Pay attention you lot. I’ll tell you once and once only. Word from on high. From nine hundred hours this morning we’ll be using a fancy new facial recognition software system from the UK. It’s called EvoFIT.
You mean like a photofit, Sarge?
An e-fit yoke then?
Sort of but not quite. We’ll come to all that in a minute. And no groaning at the back. Some of you need to shape up or ship out. This is the Garda Siochána, not The Lord of the Bleeding Rings. There’s far too many of you lot using techniques and procedures that are old-fashioned, clichéd, fantastical and downright unrealistic and unscientific. Even if it is crime fiction.
Like what d’you mean, Sarge?
Don’t get me started. Your witness interview procedures, the way you do your line-ups, your use of photofits, your police sketch artists…
What’s wrong with identity parades?
Your ones are wild dodgy.
But they look really good. In the TV adaptations I mean.
Shut up. This ain’t Red Rock or Sun Hill. It’s real life. Really. And the trouble with identity parades, even your fictional ones…
VIPER. You know. Video Identification Parade Electronic Recording.
Oh. Right. He means a video parade. The trouble with all your parades – video parades, live parades, whatever – is that they…
I love video parades, me. Like suppose you have a story with a coupla dozen characters. If you have to do a live parade instead of a video one you’d have to go round town physically rounding up dozens more characters who look like the suspect, and before you know it you’ve spent two days and three chapters just getting the details of your feckin’ line-up right. In a video parade, though, you can get hundreds of faces from the database in a jiffy and…
Can we get back to the new EvoFIT system?
Line-ups? I feckin’ hate that. You know, having to get similar faces, and the witness gets confused, and I get confused, even the bleddy reader gets confused. But suppose you’re trying to organise a line-up and the suspect is of a particular racial minority in Ireland, you know, a really really really tiny minority. Like say your prime suspect is a tall Eskimo lady with a big tattoo and a…
You can’t say Eskimo any more.
Oh sugar. I just meant it can be a right bugger trying to find all the right characters, but once you do it’s blummin’ worthwhile because a good old line-up will look far better on the finished page and…
You’re not listening. Line-ups. Are. Stone Age. When you lot do line-ups you’re always giving the wrong instructions, doing really basic mistakes like telling the witness character whether the suspect character is or isn’t present, slipping in another psychological bias here, a little flaw there.
You’re always skipping over all the tiny real-life details like whether (a) you’ve chosen the rest of the guys in the line-up to resemble the suspect, or (b) for them to match the general description that a witness has already provided, or (c) to be not too similar to your suspect because it’ll only make it much harder for the witness to tell them all apart. You’re just as bad when it comes to photo-based identity parades because…
You mean them ones with the photos in the bukes?
Yeah, the photos in the books.
But I love mugshot albums. Next you’ll be telling us we can’t use police sketch artists any more.
Got it in one.
Not even if the person doing the sketches is a police character, like, say, what’s your man’s name again, um, Camille Verhoeven?
No. No sketches. Or photofits. They’re all out.
You can’t have them any more, because…
Because. Orders. On high. For sound scientific reasons, from now on we’ll be using this new EvoFIT system.
Lookit, I could bore you with a load of psychological research claptrap about how we recognise faces, but when it comes down to it our brains don’t see a face as lots and lots of separate squiggly bits. We see it holistically…
Did he just say homeopathy?
…as in a single, whole face. That’s the main difference between an e-fit system and a photofit one. Not just that one is electronic and the other is physical strips of photographs. In an e-fit you show the witness a whole face instead of the separate components one by one, the eyes, noses, eyebrows, scars, moustache…
And that’s just the mother in law!
Listen up, you bunch of mysogynistic wankers, alcos and misfit mavericks. Get with the programme. Sketch artists are out, your witness interviews stink, your line-ups are criminal, an e-fit is ten times better than a photofit but we’ve something even better than that again this morning if you’ll only let me…
What’s e-fit stand for again?
Electronic Facial Identification Technique, I think. It presents the witness with a whole face. Whole. “Holistic”. Right? But the new EvoFIT yokemebob is even better again, because, well, it’s evolutionary.
(Complete silence for ages.)
How does it work?
At last. A sensible question. For a start, it constructs facial composites of offenders by witnesses and victims.
Like a photofit then?
No, you dozy sap.
Like an e-fit?
Yeah, but not quite. I mean what it does is show the witness loads and loads of complete faces – 18 at a time on a touchscreen. Then it evolves.
It starts with this general set of faces with random shapes and features, and the witness selects the ones that are most like the suspect (they can also touch which ones are least like the suspect to give the computer more clues, but we’ll leave that aside for the moment).
These selections become the “parents” of the next generation, a further 18 faces that are based on a mix of the shapes and features from these selections. So the sets of faces keep evolving and evolving like that, over and over until you narrow it down to the best fit.
All very Darwinian when you come to think of it. Survival of the fittest fit and all that. Gimme that brochure over there…
Current practices build composites with a single face and operating on only one face means that a witness must continuously describe changes necessary to make a face better. This ‘recall’ process is a naturally hard task for anyone to do. Recognising a face, on the other hand, is fairly easy and accurate. EvoFIT therefore attempts to tap into witnesses’ recognition rather than recall.
But does it have hats and hoods?
Yeah yeah. All the caps and hoodies and hairstyles you’ll ever need.
Does it do Philip Treacy hats?
Now feck off up the yard.
Does it do Photoshop too?
So is it any good?
It has a success rate of 60%, so the manufacturers say. That’s compared with 5% for traditional “feature” systems. So I’m expecting most of your crime fiction cases to be wrapped up far more quickly, like in say 50 pages or so. It’s novellas instead of novels from now on, lads.
A bit like the way DNA forensics did for the big fat blockbusters when it came along, Sarge?
I could murder you.
And your bedtime reading tonight…
They say all EvoFIT roads lead to Dr Charlie Frowd from the Department of Psychology at the University of Winchester. Six gardaí based in the photographic section at Garda HQ have been trained to use the system by Dr Frowd. The following are all .docx or .doc Word documents or PDFs…
Frowd, C.D., Bruce, V. & Hancock, P.J.B. (2008). Changing the face of criminal identification. The Psychologist, 21, 670-672.
Frowd, C.D. (2011). Eyewitnesses and the use and application of cognitive theory. In G. Davey (Ed.) Introduction to Applied Psychology. BPS Wiley-Blackwell.
Frowd, C.D. (2010). Varieties of biometric facial techniques for detecting offenders. International Conference and Exhibition on Biometric Technology. 6 – 7th September 2010, Coimbatore, India.
Frowd, C.D., Skelton, F., Atherton, C., & Hancock, P.J.B. (2012). Evolving an identifiable face of a criminal. The Psychologist, 25, 116 – 119.
Frowd, C.D. (2012). Facial Recall and Computer Composites. In C. Wilkinson and C. Rynn (Eds). Facial Identification. Cambridge University Press.
Frowd, C.D. (2014). Facial composite systems. In T. Valentine and J. Davis (Eds.) Forensic Facial Identification. Wiley-Blackwell.