Quote:
However, it is far from clear whether the technology will prove accurate. Security experts warn that data-fusion technologies tend to produce a huge number of false positives, flagging up perfectly innocent people as suspicious.
1. There's a subtler problem than that, which is even more dangerous. The New Scientist description implies flagging by error is quasi-linear process, that there will be simple Buttle/Tuttle errors. However, the big difficulty with intelligence data of all kinds is generally acknowledged to be interpretation.
Human interpreters are very good at seeing meaningful patterns even when they aren't there (e.g. ley lines). And institutions are severely subject to confirmation bias, where they tend to seek out information to support their theories, rather than test them (which is the source of a very high proportion of miscarriages of justice).
The danger is that such technology will exacerbate both tendencies - which would be both a threat to innocent individuals in small worlds and also dilute and undermine intelligence resources. If you were going to use such things rationally, the first thing you'd need to do would be to use it to conduct research into the structure and distribution of ordinary social networks, and what functional transformations were introduced via the data-fusion approach. No-one has an interest in conducting such research, however. When it is guaranteed to appear that something
meaningful is happening, vast funds have been committed and a priesthood has been created, then no-one is going to set out to discover it is all worse than useless, and insiders will probably make strong efforts to discount any contrary evidence. (That's a common pattern in public policy, as in other pseudo-sciences.)
It is entirely plausible when spooky types put forward the argument: 'we need data on everyone so we can spot those who stand out'. (And this is the case being made indirectly here:
http://www.guardian.co.uk/uk/2008/aug/2 ... .terrorism ) The trouble is that that case can be made entirely in good faith and still end up being led astray by a different sort of faith in method and meaning.
2. There's another thing. Using data for intelligence purposes is quite different from using it for criminal investigation and social control. The Home Office is using the former as its pretext for gathering information in vast quantities that may then be used for the latter.
Formerly there was a clear division because intelligence was gathered and deployed extra-legally. Its sources did not exist as far as the civil power was concerned, and the civil power was exerted through the courts, which required cases made beyond reasonable doubt. That division was a safety mechanism. The methods of intelligence were limited to giving leads.
In the last decade and a half the nature of courts, of evidence, of criminal prosecution, and the status of the intelligence services has radically changed: while the secret world is nominally under control (it must be: there are written laws about it, runs the naive legalist argument) the burden to be discharged by prosecutorial systems of all kinds has weakened and official executive powers have increased. Individuals have been made much more subject to arbitrary search, seizure, and arrest. Administrative penalties and conduct control orders of various kinds (both general and individualised) were almost unknown; they are now ubiquitous. This makes the spreading of surveillance methods and stochastic interpretations vastly more likely to result in oppressive behaviour - there are far fewer constraints in practice on the exercise of official power, and far more encouragements to its use, however detailed the regulatory framework.
