UK Police Learn That More Surveillance Data Doesn't Mean Better Surveillance Data
from the swamped-by-data dept
Way back in 2002, we wrote that with all the efforts (mostly in the UK) to try to allow law enforcement officials to collect more and more surveillance data that more data doesn’t mean better data, and in fact, all that data often makes it harder to find the right or necessary data. The trick is to be smarter about surveillance, not just focus on getting more. It appears that police in the UK are finally learning this lesson. Last month, we saw how all that data was leading to mistakes as patterns were being spotted that weren’t there. And, now, UK police are discovering that they’re missing important information and clues because they’re just overwhelmed by garbage data. Of course, this won’t stop the increasing collection of data, because no one seems to want to admit that too much data helps bury the needles deeper in the haystack.
Filed Under: police, surveillance, uk
Comments on “UK Police Learn That More Surveillance Data Doesn't Mean Better Surveillance Data”
The problem with surveillance is that you need the human bandwidth (ie: more people watching more surveillance) for additional surveillance data to be useful. There are millions of people doing millions of things all the time, no matter how much footage you have you have only so many eyes watching so many people at any given time. If you only have 100 people reviewing surveillance footage of a million people that’s a HUGE bottleneck. It can’t be done, for each person they watch takes away from their ability to watch someone else. So when one person watches a single other person for 24 hours he can’t thoroughly watch anyone else during that same time. If he switches to watch someone else that distracts him from the first person. When he does switch to watching someone else is he going to start from yesterday (because yesterday he was watching the first person, remember)? If so, then by the time he reaches today it’ll be tomorrow (by then it might be too late) and that’s a whole day that he missed watching other people. If he starts from today then he missed everything that person did yesterday. Sure there are filters to help weed out unimportant data (ie: by fast forwarding through much of the footage)and one can have multiple screens at once showing multiple locations and occurrence and people at once but overall these limiting factors make it impossible to monitor everyone all at once. The disparity between what needs to be monitored and the amount of people doing the monitoring is HUGE.
Re: At least they have something to study
A government won’t be able to monitor its citizens in every matter. But getting surveillance data is essential to some extent because the data, though incomplete, can inform the govt. something about the monitored citizens.
Re: Re: At least they have something to study
This is the problem. The government will be monitoring for patterns, which predisposes them to make up said patterns. In which case, those three gravel trucks that went on the local interstate end up looking like truck bombs.
i have no experience with surveillance, but i can tell you this more data is better as long as its properly data based.
The problem is that for this data to be properly processed you need enough human bandwidth to process it. By human bandwidth I mean enough humans to watch all the other humans and interpret each and every situation. It’s not a simple matter whatsoever.
“have no experience with surveillance, but i can tell you this more data is better as long as its properly data based.”
You really have to think of it in terms of information systems because that’s what it is. Data needs to be collected and properly processed (ie: interpreted) but you only have so much bandwidth (in this case, human bandwidth) so you have to be careful about how you select and filter your data.
Mo' data, pleeze...
The concept that more data is better data is truly a hoot! Anyone who has done more than five minutes of security data review knows this. The only thing more data gets you is an increased probability of finding what you are looking for. Using the ‘more data’ technique is generally referred to as ‘brute force’ methodology and is usually considered a method of last resort.
I intend to begin 24 hour video surveillance in my new home with 0 monitoring. why you might ask, to have a video record of what happens. If a crime in committed then a video of the event is very useful. Massive surveillance programs really can’t be about catching and preventing crimes, but instead about catching and punishing criminals after the fact. You can hope that criminals will be deterred by surveillance, but don’t count on it
As any good crook knows, the best way to hide is in plain sight! This cruft just makes it easier for them to do their “jobs” – the crooks that is! 🙂
Disagree With Surveillance Analysis
Mike,
I disagree with this ongoing meme in Techdirt of arguing that collecting data is foolish because it is overwhelming. It’s true that many agencies collect more data than they can competently handle, it is wrong to suggest that it is impossible to use the data effectively.
The important reason for this is the undeniable advance of technology. For now, machines do a pretty poor job of mining the mass of surveillance data. Most video data feeds need a real human to interpret them, look for suspects, or see interesting patterns that indicate crimes. For now, then, the data far outstrips our ability to use it economically. However, as machines get more capable to do that job, the costs of sifting through the haystack for the needle will drop dramatically.
While facial recognition software is still unreliable, it will get better. Machines will eventually become capable of detecting suspicious behavior and patterns. This is inevitable. They may never be as quick as the right brain of a human at understanding a complex scene, but they will steadily improve. At that point, your argument of “just too much data” becomes moot. So maybe we should collect the data now, and eventually machines will be able to use it effectively?
This, BTW, is by no means an avocation of a surveillance society. I detest the notion of cameras on every corner, as it is in the UK. But I would make a privacy argument against cameras. Your argument of the “mountain of data too big to ever manage” argument doesn’t stand up to Moore’s law very well.
Re: Disagree With Surveillance Analysis
Here is an analogy to my argument:
Google’s Street View project, which seeks to drive many of the roads of the world with cars equipped like this one:
http://www.wikinomics.com/blog/index.php/2009/04/13/google-street-view-car-pose-or-hide/
As most of us here know, the cars gather images from multiple cameras, data from high-precision GPS units, and other road characteristics and data. This is a mountain of data.
In fact, this mountain of data, just 10 years ago, would have been thought of as just too much data to be useful. Who could afford just the storage for all those TB? What computer systems could handle and manage that data in a way to do anything useful with it?
Clearly, though, Moore’s law marches on, prices come down and capacities go up for storage, processing, bandwidth, display technology. Rendering solutions like AJAX are invented, and good ideas for how to use the information are created. What was formerly a mountain of useless data is now a practical service.
So surveillance video data is much the same. Storage is available and cheap today, but the technology to have computers “watch” the video isn’t ready yet. How long will it be?
Second example would be Navteq or TeleAtlas, which gather similar data using cars for their mapping solutions sold to Garmin, Magellan, Mapquest, Google, etc. What these companies do is take the video with the cars, then pump it all to India overnight, where cheaper HR pore over the video transcribing road signs an other road characteristics into the map GIS database. Using cheaper labor to pore over video files makes a mountain of useless data VERY useful.
Good points everyone.