I'm fairly seriously prosopagnosic. That may be why I am so skeptical about identification in general, but the research into the condition is beginning to militate quite strongly against the presuppositions many people in the law-and-order business make about the utility of photo-ID.
The BBC has an interesting programme on how bad people are at facial-recognition, here:http://www.bbc.co.uk/programmes/p007bf7z
The points in this that I suggest are important for policy are:
people are not in fact very good at matching strangers to their photographs. People tend to be much better at recognising people they know than people they don't, and mistakenly generalise what you could call 'a theory of matching' from their experience with their familiars. (This is much less easy to do if, like me, you find recognising people you know quite hard.) Looking at someone's 'photo-ID' on a one-off occasion will ordinarily be hit-and-miss, unless you are one of the rare "super-recognisers".
2. A significant number of people (the programme suggests 3%) are sufficiently bad that it handicaps them in everyday life, but generally they don't realise it. I was 30 before I understood I had a problem, though I can recall incidents back to 6 or 7 years old, which are examples. Yet officialdom assumes that anyone can recognise others from pictures, to the degree suggested by the false theory of matching. There is no testing of passport control staff, police, security guards, bar-staff... anyone, who is expected to do the matching. Yet 3% or thereabouts not only are ordinarily useless at it but are being put to an impossible task. They may well compensate intuitively by responding to other behavioural cues that have nothing to do with facial features. The Clever Hans http://www.skeptics.org.uk/article.php?dir=articles&article=clever_hans.php
syndrome. Some Hypotheses:
Photo-ID for age-checking is rather like voice-stress or polygraph "fraud detection". There is no real evidence for its accuracy, yet the story that it "obviously" works is so plausible to so many, that few even question it. There is a massive confirmation bias, and it is probably not acting as more than an intimidatory deterrent.
Software "facial-recognition" is thought of as pretty bad for the purpose that it is put to (fact) - but it may well be better than 90%+ of people under the same circumstances of matching strangers in large numbers, and infinitely better than the small fraction of checkers who, unknown to themselves and their employers, are getting it wrong almost all the time. ID-ing travellers and drinkers doesn't do a lot. It is an imposition and a cost on everyone, but posturing security doesn't stop competent imposture.
Flash and dash is close to useless, but there's a huge industry of ID badges built on it. The false assumption is that replacing a doorkeeper who knows everyone (and in most cases will therefore recognise them quite well), with picture-passes that 'anyone can check', is more efficient and more secure. Quite the reverse. But look at the reception area in any large firm and what to you see? Picture passes with RFID tracking of the pass and bored temps concerned that you display a badge properly, operating on the assumption that your badge is you.
There is now in the system a prejudice and an interest in not facing (ha!) these ideas. Everything in fact tends towards dismissing them. The authoritarian mindset is particularly prone to confirmation bias (Cf. the DNA arguments), and Clever Hans will sucker them every time.