03:34And second, there needs to be a clear legal framework

03:37on what data authorities can access.

03:39Abuse of this tech could threaten our privacy

03:42and civil rights.

03:43What do you think of the police using AI?

03:46Let us know.

Category

Transcript

Be the first to comment

Recommended