1 00:00:02,570 --> 00:00:04,250 S1: A listener production. 2 00:00:09,090 --> 00:00:13,739 S2: Welcome to Crime Insider's forensics. For those joining us for 3 00:00:13,740 --> 00:00:17,820 S2: the first time. My name's Catherine Fox. I'm a former GP, 4 00:00:18,000 --> 00:00:23,400 S2: crime author and screenwriter. I'm enthralled by forensics and have 5 00:00:23,400 --> 00:00:27,720 S2: spent thousands of hours researching for books and screenplays. So 6 00:00:27,720 --> 00:00:32,340 S2: I thought, why not turn my research into a podcast? 7 00:00:32,909 --> 00:00:36,810 S2: Every week you'll be joining me in discovering how forensic 8 00:00:36,810 --> 00:00:40,740 S2: science is helping solve high profile crimes in Australia and 9 00:00:40,740 --> 00:00:46,920 S2: around the world. This week how CCTV footage can lead 10 00:00:46,920 --> 00:00:51,360 S2: to convictions, but also miscarriages of justice. 11 00:00:51,690 --> 00:00:54,480 S3: Unless you can prove it scientifically, you can't really make 12 00:00:54,480 --> 00:00:57,450 S3: that claim. And that was the fallacy of individualization. If 13 00:00:57,450 --> 00:00:59,520 S3: you can't prove it scientifically, don't use it. 14 00:00:59,970 --> 00:01:03,060 S2: Professor Glenn Porter is an expert in the field of 15 00:01:03,060 --> 00:01:08,550 S2: forensic photography. He spent decades taking photos, analyzing them and 16 00:01:08,550 --> 00:01:10,830 S2: presenting them as evidence in court. 17 00:01:10,830 --> 00:01:14,940 S3: So it's very much a filmic visual narrative that crime 18 00:01:14,940 --> 00:01:16,920 S3: scene operatives work on. 19 00:01:17,370 --> 00:01:20,520 S2: It's easy to think that video and photographic evidence is 20 00:01:20,520 --> 00:01:23,699 S2: the smoking gun and the slam dunk for a conviction, 21 00:01:25,110 --> 00:01:28,470 S2: but it's actually far more complicated and not as clear 22 00:01:28,470 --> 00:01:32,400 S2: cut as you might think. You'll hear from Glenn about 23 00:01:32,400 --> 00:01:36,780 S2: how photos can be extremely helpful, but also can sometimes 24 00:01:36,780 --> 00:01:41,770 S2: hinder a case for the prosecution. Glenn's taking us back 25 00:01:41,770 --> 00:01:44,170 S2: to a high profile double murder. 26 00:01:46,940 --> 00:01:51,950 S3: Peter Johnson was accused of murdering a couple at Cadi. 27 00:01:52,490 --> 00:01:57,380 S3: The police facts allege that Mr. Johnson tortured these two 28 00:01:57,380 --> 00:02:00,770 S3: people to get their Pin number, and then went around 29 00:02:00,770 --> 00:02:04,790 S3: ATMs extracting money. Now, there was an ATM in Windsor, 30 00:02:05,150 --> 00:02:11,030 S3: in western New South Wales, where the CCTV captured what 31 00:02:11,030 --> 00:02:14,810 S3: police allege. Mr. Johnson taking the money out. But the 32 00:02:14,810 --> 00:02:18,080 S3: quality of the CCTV was quite poor and you could 33 00:02:18,080 --> 00:02:21,020 S3: see by the footage that Mr. Johnson was aware that 34 00:02:21,020 --> 00:02:23,330 S3: there was a camera there. He did what I would 35 00:02:23,330 --> 00:02:26,660 S3: call a recce. He walked past and noticed that the 36 00:02:26,660 --> 00:02:29,329 S3: camera was there. So when he was extracting the money, 37 00:02:29,360 --> 00:02:32,450 S3: he held up what looked like a white handkerchief or 38 00:02:32,450 --> 00:02:34,940 S3: something like that over his face, so his face couldn't 39 00:02:34,940 --> 00:02:39,260 S3: be identified. So police had the details of the ATM, 40 00:02:39,260 --> 00:02:43,550 S3: obviously the account of the deceased, um, there was a 41 00:02:43,550 --> 00:02:48,800 S3: slight time difference between the ATM and the CCTV timestamp, 42 00:02:48,800 --> 00:02:51,890 S3: but that's that's fairly normal. The technology never lines up 43 00:02:51,889 --> 00:02:54,980 S3: all the time, but the couple were dead in their 44 00:02:54,980 --> 00:02:58,250 S3: home at that time that the money was extracted. You 45 00:02:58,250 --> 00:03:03,320 S3: saw this person extracting the money. The police went and 46 00:03:03,320 --> 00:03:09,889 S3: interviewed two people, his ex-wife and a friend of Mr. Johnson, 47 00:03:10,220 --> 00:03:14,150 S3: and showed them the CCTV footage, which they were able 48 00:03:14,150 --> 00:03:18,800 S3: to then identify Mr. Johnson through that video footage or 49 00:03:18,800 --> 00:03:22,369 S3: the CCTV footage. When it came to Mr. Johnson's trial, 50 00:03:22,370 --> 00:03:26,540 S3: I gave evidence for for the defence on the issues 51 00:03:26,540 --> 00:03:30,200 S3: that may occur because in both of the statements and 52 00:03:30,200 --> 00:03:34,340 S3: their oral evidence, they said they couldn't identify Mr. Johnson 53 00:03:34,340 --> 00:03:36,770 S3: from his face because he was covering his face and 54 00:03:36,770 --> 00:03:41,330 S3: it was fairly grainy, fairly poor quality. But they could 55 00:03:41,330 --> 00:03:45,350 S3: recognise or identify Mr. Johnson through the way he walked. 56 00:03:45,680 --> 00:03:50,390 S3: Now the problem I had with that, from an evidential perspective, 57 00:03:50,390 --> 00:03:54,680 S3: I'm I'm not doubting what they claim, but the issue 58 00:03:54,680 --> 00:03:58,970 S3: there from a technical perspective, was that the CCTV was 59 00:03:58,970 --> 00:04:02,210 S3: shot around two frames a second, which results in a 60 00:04:02,210 --> 00:04:06,860 S3: very jagged, animated type of motion. It's certainly not a 61 00:04:06,860 --> 00:04:10,100 S3: natural motion, which is around 30 frames per second. So 62 00:04:10,100 --> 00:04:14,060 S3: I was tasked to explain to the court how this 63 00:04:14,060 --> 00:04:18,020 S3: could be misleading. The motion isn't a natural motion, so 64 00:04:18,020 --> 00:04:23,360 S3: to make some form of recognition or identification could be dangerous. 65 00:04:23,720 --> 00:04:27,530 S3: I gave my evidence and Judge Anthony Wheely very good. 66 00:04:27,770 --> 00:04:30,770 S3: Supreme Court judge asked me a question. He said. So, 67 00:04:30,770 --> 00:04:35,089 S3: doctor Porter, are you referring to a situation similar to 68 00:04:35,089 --> 00:04:36,320 S3: a Charlie Chaplin film? 69 00:04:36,320 --> 00:04:37,760 S2: The silent movies. 70 00:04:37,760 --> 00:04:41,810 S3: The silent movies where the movement is quite animated, jerky 71 00:04:41,810 --> 00:04:43,310 S3: and not natural? 72 00:04:43,490 --> 00:04:47,360 S2: Well, that's probably because back then the silent films were 73 00:04:47,360 --> 00:04:52,070 S2: recorded at about 16 to 18 frames per second. And 74 00:04:52,070 --> 00:04:54,470 S2: I think for people to understand, it's almost like one 75 00:04:54,470 --> 00:04:57,500 S2: of those flip books. So we used to draw as kids, 76 00:04:57,500 --> 00:04:59,600 S2: you know, a whole thing of papers in a, in 77 00:04:59,600 --> 00:05:01,490 S2: a book, in a corner, and then you can get 78 00:05:01,490 --> 00:05:05,299 S2: your own little cartoon going. So it's easy to then 79 00:05:05,300 --> 00:05:09,350 S2: understand in that context why this gentleman's gait would have 80 00:05:09,350 --> 00:05:11,240 S2: appeared quite odd. 81 00:05:11,540 --> 00:05:14,330 S3: And that distortion, I don't know how much it would 82 00:05:14,330 --> 00:05:18,080 S3: affect their ability to make that judgment. That's that's not 83 00:05:18,080 --> 00:05:21,650 S3: my call. But what was interesting, too, is, uh, Judge 84 00:05:21,650 --> 00:05:25,339 S3: Whaley also instructed the jury, um, as a result of 85 00:05:25,339 --> 00:05:30,860 S3: another Supreme Court matter, uh, Tang, he instructed the jury 86 00:05:30,860 --> 00:05:34,280 S3: that the evidence given by the two witnesses is not 87 00:05:34,279 --> 00:05:39,350 S3: identification evidence, but recognition evidence that they recognize Mr. Johnson 88 00:05:39,350 --> 00:05:40,280 S3: on the video. 89 00:05:40,370 --> 00:05:45,020 S2: The difference between thinking they recognize and identifying correctly. There 90 00:05:45,020 --> 00:05:47,990 S2: seems to be a very big gap between the two. 91 00:05:48,020 --> 00:05:49,130 S2: Just as a layperson. 92 00:05:49,580 --> 00:05:53,180 S3: Yes. And there's a couple of distinctions. I mean, in 93 00:05:53,450 --> 00:05:58,969 S3: facial identification, there's a distinction between facial recognition software, which 94 00:05:58,970 --> 00:06:02,539 S3: enables a search of a database, and it gets to 95 00:06:02,540 --> 00:06:05,540 S3: a what they call a recognized level. But to go 96 00:06:05,540 --> 00:06:10,310 S3: the step to identification, it needs a forensic expert that's 97 00:06:10,310 --> 00:06:13,489 S3: trained in facial identification to be able to make that 98 00:06:13,490 --> 00:06:16,099 S3: claim a little bit like a fingerprint. They were used 99 00:06:16,100 --> 00:06:21,380 S3: the automatic fingerprint identification system, Aphis or Nifrs, and to 100 00:06:21,380 --> 00:06:23,210 S3: do a search of the fingerprint that they might have 101 00:06:23,210 --> 00:06:27,140 S3: found in the crime scene, the AI or the computer 102 00:06:27,140 --> 00:06:32,000 S3: technology can do a search of thousands in the database, 103 00:06:32,000 --> 00:06:36,049 S3: but it's actually the forensic examiner that actually takes it 104 00:06:36,050 --> 00:06:40,550 S3: to the identification level. And that's that's a system similar 105 00:06:40,550 --> 00:06:44,659 S3: to forensic imaging. And there's been quite a bit of 106 00:06:44,660 --> 00:06:49,610 S3: work done from. Universities around the world, in psychology departments, 107 00:06:49,610 --> 00:06:54,680 S3: around people's ability to recognize people in images, and even 108 00:06:54,680 --> 00:06:59,630 S3: in poor quality CCTV or poor quality video. And the 109 00:06:59,630 --> 00:07:03,710 S3: results are quite consistent. The psychologists split it up into 110 00:07:03,710 --> 00:07:07,969 S3: two categories familiar and unfamiliar faces. So if you know 111 00:07:07,970 --> 00:07:13,970 S3: somebody even with poor quality CCTV, the accuracy is quite high. 112 00:07:14,150 --> 00:07:18,680 S3: But if it's an unknown, an unfamiliar face, the error 113 00:07:18,680 --> 00:07:23,150 S3: rate goes up quite significantly, up around 3,040%. So it's 114 00:07:23,150 --> 00:07:27,170 S3: an interesting part of our innate ability to recognize people 115 00:07:27,170 --> 00:07:30,230 S3: when we see them. Familiar faces. We do a pretty 116 00:07:30,230 --> 00:07:33,650 S3: good job at recognizing them in images, even if they're 117 00:07:33,650 --> 00:07:34,400 S3: poor quality. 118 00:07:35,450 --> 00:07:38,180 S2: One of the interesting things, though, is comparing to a fingerprint. 119 00:07:38,180 --> 00:07:42,140 S2: There's only a finite number of whorls, loops, things that 120 00:07:42,140 --> 00:07:45,950 S2: they can find in combination. Surely in a human face 121 00:07:45,950 --> 00:07:54,380 S2: there are so many variables length, distance, um, depth of muscle, fat, tissue. 122 00:07:54,380 --> 00:07:59,420 S2: And then you have possible cosmetic interventions as well. Is 123 00:07:59,420 --> 00:08:02,660 S2: a face like as unique as a fingerprint? 124 00:08:03,140 --> 00:08:07,160 S3: Well, the difference major difference between face and fingerprint is 125 00:08:07,160 --> 00:08:12,020 S3: that faces change. Fingerprints don't. So weight gain, weight loss, 126 00:08:12,020 --> 00:08:15,650 S3: cosmetic surgery as you mentioned, age. That's one of the 127 00:08:15,650 --> 00:08:20,390 S3: things that's made facial identification quite difficult that it's not consistent. 128 00:08:20,390 --> 00:08:24,260 S3: You need a known source and exemplar to make that comparison. 129 00:08:24,260 --> 00:08:28,130 S3: And that exemplar is obviously done sometime in the past. 130 00:08:28,130 --> 00:08:32,030 S3: So depending upon how far in the past it was 131 00:08:32,420 --> 00:08:37,309 S3: may affect the the accuracy. Martin Evison in 2014 wrote 132 00:08:37,309 --> 00:08:41,540 S3: a brilliant paper called The Third Forensics, and he suggested 133 00:08:41,540 --> 00:08:46,760 S3: that forensic identification is entering its third phase. So fingerprints 134 00:08:46,760 --> 00:08:51,410 S3: being the first, DNA being the second method, and faces 135 00:08:51,410 --> 00:08:55,430 S3: identification was going to be the third forensic identifier, but 136 00:08:55,429 --> 00:08:57,920 S3: it just hasn't played out that that way. 137 00:08:58,309 --> 00:09:03,579 S2: You mentioned the database with fingerprints. Facial databases. Do they 138 00:09:03,580 --> 00:09:07,630 S2: help you in terms of helping to identify a person, 139 00:09:07,630 --> 00:09:12,250 S2: in terms of the likelihood that these facial features are 140 00:09:12,250 --> 00:09:16,290 S2: consistent with this? Person in this photograph. 141 00:09:16,920 --> 00:09:20,730 S3: Yeah. So there are some systems out there that the 142 00:09:20,730 --> 00:09:24,839 S3: Australian government are now using for facial identification. And we 143 00:09:24,840 --> 00:09:27,360 S3: see it now on our mobile phones when we sign 144 00:09:27,450 --> 00:09:32,340 S3: deeds and contracts, you can there, there are verification sort 145 00:09:32,340 --> 00:09:35,460 S3: of processes that are using the face as a, as 146 00:09:35,460 --> 00:09:39,810 S3: a form of verification based on your passport photographs or 147 00:09:39,809 --> 00:09:44,220 S3: your license photograph. So that technology certainly coming to play. 148 00:09:44,220 --> 00:09:45,990 S2: That's facial recognition isn't it. 149 00:09:45,990 --> 00:09:49,559 S3: Facial recognition. Yeah. Not identification from a forensic perspective. Yeah. 150 00:09:49,559 --> 00:09:53,610 S3: But but certainly the facial recognition is is playing a 151 00:09:53,610 --> 00:09:57,360 S3: part with facial databases because we, we've been in New 152 00:09:57,360 --> 00:10:01,410 S3: South Wales and other states, we've been collecting biometric data 153 00:10:01,410 --> 00:10:05,640 S3: from our driver's licenses for 15, 20 years now. It's 154 00:10:05,640 --> 00:10:08,820 S3: been quite a while since photo ID has been used 155 00:10:08,820 --> 00:10:13,410 S3: in our driver's licenses. Those authorities, government authorities have been 156 00:10:13,410 --> 00:10:16,469 S3: collecting that. Same with our passport. Our passport now has 157 00:10:16,470 --> 00:10:21,000 S3: biometric information. A passport photograph has biometric information. So the 158 00:10:21,000 --> 00:10:24,990 S3: state actually has as a resource quite a lot of 159 00:10:24,990 --> 00:10:27,930 S3: people's faces in databases. How is the state going to 160 00:10:27,929 --> 00:10:31,260 S3: use that. They've got some legislation at the moment that 161 00:10:31,530 --> 00:10:36,240 S3: allows organisations to use it in those software applications. I 162 00:10:36,240 --> 00:10:39,630 S3: just had a one of my PhD students who finished 163 00:10:39,630 --> 00:10:45,060 S3: a research project looking at whether the facial databases and 164 00:10:45,059 --> 00:10:49,619 S3: the facial recognition software, when you get poor quality CCTV 165 00:10:49,620 --> 00:10:53,970 S3: where you can't really see facial detail. She tested whether 166 00:10:53,970 --> 00:10:59,130 S3: forensic artists could be a transition between the CCTV of 167 00:10:59,130 --> 00:11:03,059 S3: poor quality and the facial recognition software. So an artist 168 00:11:03,059 --> 00:11:07,950 S3: sketch off the images and then the facial database can 169 00:11:07,950 --> 00:11:12,120 S3: search the actual line drawing or the sketch, and she 170 00:11:12,120 --> 00:11:15,510 S3: come up with some very interesting results that said that, yes, 171 00:11:15,510 --> 00:11:19,740 S3: the technology can recognize, um, drawings. And in some aspects 172 00:11:19,740 --> 00:11:23,610 S3: they were better than poor quality CCTV photographs. The artists 173 00:11:23,610 --> 00:11:27,929 S3: are outperformed in in some areas, particularly areas of really 174 00:11:27,929 --> 00:11:31,170 S3: bad angles when there was a sharp angle, not a 175 00:11:31,170 --> 00:11:35,339 S3: normal frontal view, which is, um, straight on which I'm 176 00:11:35,340 --> 00:11:37,590 S3: looking at the camera. Now, that's a normal frontal view, 177 00:11:37,590 --> 00:11:42,060 S3: but any side angles or uh, high angles, which CCTV 178 00:11:42,059 --> 00:11:47,100 S3: often is high, um, the, the artists actually outperformed in 179 00:11:47,100 --> 00:11:50,580 S3: some instances than the photograph. So that was an interesting 180 00:11:50,580 --> 00:11:54,930 S3: experiment that Vanessa, uh, gained a PhD out of. 181 00:11:55,200 --> 00:11:57,660 S2: In terms of what can people do if there are 182 00:11:57,660 --> 00:12:00,090 S2: so many images out there? And I just think how 183 00:12:00,090 --> 00:12:04,350 S2: many selfies are being taken by criminals and non criminals 184 00:12:04,350 --> 00:12:08,130 S2: every single day. There's got to be millions and millions 185 00:12:08,130 --> 00:12:13,140 S2: of of photographs taken and uploaded. So potentially there's a 186 00:12:13,140 --> 00:12:17,940 S2: massive database, a private companies utilizing that database. 187 00:12:19,020 --> 00:12:23,220 S3: Yeah, it's a bit. It's. There is some legislation around 188 00:12:23,220 --> 00:12:28,350 S3: where private enterprise can use software from the federal government, 189 00:12:28,620 --> 00:12:32,460 S3: and they can access certain elements, but they don't access 190 00:12:32,460 --> 00:12:36,840 S3: the database, but they can apply for a recognition search 191 00:12:36,840 --> 00:12:40,829 S3: from different faces. I'm not familiar with the 100% with 192 00:12:40,830 --> 00:12:43,110 S3: the legislation, but this is something that's come over the 193 00:12:43,110 --> 00:12:46,290 S3: last couple of years. It's going to be obviously more 194 00:12:46,290 --> 00:12:50,070 S3: prominent as we move into the AI tool of technology, 195 00:12:50,070 --> 00:12:53,550 S3: I would imagine. I also had an honor student who's 196 00:12:53,550 --> 00:12:57,000 S3: just finished an ex police officer, Tony Caledon, and his 197 00:12:57,000 --> 00:13:00,089 S3: career was working in the police, in the BCI and 198 00:13:00,090 --> 00:13:05,850 S3: also in intelligence and then privately over in private security 199 00:13:05,850 --> 00:13:08,099 S3: before he retired. So he's got a sort of a 200 00:13:08,100 --> 00:13:10,920 S3: background on both of those. And what he found in 201 00:13:10,920 --> 00:13:14,699 S3: his research is these notion of fusion centers going to 202 00:13:14,700 --> 00:13:19,290 S3: occur where surveillance of the of areas are going to 203 00:13:19,290 --> 00:13:24,480 S3: be black box. So with AI being able to interpret 204 00:13:24,480 --> 00:13:28,740 S3: activity rather than if you can imagine a security officer 205 00:13:28,740 --> 00:13:32,250 S3: sitting in front of, you know, 200 screens, the the 206 00:13:32,250 --> 00:13:35,819 S3: amount of detail becomes kind of white noise and detecting 207 00:13:35,820 --> 00:13:38,940 S3: things that are happening is not very easily. Well, in 208 00:13:38,940 --> 00:13:42,890 S3: fusion centers, they're all blacked out. Sort of uses a simple, 209 00:13:42,890 --> 00:13:48,530 S3: oversimplified analogy. They're all blacked out until the AI detects 210 00:13:48,530 --> 00:13:53,929 S3: some suspicious behavior, then it becomes visible, and then the 211 00:13:53,929 --> 00:13:58,460 S3: security agency can then may make a decision on the action. 212 00:13:58,700 --> 00:14:03,020 S3: So it's an interesting concept that Tony kind of discovered 213 00:14:03,020 --> 00:14:06,440 S3: that with fusion centers, you're going to get more surveillance 214 00:14:06,440 --> 00:14:10,969 S3: of the civilians or of the citizens, but more privacy, 215 00:14:11,300 --> 00:14:14,960 S3: which is a bit, you know, sounds contradictory, but, um, 216 00:14:14,960 --> 00:14:17,689 S3: there's some protection of the privacy because it's all black 217 00:14:17,690 --> 00:14:21,290 S3: box until you see someone in a car park looking 218 00:14:21,290 --> 00:14:24,020 S3: into car windows to see if there's anything you know 219 00:14:24,020 --> 00:14:26,630 S3: in there to steal the. I would pick that up 220 00:14:26,630 --> 00:14:30,230 S3: as suspicious activity and then turn on and alert the 221 00:14:30,230 --> 00:14:33,590 S3: security agency to. Then they can make a decision on 222 00:14:33,590 --> 00:14:36,320 S3: what that action is. So the technology is going to 223 00:14:36,320 --> 00:14:39,980 S3: get very clever. Um, but I think it's good in 224 00:14:39,980 --> 00:14:43,340 S3: one way that there thinking about the privacy because we 225 00:14:43,340 --> 00:14:46,970 S3: don't want an over surveillance state. I don't think most 226 00:14:46,970 --> 00:14:50,840 S3: people would want that. But this fusion center technology may 227 00:14:50,840 --> 00:14:55,610 S3: have our interest in in heart, I guess, around privacy. 228 00:14:55,610 --> 00:14:57,620 S3: It'd be interesting to see how it pans out over 229 00:14:57,620 --> 00:14:59,270 S3: the next ten, 15 years. 230 00:14:59,480 --> 00:15:04,100 S2: Well, we've also seen deepfake videos, and this is obviously 231 00:15:04,100 --> 00:15:07,670 S2: a very divisive thing that you can, um, if you 232 00:15:07,670 --> 00:15:12,940 S2: have the technology and the ability you can. Completely alter 233 00:15:12,940 --> 00:15:16,900 S2: someone's face in a video and replace it with someone else's. 234 00:15:17,140 --> 00:15:20,440 S2: For example, at the recent international forensic conference, one of 235 00:15:20,440 --> 00:15:26,670 S2: the forensic scientists had Obama, President Obama. Actually introducing her, 236 00:15:26,910 --> 00:15:30,300 S2: and everyone initially was like, oh, and then 11 or 237 00:15:30,300 --> 00:15:34,080 S2: 12 seconds into it, the mouth didn't sync with the dialogue, 238 00:15:34,080 --> 00:15:36,510 S2: and so we all realized that it was a deep fake. 239 00:15:36,510 --> 00:15:40,110 S2: But just initially there was a room full of forensic 240 00:15:40,110 --> 00:15:42,900 S2: experts who were, oh my gosh, is that actually a 241 00:15:42,900 --> 00:15:46,920 S2: bummer introducing her? And that was the whole point. If 242 00:15:46,920 --> 00:15:51,300 S2: you could even convince a roomful of one of 1500 243 00:15:51,510 --> 00:15:56,820 S2: forensic experts how easy it is to fool the public. 244 00:15:57,180 --> 00:16:00,780 S3: That is very, very worrying. And, you know, can you 245 00:16:00,780 --> 00:16:03,990 S3: trust anything in imaging? That's that's where we might go. 246 00:16:03,990 --> 00:16:08,880 S3: But the filmmaker, Errol Morris, uh, has published a book 247 00:16:08,880 --> 00:16:12,210 S3: called Believing Is Seeing. And I've got a quote from 248 00:16:12,210 --> 00:16:16,860 S3: Morris book. It says photographs provide evidence, but no shortcut 249 00:16:16,860 --> 00:16:20,850 S3: to reality. It is often said that seeing is believing. 250 00:16:20,850 --> 00:16:24,210 S3: But what we do not form our beliefs on the 251 00:16:24,210 --> 00:16:28,290 S3: basis of what we see. Rather, what we see is 252 00:16:28,290 --> 00:16:31,800 S3: often determined by our beliefs. Believing is seeing, not the 253 00:16:31,800 --> 00:16:35,640 S3: other way around. It sort of highlights the ambiguity sometimes 254 00:16:35,640 --> 00:16:39,810 S3: of imaging as evidence. It's what we want to, uh, 255 00:16:40,050 --> 00:16:43,950 S3: believing is seeing the imagery, the fake, the deep fake, 256 00:16:43,950 --> 00:16:47,310 S3: or you know, that it's Obama and you know that 257 00:16:47,310 --> 00:16:51,000 S3: he's talking. So you often would come to the belief 258 00:16:51,000 --> 00:16:53,010 S3: that that's Obama talking. 259 00:16:54,280 --> 00:16:57,970 S2: In terms of having a suspect in custody. We all 260 00:16:57,970 --> 00:17:02,200 S2: know about mug shots. Profiles front on and each profile 261 00:17:02,200 --> 00:17:05,050 S2: can vary too, so I gather they take each side. 262 00:17:06,330 --> 00:17:10,470 S2: Do the police have a right to actually try and 263 00:17:10,470 --> 00:17:15,330 S2: photograph from the same perspective, for example, as the ATM captured, 264 00:17:15,330 --> 00:17:21,210 S2: to try and then superimpose the accused face onto the 265 00:17:21,210 --> 00:17:23,850 S2: evidence that they have and see if that matches. 266 00:17:24,680 --> 00:17:29,990 S3: We've seen this in some cases where anatomists have superimposed 267 00:17:29,990 --> 00:17:40,609 S3: faces onto forensically obtain images. It's it's dangerous. Um. Superimposition can. 268 00:17:41,480 --> 00:17:45,530 S3: Hide something as well as show something. So again, it 269 00:17:45,530 --> 00:17:49,970 S3: can really superimposition can really trick people's brains. So it's 270 00:17:49,970 --> 00:17:53,060 S3: I'm not much of a fan of it. This technology 271 00:17:53,060 --> 00:17:57,290 S3: came out with identification of, of skeletal remains. So they 272 00:17:57,290 --> 00:18:00,679 S3: would get a skull and get a photograph of the 273 00:18:00,680 --> 00:18:03,290 S3: person or the deceased who they think it is, and 274 00:18:03,290 --> 00:18:08,780 S3: then they would superimpose that image onto the skeletal, onto 275 00:18:08,780 --> 00:18:12,649 S3: the skull, and match up certain anatomical points of the 276 00:18:12,650 --> 00:18:15,500 S3: eye sockets and the teeth and, and those type of 277 00:18:15,500 --> 00:18:18,110 S3: things to try to identify the body. And this is 278 00:18:18,109 --> 00:18:22,700 S3: a technique designed before DNA. So DNA is obviously the 279 00:18:22,700 --> 00:18:25,760 S3: the premium method for identification now. But if there's no 280 00:18:25,760 --> 00:18:29,570 S3: DNA to match it to or no maternal DNA with 281 00:18:29,570 --> 00:18:33,919 S3: mitochondrial DNA, this technique I guess, is still valid. If 282 00:18:33,920 --> 00:18:36,439 S3: you talk to a fingerprint expert. What they don't do 283 00:18:36,440 --> 00:18:40,670 S3: is superimposition, fingerprints, and there's a good reason for it 284 00:18:40,670 --> 00:18:41,480 S3: because of the. 285 00:18:41,480 --> 00:18:43,850 S2: Distortion they do on telly all the time. 286 00:18:44,750 --> 00:18:47,869 S3: Yeah, they do a lot of things on tele. Um, 287 00:18:47,869 --> 00:18:51,919 S3: it can be more misleading than give information so side 288 00:18:51,920 --> 00:18:57,200 S3: by side, like for like combination is by a forensic 289 00:18:57,200 --> 00:19:01,609 S3: expert is the method of choice rather than superimposition. It's 290 00:19:01,609 --> 00:19:06,590 S3: too difficult. But image perspective which can change the face 291 00:19:06,590 --> 00:19:11,600 S3: quite a lot. Um and image perspective is. The spice, 292 00:19:11,600 --> 00:19:16,280 S3: how three dimensional objects are now recorded into two dimensional. 293 00:19:16,280 --> 00:19:19,460 S3: And basically, if you got and you'll see this on 294 00:19:19,460 --> 00:19:22,820 S3: a lot of selfies, you know, close up any close 295 00:19:22,820 --> 00:19:29,510 S3: up photographs of people's faces. The image perspective is quite deep. 296 00:19:29,510 --> 00:19:32,240 S3: So you might notice that, you know, the person's nose 297 00:19:32,240 --> 00:19:36,860 S3: seems exaggerated in a very close up photograph. The further 298 00:19:36,859 --> 00:19:42,290 S3: you move away, the flatter the perspective comes. And I've 299 00:19:42,290 --> 00:19:46,550 S3: measured about four meters. After four meters it becomes fairly stable. 300 00:19:46,820 --> 00:19:50,750 S3: So fashion photographers use this quite a lot as well. 301 00:19:50,750 --> 00:19:53,209 S3: This technique, in the sense that, you know, when you're 302 00:19:53,210 --> 00:19:58,520 S3: using people that have this beautiful proportionality of their face 303 00:19:58,520 --> 00:20:01,040 S3: and that's what makes them pretty and attractive and models 304 00:20:01,190 --> 00:20:03,470 S3: what you want to do is a fashion photographer or 305 00:20:03,470 --> 00:20:07,700 S3: a beauty photographer is capture that natural beauty. The approach 306 00:20:07,700 --> 00:20:13,340 S3: to maintaining that natural proportion and beauty is usually shot 307 00:20:13,340 --> 00:20:17,210 S3: on a long focal length lens at a distance, so 308 00:20:17,480 --> 00:20:21,050 S3: so that those proportions and parts of the face are 309 00:20:21,050 --> 00:20:24,109 S3: maintained naturally. As soon as you start bringing the camera 310 00:20:24,109 --> 00:20:27,919 S3: in closer, the proportions of the face, particularly the nose 311 00:20:27,920 --> 00:20:31,160 S3: being closer to the camera, will start to exaggerate and 312 00:20:31,160 --> 00:20:34,760 S3: those proportions must be disrupted. So part of the problem 313 00:20:34,760 --> 00:20:38,600 S3: with superimposition is that you don't often know the distance 314 00:20:38,720 --> 00:20:42,679 S3: between when that CCTV camera you can go and measure it, 315 00:20:42,680 --> 00:20:46,129 S3: which I've done at crime scenes as well. But then 316 00:20:46,130 --> 00:20:48,260 S3: you have to try to reproduce it. And there's a 317 00:20:48,260 --> 00:20:52,820 S3: lot of, um, difficulties with trying that approach. It's not impossible, 318 00:20:52,820 --> 00:20:55,820 S3: but superimposition is is certainly not the way to go 319 00:20:55,820 --> 00:20:59,210 S3: with facial identification. A side by side like for like 320 00:20:59,210 --> 00:21:02,359 S3: is is the method if it's an infrared versus a 321 00:21:02,359 --> 00:21:07,340 S3: color image, that could be problematic because of the differences. 322 00:21:07,340 --> 00:21:10,880 S3: They should be roughly the same angle, normal font size 323 00:21:10,880 --> 00:21:13,940 S3: is the perfect angle, of course, but often that's not 324 00:21:13,940 --> 00:21:15,650 S3: the case in surveillance photography. 325 00:21:16,830 --> 00:21:22,430 S2: What about ease in terms of uniqueness? How reliable is 326 00:21:22,430 --> 00:21:25,730 S2: analysis of ease and comparison in photo? 327 00:21:26,060 --> 00:21:30,230 S3: The term uniqueness is being a bit of a dirty 328 00:21:30,230 --> 00:21:35,449 S3: word in forensics over the last few years, because we've 329 00:21:35,450 --> 00:21:40,430 S3: always claimed that fingerprints are unique. But from a scientific position, 330 00:21:40,430 --> 00:21:41,480 S3: we can't prove that. 331 00:21:41,480 --> 00:21:44,149 S2: Unless you've fingerprinted every single person who's ever lived on 332 00:21:44,150 --> 00:21:45,199 S2: the planet, you. 333 00:21:45,200 --> 00:21:49,760 S3: Can model it. The US government tried to model. Model fingerprint? Statistically, 334 00:21:49,790 --> 00:21:53,750 S3: DNA does it through likelihood ratios. You know, to say 335 00:21:53,750 --> 00:21:57,350 S3: that 1 in 20,000,000 chance of someone having the same DNA. 336 00:21:57,590 --> 00:22:02,990 S3: So it can be modeled statistically. But uniqueness is something 337 00:22:02,990 --> 00:22:06,410 S3: that we don't like to sort of, uh, say, well, 338 00:22:06,410 --> 00:22:09,560 S3: we can say that they, they match or they have 339 00:22:09,560 --> 00:22:14,929 S3: similar characteristics, um, as to satisfy a match or whatever, 340 00:22:14,930 --> 00:22:19,040 S3: but not the concepts of uniqueness. So like fingerprints, there's 341 00:22:19,040 --> 00:22:22,190 S3: a claim that is, ah, unique. But we can't prove 342 00:22:22,190 --> 00:22:26,270 S3: that scientifically. And, you know, when you're comparing side by 343 00:22:26,270 --> 00:22:29,600 S3: side is which is what we can do, you can 344 00:22:29,600 --> 00:22:32,659 S3: say that there are a lot of strong anatomical similarities 345 00:22:32,660 --> 00:22:35,180 S3: between the two ears, and the likelihood of it being 346 00:22:35,180 --> 00:22:38,960 S3: the same is fairly high, but we we don't use 347 00:22:38,960 --> 00:22:43,159 S3: the word uniqueness anymore. Um, that's something we've, we've got 348 00:22:43,160 --> 00:22:48,020 S3: over it took fingerprinting, uh, bureaus quite a, quite a 349 00:22:48,020 --> 00:22:50,750 S3: while to go away from that. And I think they 350 00:22:50,900 --> 00:22:56,000 S3: may still not believe that fingerprints can't be proven. You know, um, 351 00:22:56,000 --> 00:23:01,370 S3: Michael Sachs, uh, us sociologists has written really well on this, uh, 352 00:23:01,369 --> 00:23:05,960 S3: around the, the fallacy of uniqueness and that, you know, 353 00:23:05,960 --> 00:23:12,680 S3: if we're using scientific methods within our forensic analysis, you 354 00:23:12,680 --> 00:23:17,060 S3: can't then come up with this qualitative claim that every fingerprints, 355 00:23:17,060 --> 00:23:21,050 S3: unique or every is unique unless you can prove it scientifically, 356 00:23:21,050 --> 00:23:23,209 S3: you can't really make that claim. And that was the 357 00:23:23,210 --> 00:23:26,720 S3: fallacy of individualization. If you can't prove it scientifically, don't 358 00:23:26,720 --> 00:23:27,260 S3: use it. 359 00:23:41,780 --> 00:23:48,050 S2: With all these variables light, angles, perspectives, distance. How on 360 00:23:48,050 --> 00:23:55,040 S2: earth is anyone actually identified by a photograph? Beyond all 361 00:23:55,040 --> 00:23:56,180 S2: reasonable doubt. 362 00:23:57,350 --> 00:23:59,780 S3: It seems you know, as much as I mentioned Martin 363 00:23:59,780 --> 00:24:02,540 S3: Stevenson's paper, the third forensics, it looked like, you know, 364 00:24:02,540 --> 00:24:05,899 S3: back in 2014 that this was going to be a revelation, 365 00:24:05,930 --> 00:24:09,890 S3: a new way of identifying people with the predominant amount 366 00:24:09,890 --> 00:24:13,550 S3: of imagery in the community, cameras in the community. This 367 00:24:13,550 --> 00:24:16,129 S3: is going to be the the next best thing since 368 00:24:16,130 --> 00:24:22,910 S3: DNA and fingerprints, but it just hasn't eventuated. I think 369 00:24:22,910 --> 00:24:27,739 S3: the facial recognition software and the tools that might be 370 00:24:27,740 --> 00:24:31,879 S3: used in the private sector and the verification of identity, 371 00:24:31,880 --> 00:24:35,389 S3: I think that's that's where facial recognition is going to 372 00:24:35,390 --> 00:24:39,380 S3: really play a part. But I'm not sure whether in 373 00:24:39,380 --> 00:24:43,910 S3: a forensic situation that this form of identification evidence is 374 00:24:43,910 --> 00:24:48,020 S3: going to resonate like the, like the fingerprints and DNA. 375 00:24:48,109 --> 00:24:52,310 S3: I know the AFP's facial team, the last time I 376 00:24:52,310 --> 00:24:54,950 S3: spoke with them, they hadn't gone to court. But they're 377 00:24:54,950 --> 00:24:59,660 S3: working very hard in developing standards and developing training for 378 00:24:59,660 --> 00:25:02,270 S3: their people. They're doing a very good job. But I 379 00:25:02,270 --> 00:25:06,410 S3: think our initial thinking that being able to identify from 380 00:25:06,410 --> 00:25:10,400 S3: it because it is in some way a trace like 381 00:25:10,400 --> 00:25:13,670 S3: fingerprints and DNA left at the scene. But it's not 382 00:25:13,670 --> 00:25:19,070 S3: a real direct trace. It's a visual trace. So there 383 00:25:19,070 --> 00:25:23,480 S3: is some ambiguities and or differences between, I think that 384 00:25:23,480 --> 00:25:28,310 S3: biological trace of a fingerprint or DNA versus a biological 385 00:25:28,310 --> 00:25:33,080 S3: trace from a camera. There are just some, some complexities 386 00:25:33,080 --> 00:25:35,180 S3: that it's just not made it as easy as what 387 00:25:35,180 --> 00:25:38,300 S3: we thought it was. You know, if we time traveled, uh, 388 00:25:38,300 --> 00:25:42,830 S3: 50 years, not even 50 years, probably 20 years, 25 389 00:25:42,830 --> 00:25:47,000 S3: years and say to detect these right in the modern you, 390 00:25:47,030 --> 00:25:50,479 S3: you're going to have this resource where every crime is 391 00:25:50,480 --> 00:25:53,150 S3: going to be recorded by a camera. Do you think 392 00:25:53,150 --> 00:25:56,000 S3: they'll need detectives? And okay, well, of course not, because 393 00:25:56,000 --> 00:26:00,290 S3: there it is. But it hasn't eventuated. Um, identifying people 394 00:26:00,290 --> 00:26:01,280 S3: with difficult. 395 00:26:01,280 --> 00:26:05,780 S2: It sounds to me like photographs are actually circumstantial evidence 396 00:26:06,020 --> 00:26:11,389 S2: as opposed to definitive evidence. And I would ask, and 397 00:26:11,390 --> 00:26:16,370 S2: we actually giving defense teams more to pick holes in 398 00:26:16,730 --> 00:26:21,290 S2: by bringing photography and CCTV into it. 399 00:26:21,560 --> 00:26:24,380 S3: That's a very important element, because what I've seen over the, 400 00:26:24,530 --> 00:26:28,040 S3: in my experience, is the misuse of photographs as well. 401 00:26:28,040 --> 00:26:31,490 S3: And this is this is a problem. I remember being 402 00:26:31,490 --> 00:26:34,100 S3: in a biometric conference and I was on invited on 403 00:26:34,100 --> 00:26:37,879 S3: as a panel of experts and, um, a guy I 404 00:26:37,880 --> 00:26:40,220 S3: have a lot of respect for. Richard Brooke from the 405 00:26:40,220 --> 00:26:44,149 S3: FBI was on that panel as well. And the question 406 00:26:44,150 --> 00:26:47,240 S3: was asked of Richard. Richard was using a lot of 407 00:26:47,240 --> 00:26:52,820 S3: facial identification in, um, in the FBI. Uh, the famous 408 00:26:52,820 --> 00:26:56,270 S3: case that he worked on was the Afghan girl. I 409 00:26:56,270 --> 00:26:59,780 S3: don't know whether you're familiar with Afghan girl, with, uh, 410 00:26:59,780 --> 00:27:04,880 S3: Stephen McCurry. Um, National Geographic, the green eyes. The green eyes. Yeah. 411 00:27:04,880 --> 00:27:08,120 S3: The green eye. Beautiful Afghan. Beautiful. Yeah, yeah. Absolutely stunning. 412 00:27:08,359 --> 00:27:10,760 S3: They went back a few years later and tried to 413 00:27:10,760 --> 00:27:15,770 S3: identify it. And I believe the FBI certainly made the identification. 414 00:27:15,890 --> 00:27:20,240 S3: But it was asked of the panel, particularly Richard, of um, 415 00:27:20,930 --> 00:27:24,200 S3: so what happens if you can't identify them, you know, 416 00:27:24,200 --> 00:27:27,859 S3: so you've got a mugshot and exemplar photographed a source 417 00:27:27,859 --> 00:27:31,730 S3: that you have known, known and known identity. And you've 418 00:27:31,730 --> 00:27:35,150 S3: got a, a still of a, of a face from 419 00:27:35,150 --> 00:27:40,219 S3: CCTV or ATM or a mobile phone, but the experts 420 00:27:40,220 --> 00:27:43,220 S3: just can't identify it because, you know, maybe the quality 421 00:27:43,220 --> 00:27:45,830 S3: isn't there or just it just you just can't identify it. 422 00:27:45,830 --> 00:27:48,470 S3: What do you do? And he said, that's simple. We 423 00:27:48,470 --> 00:27:50,869 S3: just put the two photographs together, paste them on a 424 00:27:50,869 --> 00:27:55,160 S3: board and give them to the jury. And I sort 425 00:27:55,160 --> 00:27:58,400 S3: of went, what do you do? What I said, so 426 00:27:58,400 --> 00:28:03,230 S3: here you are, the FBI, who have all these trained 427 00:28:03,230 --> 00:28:08,720 S3: science scientists and forensic experts trained in facial identification. They 428 00:28:08,720 --> 00:28:11,540 S3: can't make the identification. But what you're asking the jury 429 00:28:11,540 --> 00:28:14,210 S3: is to make that identification for you, because you can't. 430 00:28:14,450 --> 00:28:18,650 S3: That's outrageous. That's that's biasing the evidence in a, in 431 00:28:18,650 --> 00:28:20,540 S3: a in a very big way. And a good friend 432 00:28:20,540 --> 00:28:24,139 S3: of mine, Gary Edmund from the unionist South Wales, he's 433 00:28:24,140 --> 00:28:27,740 S3: position that he I've heard him say quite often is 434 00:28:27,740 --> 00:28:31,940 S3: that particularly with fingerprints fingers fingerprints do the same. Fingerprint 435 00:28:31,940 --> 00:28:34,340 S3: experts will have the exemplar and the fingerprint and the 436 00:28:34,340 --> 00:28:37,730 S3: points of identification. Give that to the jury and they 437 00:28:37,730 --> 00:28:40,700 S3: will see, you know, ten or 20 or 15 points 438 00:28:40,700 --> 00:28:43,110 S3: of identification. The fingerprints. I don't have too much of 439 00:28:43,110 --> 00:28:47,820 S3: a problem with that. But. But Professor Edmond suggests that. Well, 440 00:28:47,820 --> 00:28:49,980 S3: if you've got the expert there, you don't need to 441 00:28:49,980 --> 00:28:53,490 S3: show them that photographic evidence. And I think he's got 442 00:28:53,490 --> 00:28:57,240 S3: a really strong point there that whether showing that sort 443 00:28:57,240 --> 00:29:00,660 S3: of visual aspect to the jury, does it better inform 444 00:29:00,660 --> 00:29:07,050 S3: them or is it influencing them in a wrong way? Another, um, 445 00:29:07,050 --> 00:29:09,930 S3: encounter I had, it was with the New South Wales Police. 446 00:29:09,930 --> 00:29:12,720 S3: I was asked by the chief scientists at the time 447 00:29:12,720 --> 00:29:17,729 S3: to do a review of the forensic forensic photography, a 448 00:29:17,730 --> 00:29:21,180 S3: forensic imaging unit in the New South Wales Police. And 449 00:29:21,180 --> 00:29:26,220 S3: I did that with another AFP senior AFP officer. And 450 00:29:26,610 --> 00:29:29,580 S3: we were shown around the the unit and we went 451 00:29:29,580 --> 00:29:32,250 S3: to this, uh, one room where they had a poster 452 00:29:32,910 --> 00:29:39,120 S3: and there was a CCTV photograph of two people holding 453 00:29:39,120 --> 00:29:41,580 S3: up another two people in a park. I think it 454 00:29:41,580 --> 00:29:46,230 S3: was in King's Cross. So then they had, uh, so 455 00:29:46,350 --> 00:29:48,810 S3: just try to visualise this poster. So this big image 456 00:29:48,810 --> 00:29:53,010 S3: in the middle was the CCTV image of the incident, 457 00:29:53,430 --> 00:29:57,450 S3: a still image. There was then close ups of each 458 00:29:57,450 --> 00:30:01,800 S3: person's head on the left hand side of the right 459 00:30:01,800 --> 00:30:05,100 S3: hand side. Sorry of the poster. Then on the other 460 00:30:05,100 --> 00:30:08,580 S3: opposite side of the poster was the two mug shots 461 00:30:08,580 --> 00:30:12,490 S3: of the people. And I said, well, so what's this 462 00:30:12,490 --> 00:30:16,150 S3: poster for? They said, oh, it's for court. Uh, identification 463 00:30:16,150 --> 00:30:19,240 S3: of the people. Oh, you say you've identified them? Oh, no, no, no, 464 00:30:19,240 --> 00:30:21,940 S3: we can't do that. Um, but we show them these 465 00:30:21,940 --> 00:30:25,150 S3: posters so that the jury can, you know, uh, correlate 466 00:30:25,330 --> 00:30:28,960 S3: between the the two close ups of the faces. So 467 00:30:28,960 --> 00:30:33,730 S3: this is an example of issues I've come across where 468 00:30:33,730 --> 00:30:38,200 S3: it's treating the jury as, uh, pseudo experts by showing 469 00:30:38,200 --> 00:30:42,460 S3: them images. And we all believe we're very visually literate. 470 00:30:42,460 --> 00:30:47,410 S3: We're a community that embellishes images all in everything we do. Even, 471 00:30:47,410 --> 00:30:49,600 S3: you know, we're doing that now. We're recording this on 472 00:30:49,600 --> 00:30:53,560 S3: on a camera, this conversation. So everything we do just 473 00:30:53,560 --> 00:30:57,940 S3: just about is recorded in some visual form. I have 474 00:30:57,940 --> 00:31:02,680 S3: some concerns when photographs are misrepresented to a jury in 475 00:31:02,680 --> 00:31:07,300 S3: a seemingly unbiased, unaltered way for them to make up 476 00:31:07,300 --> 00:31:11,230 S3: their mind. So using photographs, I think can be quite dangerous. 477 00:31:11,230 --> 00:31:14,110 S3: And there's no filter with this at the moment in 478 00:31:14,110 --> 00:31:19,180 S3: the criminal justice system, the judges, uh, that I've come across, uh, 479 00:31:19,480 --> 00:31:22,870 S3: believe that the more information you can give the jury, 480 00:31:22,870 --> 00:31:25,990 S3: the better the jury are to make the right decision. 481 00:31:26,230 --> 00:31:29,860 S3: I certainly agree with that. But whether images can help 482 00:31:29,860 --> 00:31:33,880 S3: or hinder the truth is something that's a bit concerning 483 00:31:33,880 --> 00:31:36,670 S3: when it comes to visual evidence in criminal cases. 484 00:31:37,480 --> 00:31:41,770 S2: There was a case that involved Larry David's show, Curb 485 00:31:41,770 --> 00:31:46,480 S2: Your Enthusiasm, and the role that that played in a 486 00:31:46,480 --> 00:31:49,330 S2: criminal trial. Can you go through that for us? 487 00:31:49,660 --> 00:31:53,230 S3: Sure. Um, I wrote about it in my PhD thesis 488 00:31:53,230 --> 00:31:56,380 S3: many years ago, but it's, um, it's a bit of 489 00:31:56,380 --> 00:32:02,560 S3: a reverse, uh, identification. There was a gang murder, and 490 00:32:02,560 --> 00:32:06,100 S3: one of the witnesses to that gang murder was actually 491 00:32:06,100 --> 00:32:11,200 S3: murdered herself. And the accused of the original murder trial. 492 00:32:11,230 --> 00:32:15,850 S3: His brother was charged with her murder, and he claimed 493 00:32:15,850 --> 00:32:18,670 S3: that it couldn't have been him because he was at 494 00:32:18,670 --> 00:32:22,810 S3: the baseball with his daughter. And he produced the seating 495 00:32:22,810 --> 00:32:25,030 S3: stubs to the police. And they said, well, you could 496 00:32:25,030 --> 00:32:28,720 S3: have got this off anyone. They tried to look it 497 00:32:28,720 --> 00:32:33,190 S3: up with the CCTV cameras of the stadium, and they 498 00:32:33,190 --> 00:32:35,830 S3: saw where his seat was, and they tried to enlarge it, 499 00:32:35,830 --> 00:32:40,030 S3: but it was just too it just didn't allow definition 500 00:32:40,030 --> 00:32:43,240 S3: that they could identify him sitting there with his daughter. 501 00:32:43,600 --> 00:32:47,800 S3: But for some, uh, strange reason, Larry David was shooting, 502 00:32:47,800 --> 00:32:51,910 S3: I think, Curb Your Enthusiasm show. And there was the 503 00:32:51,910 --> 00:32:54,490 S3: accused in the background, sitting in his seat. 504 00:32:54,610 --> 00:32:55,510 S2: With a good quality. 505 00:32:55,510 --> 00:33:00,370 S3: Film. Yeah, broadcast quality film. And they use that. And 506 00:33:00,370 --> 00:33:03,670 S3: of course, they got, um, uh, statements off the producers 507 00:33:03,670 --> 00:33:05,500 S3: of the time and the and the time and date 508 00:33:05,500 --> 00:33:09,670 S3: stamps on the, on the video. Um, and his alibi 509 00:33:09,670 --> 00:33:12,520 S3: then was confirmed that he was at the baseball through 510 00:33:12,670 --> 00:33:16,480 S3: the happened chance of being in the background of Larry 511 00:33:16,480 --> 00:33:19,240 S3: David's show. So quite a quite an interesting. 512 00:33:19,300 --> 00:33:19,930 S2: High definition. 513 00:33:19,930 --> 00:33:22,810 S3: In high definition. Yeah, yeah. And he was telling the truth. 514 00:33:22,810 --> 00:33:25,570 S2: So it can actually be used to exonerate. 515 00:33:25,720 --> 00:33:28,870 S3: It can be used to exonerate. Absolutely. Any forensic evidence 516 00:33:28,870 --> 00:33:33,190 S3: is about the first of all, trying to exonerate innocent 517 00:33:33,190 --> 00:33:37,840 S3: people and secondly, trying to, uh, to convict or use 518 00:33:37,840 --> 00:33:41,200 S3: to in the conviction of perpetrators of crime. But the 519 00:33:41,200 --> 00:33:44,500 S3: primary objective is always to try to eliminate first, not 520 00:33:44,500 --> 00:33:48,250 S3: not identify. Being an academic in forensic science at Western 521 00:33:48,250 --> 00:33:51,970 S3: Sydney University for several years, being one of the founding 522 00:33:51,970 --> 00:33:55,900 S3: academics or the founding academic there. For that course, we 523 00:33:55,900 --> 00:33:59,950 S3: built a crime scene facility with the New South Wales Police, 524 00:33:59,950 --> 00:34:03,820 S3: where students would go through and actually process scenes of crime. 525 00:34:04,120 --> 00:34:09,219 S3: Forensic photography is more of the specialisation, the the enhancement 526 00:34:09,219 --> 00:34:13,989 S3: of the of the evidence. Uh, the understanding reading of 527 00:34:13,989 --> 00:34:19,660 S3: how vision works and how it how it affects the interpretation. Um, 528 00:34:19,660 --> 00:34:23,049 S3: I think that's more about forensic crime scene photography really 529 00:34:23,050 --> 00:34:27,850 S3: is more about recording in a document style, um, in 530 00:34:27,850 --> 00:34:31,360 S3: a vernacular kind of photography, how the scene was found, 531 00:34:31,360 --> 00:34:35,680 S3: items of evidence, murder, weapons, shoe marks and so forth. 532 00:34:35,680 --> 00:34:38,410 S3: There are particular ways that they need to be photographed, 533 00:34:38,410 --> 00:34:41,530 S3: but what a crime scene officer does in the processing 534 00:34:41,530 --> 00:34:45,040 S3: of a scene is document. And they would they would 535 00:34:45,040 --> 00:34:49,990 S3: document in images, usually video or still photography. They would 536 00:34:49,989 --> 00:34:53,890 S3: write contemporaneous notes on where they saw things and how 537 00:34:53,890 --> 00:34:57,370 S3: they found it, and also sketching, which is another visual, 538 00:34:57,370 --> 00:35:01,029 S3: cultural sort of aspect to understanding crime scenes. So crime 539 00:35:01,030 --> 00:35:03,670 S3: scene investigators would do those, those three things. They would 540 00:35:03,670 --> 00:35:07,810 S3: also collect the evidence, but say shoe marks and fingerprints, 541 00:35:07,810 --> 00:35:11,270 S3: items that are going to be analyzed. By other forensic 542 00:35:11,270 --> 00:35:15,770 S3: experts need to be taken in a particular way so 543 00:35:15,770 --> 00:35:21,230 S3: that distortion and artifacts are eliminated. And that's where forensic 544 00:35:21,230 --> 00:35:26,299 S3: photography training has to occur with crime scene investigators. So 545 00:35:26,300 --> 00:35:29,660 S3: they do do some, I guess, forensic photography on site 546 00:35:29,660 --> 00:35:33,830 S3: when it comes to recording, documenting those types of evidence. 547 00:35:34,250 --> 00:35:39,230 S3: But in a whole it's more around just recording the 548 00:35:39,230 --> 00:35:44,600 S3: site as it's found and the critical elements of the site. 549 00:35:44,870 --> 00:35:48,470 S3: They sort of work in a photo narrative form more 550 00:35:48,469 --> 00:35:52,790 S3: so than anything, and they would use filmic techniques. So 551 00:35:52,790 --> 00:35:55,310 S3: for instance, if there was a shoe mark in blood, 552 00:35:55,430 --> 00:35:59,359 S3: they would photograph it using a style of photography, but 553 00:35:59,360 --> 00:36:02,000 S3: in a way that that a filmmaker would approach it. 554 00:36:02,000 --> 00:36:04,460 S3: So you do what we call an establishing shot. So 555 00:36:04,460 --> 00:36:07,100 S3: you do a wide shot to show the location of 556 00:36:07,100 --> 00:36:08,989 S3: the shoe mark in blood, and then you do a 557 00:36:08,989 --> 00:36:11,810 S3: close up shot, which would be the forensic kind of 558 00:36:11,810 --> 00:36:14,480 S3: where you would put a scale down. You would make 559 00:36:14,480 --> 00:36:17,930 S3: sure that it's perpendicular to the subject. You would use 560 00:36:17,930 --> 00:36:21,440 S3: lighting to show that the, the maximum detail of the, 561 00:36:21,440 --> 00:36:24,530 S3: of the shoe mark. So it's very much a filmic 562 00:36:24,530 --> 00:36:29,480 S3: kind of visual narrative that crime scene operatives work on 563 00:36:29,480 --> 00:36:33,290 S3: in combination with their contemporaneous notes. And they use tools 564 00:36:33,290 --> 00:36:37,640 S3: like markers, evidence markers to show and relate the photographs 565 00:36:37,640 --> 00:36:40,850 S3: in the sketch to the contemporaneous notes. That's all a 566 00:36:40,850 --> 00:36:44,810 S3: way of communicating a bloody shoe mark. That could be 567 00:36:44,870 --> 00:36:48,890 S3: a item I found in the the lounge room or whatever, 568 00:36:48,890 --> 00:36:53,660 S3: but that marker that's placed next to the the shoe mark, 569 00:36:53,660 --> 00:36:57,380 S3: that item of evidence is only there really, so that 570 00:36:57,380 --> 00:37:01,730 S3: the contemporaneous notes and the sketches and the photographs all 571 00:37:01,730 --> 00:37:03,920 S3: tie in in a communicative way. And you. 572 00:37:03,920 --> 00:37:05,540 S2: Can get a sense of scale and you can. 573 00:37:05,540 --> 00:37:08,210 S3: Get a sense of scale. So we use linear scales 574 00:37:08,210 --> 00:37:12,140 S3: to do that. Um, and, and, and all sorts of things. Yeah. 575 00:37:12,320 --> 00:37:14,660 S3: And now with drones, you know, you can get aerial 576 00:37:14,660 --> 00:37:18,440 S3: shots as well. Um, New South Wales Police, we were 577 00:37:18,440 --> 00:37:22,160 S3: very big on using photogrammetry as well, where they would 578 00:37:22,160 --> 00:37:27,440 S3: go in and use stereo cameras and record it, particularly for, uh, 579 00:37:27,440 --> 00:37:33,170 S3: motor vehicle accidents particularly, and then reproduce that accurately. So 580 00:37:33,170 --> 00:37:37,040 S3: measurements can be taken from their line drawings very similar 581 00:37:37,040 --> 00:37:39,560 S3: to an old plan of a building. It's they sort 582 00:37:39,560 --> 00:37:42,680 S3: of they can draw scenes in that sort of line sketch, 583 00:37:42,680 --> 00:37:45,890 S3: but it's to scale and things can be measured. 584 00:37:47,050 --> 00:37:52,089 S2: We've actually seen lots of visual illusions, and there's been 585 00:37:52,090 --> 00:37:55,360 S2: so much fighting on the internet about, for example, the 586 00:37:55,360 --> 00:37:58,480 S2: dress there was blue and gold and whatever people would 587 00:37:58,480 --> 00:38:02,230 S2: argue to the death that they were totally different colours 588 00:38:02,230 --> 00:38:04,570 S2: from what someone else was seeing. Can you give us 589 00:38:04,570 --> 00:38:09,310 S2: an example of another visual illusion that people can actually understand? 590 00:38:09,310 --> 00:38:13,960 S2: How our minds can be tricked, even though we're convinced otherwise? 591 00:38:14,530 --> 00:38:18,580 S3: Yeah, there's lots of examples, Cathy. Um, I give a 592 00:38:18,580 --> 00:38:22,720 S3: lecture in forensic photography or forensic criminology subject that I 593 00:38:22,719 --> 00:38:26,140 S3: teach here at the university on how to interpret images. 594 00:38:26,140 --> 00:38:28,600 S3: And basically what I'm trying to show the students is 595 00:38:28,600 --> 00:38:33,910 S3: the ambiguity, how ambiguous a lot of the the vision is. 596 00:38:33,940 --> 00:38:37,930 S3: An example is a horse race, uh, photo finish. You 597 00:38:37,930 --> 00:38:40,480 S3: you would believe that they're in different places of the track, 598 00:38:40,480 --> 00:38:42,940 S3: but in actual fact, the way that it's recorded through 599 00:38:42,940 --> 00:38:45,940 S3: a slot, the horses are at the same location, but 600 00:38:45,940 --> 00:38:49,750 S3: at different time. So in this type of recording, the 601 00:38:49,750 --> 00:38:53,529 S3: variation is the time, not the distance, but our interpretation 602 00:38:53,530 --> 00:38:56,500 S3: is the distance because we see a horse in front 603 00:38:56,500 --> 00:38:59,890 S3: and a horse behind that the horse in in the crossing. 604 00:38:59,890 --> 00:39:02,590 S3: The line first is though is in front of the 605 00:39:02,590 --> 00:39:06,700 S3: horse coming second. But in actual fact, the photo, the 606 00:39:06,700 --> 00:39:10,719 S3: imagery is all taken at the same location. The difference? 607 00:39:10,719 --> 00:39:15,730 S3: There is time. But another great example is a checkerboard 608 00:39:15,730 --> 00:39:21,430 S3: example by, um, a researcher called Edward Allison, who uses 609 00:39:21,430 --> 00:39:27,340 S3: a visual anomaly where there's a if you can vision 610 00:39:27,430 --> 00:39:30,340 S3: a black and white checkerboard or a chessboard like, well, 611 00:39:30,340 --> 00:39:34,480 S3: like we all know, and there's a green cylinder sitting 612 00:39:34,480 --> 00:39:39,609 S3: on the, uh, checkerboard casting a shadow over the checkerboard. 613 00:39:39,610 --> 00:39:44,500 S3: And there are two, uh, squares that Alderson identifies, square 614 00:39:44,500 --> 00:39:48,640 S3: A and square B, and the question is, is, is 615 00:39:48,640 --> 00:39:52,089 S3: the tone of A and B square A and B 616 00:39:52,120 --> 00:39:56,860 S3: the same or different? Now, unlike the dress that has people, 617 00:39:56,860 --> 00:40:02,620 S3: C perceive the colors differently in this experiment, everyone that 618 00:40:02,620 --> 00:40:07,210 S3: I'm aware of sees the two squares as different tones, 619 00:40:07,210 --> 00:40:10,960 S3: blatantly different tones, and they would believe and swear that 620 00:40:10,960 --> 00:40:14,530 S3: they are different tones. But if you were to isolate 621 00:40:14,530 --> 00:40:19,870 S3: everything else and just look at those two squares like 622 00:40:19,870 --> 00:40:22,239 S3: I do in my class, you'll see that that they're 623 00:40:22,239 --> 00:40:26,259 S3: exactly the same tone of gray. I've even use Photoshop 624 00:40:26,260 --> 00:40:30,339 S3: and measured the A square and the B square, and 625 00:40:30,340 --> 00:40:35,950 S3: I get the same RGB value, uh, in both. So empirically, 626 00:40:35,950 --> 00:40:39,670 S3: I've also proven that it's the same tonal range, but 627 00:40:39,670 --> 00:40:42,190 S3: if you see it, uh, you would believe that a 628 00:40:42,190 --> 00:40:47,770 S3: square A and square B is absolutely different tonality. But 629 00:40:47,770 --> 00:40:51,219 S3: the reality is it's the same. And I'll go back 630 00:40:51,219 --> 00:40:54,399 S3: to the Errol Morris quote, believing is seeing. It's not 631 00:40:54,400 --> 00:40:57,490 S3: the other way around. You believe that it's different, but 632 00:40:57,489 --> 00:40:59,020 S3: the facts are they're the same. 633 00:40:59,620 --> 00:41:03,370 S2: I think my brain is just absolutely mashed up now 634 00:41:03,370 --> 00:41:08,770 S2: because I thought photographs were definitive, and now I'm actually 635 00:41:08,770 --> 00:41:12,250 S2: thinking it's all an illusion. Thank you so much, Glenn, 636 00:41:12,250 --> 00:41:18,219 S2: for actually getting us thinking about photographs and reliability and 637 00:41:18,219 --> 00:41:24,130 S2: evidence and facial recognition versus identification. It is so complicated. 638 00:41:24,340 --> 00:41:27,250 S2: Thanks very much for joining us and explaining all of 639 00:41:27,250 --> 00:41:28,060 S2: that today. 640 00:41:28,090 --> 00:41:29,710 S3: You most welcome, Kathy anytime. 641 00:41:37,930 --> 00:41:42,400 S2: Crime Insider's Forensics is a listener. Original production. It's hosted 642 00:41:42,400 --> 00:41:45,670 S2: by me, Catherine Fox, and is produced by Ed Gordon. 643 00:41:46,120 --> 00:41:48,759 S2: Sound Design and imaging is by Link Kelly.