The Darker Side Of Analytics

I recently read an article in one of the major Dutch newspapers (warning - it's in Dutch!) about their Government's moves to prevent crime by profiling citizens. The announcement was made by the Minister Of Justice and, not surprisingly, it has caused some discussion within The Netherlands. I wonder what the reaction in Australia would be if the Federal Government announced that all Australians were to be profiled?

Here's what the Dutch Government is proposing:

• using 'digital techniques' (whatever they are) to create profiles of people
• initially this will consist of information on: 
• • banking 
  • flight 
  • internet use 
• citizen profiles will then be compared to profiles of convicted criminals
• if/when a match is found, government officials will be sent and email or sms alerting them of this 'fact'. 

The example use quoted by the minister is for a very worthy cause: preventing pedophiles from traveling to countries where child prostitution occurs.

But if you look at the wider picture: wow, the implications are serious and widespread!

I wonder how many times their 'digital techniques' will create false positives and (presumably) false information being sent out?

I remember a situation some years ago when I owned a company in Silicon Valley and still lived in Amsterdam. I was traveling very regularly between the two places and I was profiled by US Customs solely on the basis that Amsterdam is famous for tulips and a couple of more seedy chemical habits. The conversations were pleasant and short-lived - but I would hate to be having similar conversations if, say, my next business was in Bangkok and my own Government was telling a foreign country "watch out" for this guy simply because I had a pattern of travel to Thailand that was similar to some disgusting criminal who had served time in The Netherlands!

It couldn't happen, could it?

Comments

[SninD]

аська для телефона
http://icqmobilephones.net/ - icq для мобильного

[Will Dwinnell]

Whether and how predictive analysis represents a threat depends entirely on the nature and context of their use.

Whether data mining and statistics could ever be usefully and fairly applied to the direct identification of individuals as criminals or terrorists remains, in my mind, an open question.

On the other hand, I believe that most people would regard the monitoring of business activities and transactions for illegal activity, such as money laundering, as a mostly beneficial thing. Such monitoring often includes data mining. Note that money laundering is an important source of funding for some criminal and terrorist organizations.

[OzAnalytics]

James , I agree. Perhaps my main worries are twofold:

1. That most people accept analytics as unbiased when we know that they are only as good and unbiased as the analysts who created them. Bias can be introduced though deliberate means (take a look at the 'data' produced by most lobbyists) or simple ignorance of your craft.

2. Analytics does what you suggest - it swings attention from racial prejudices ('against the Indian with a beard' you mention in your blog) and focusses in on the white guy - me! How unfair ;-)

Thanks for commenting - I've been lurking around your blog for some months and enjoy it very much. Smart (Enough) Systems? Already has a proud place in my library!

Steve

[James Taylor]

Interesting post. I think analytics are ethically neutral (as I discuss in http://www.b-eye-network.com/blogs/taylor/archives/2009/07/oz_analytics_-_the_darker_side_of_analytics.php) and the risk is the risk of personal bias with or without analytics
JT

Author, with Neil Raden, of Smart (Enough) Systems
http://jtonedm.com

[OzAnalytics]

Good points.

Also, what the state publicly declares it is doing is not always matched by what is happening secretly. Just look back at Australian Government declarations about the Echelon signals intelligence network for proof.

Perhaps I should be arguing for more sophisticated profiling which wouldn't mistake my behaviour for that of a criminal?

Of course overall that will never be achieved as there will always be assumptions made during profiling that lead to incorrect matches.

Thanks for posting, they always make me think further.

[James Beresford]

It is not only inevitable, I wouldn't be in the least bit surprised if it is already happening.

Semi-consciously security officers have been doing their own predictive modelling in terms of who to keep an eye on - by race, sex, behaviour - it should come as no surprise that as the systems get bigger, these decisions are analysed and formalised.

Of course, you are arguing from the perspective that as an individual you act and behave uniquely - which is a false assumption. Human behaviour is often surprisingly predictable, and criminals are not always the brightest and will often act in simple, easy to spot ways.

The scary element of this activity is not that you might get unwanted attention from the state - but what behaviours the state defines as criminal.