1 00:00:03,200 --> 00:00:05,880 Welcome to Probation Perspectives insights into 2 00:00:05,880 --> 00:00:08,960 Community Justice, the podcast where we dive into the world of 3 00:00:08,960 --> 00:00:11,760 probation, rehabilitation and community justice. 4 00:00:12,160 --> 00:00:15,040 Each episode we explore the themes, challenges and skills 5 00:00:15,040 --> 00:00:17,280 that underpin effective probation practise. 6 00:00:17,800 --> 00:00:20,560 Whether you're just starting out on the peak WHIP, have years of 7 00:00:20,560 --> 00:00:23,480 experience in the field, or are simply curious about the 8 00:00:23,480 --> 00:00:26,680 criminal justice system, this podcast is your space for 9 00:00:26,680 --> 00:00:30,000 practical insight, expert perspectives, and meaningful 10 00:00:30,000 --> 00:00:33,080 conversations. In today's episode, we're joined 11 00:00:33,080 --> 00:00:35,720 by Professor Mike Nellis, Professor of Criminal and 12 00:00:35,720 --> 00:00:39,000 Community Justice at the University of Strathclyde, and 13 00:00:39,000 --> 00:00:42,080 Louis All Good from the AI and Justice Government Department. 14 00:00:42,560 --> 00:00:45,240 Together, we explore the evolving role of technology in 15 00:00:45,240 --> 00:00:47,960 probation, focusing on electronic monitoring and the 16 00:00:47,960 --> 00:00:50,400 growing influence of artificial intelligence. 17 00:00:51,640 --> 00:00:54,960 Electronic monitoring is a technology in its original form 18 00:00:54,960 --> 00:00:59,520 for monitoring the locations and schedules of remanded, convicted 19 00:00:59,600 --> 00:01:03,280 or released offenders, and the initial version of it, which 20 00:01:03,320 --> 00:01:06,040 like most versions of electronic monitoring have originated in 21 00:01:06,040 --> 00:01:10,320 America, is just radio frequency curfew technology. 22 00:01:11,520 --> 00:01:14,320 We talk about the origins and development of electronic 23 00:01:14,320 --> 00:01:18,120 monitoring, the promises and pitfalls of AI and justice, and 24 00:01:18,120 --> 00:01:21,680 what these technologies mean for P quips probation officers and 25 00:01:21,680 --> 00:01:23,560 the future of the probation service. 26 00:01:25,040 --> 00:01:28,480 So Gentive AI in particular, how can we help provide assistance 27 00:01:28,520 --> 00:01:31,280 that can do things like help you spend more time listening to the 28 00:01:31,280 --> 00:01:34,720 personal probation and less time manually writing up notes, more 29 00:01:34,720 --> 00:01:37,720 time analysing information, joining the dots together, less 30 00:01:37,720 --> 00:01:40,400 time having to reread the same information again? 31 00:01:41,520 --> 00:01:44,160 It's a timely and thought provoking conversation that 32 00:01:44,160 --> 00:01:47,320 invites us to reflect on how technology is reshaping 33 00:01:47,320 --> 00:01:50,800 probation and what it means to practise ethically in a digital 34 00:01:50,800 --> 00:01:53,200 age. Let's get into our conversation 35 00:01:53,200 --> 00:02:01,960 with Mike and Louie. Great. 36 00:02:01,960 --> 00:02:04,360 So thank you so much for being here today. 37 00:02:04,600 --> 00:02:07,040 If you could both introduce yourselves and tell us a bit 38 00:02:07,040 --> 00:02:09,120 about your background and your current work. 39 00:02:09,120 --> 00:02:11,080 Mike, are we happy to start with you, please? 40 00:02:11,200 --> 00:02:12,800 Well, thank you very much, Megan Kerry. 41 00:02:12,800 --> 00:02:15,800 It's very nice to be here. I'm Mike Nelis, and I'm an 42 00:02:15,800 --> 00:02:18,560 emeritus professor, a retired professor of criminal and 43 00:02:18,560 --> 00:02:21,800 community justice at the University of Strathclyde, and I 44 00:02:21,800 --> 00:02:24,080 am retired. But every now and again I get 45 00:02:24,080 --> 00:02:27,840 quite involved in things, and AI turns out to be one of them. 46 00:02:27,840 --> 00:02:31,160 Over the years I've done work with the Council of Europe on 47 00:02:31,160 --> 00:02:34,120 electronic monitoring, which I was very pleased with and which 48 00:02:34,120 --> 00:02:38,120 seemed to go quite well. But when I was asked to work 49 00:02:38,120 --> 00:02:40,360 with the Council of Europe committee on an ethical 50 00:02:40,360 --> 00:02:43,400 recommendation about AI in prisons and probation, I didn't 51 00:02:43,400 --> 00:02:45,080 feel quite as comfortable with that. 52 00:02:45,080 --> 00:02:49,520 So I left after a year, but nonetheless got the book about 53 00:02:49,680 --> 00:02:52,720 AI and criminal justice, prisons and probation, and I've remained 54 00:02:52,720 --> 00:02:55,720 interested in it ever since. And I started to write about it, 55 00:02:55,840 --> 00:02:58,400 linking it to my work on electronic monitoring. 56 00:02:58,480 --> 00:03:01,960 Thank you so much, Mike. And then Louis, happy to say a 57 00:03:01,960 --> 00:03:04,400 bit about yourself, please. Of course, lovely to meet you 58 00:03:04,400 --> 00:03:05,920 all. I'm Louis Orgut. 59 00:03:05,920 --> 00:03:08,240 I'm an AI product manager from the Justice AI units. 60 00:03:08,240 --> 00:03:10,120 We work directly within the membership Justice. 61 00:03:10,120 --> 00:03:14,400 We are building solutions and also partnering of some of the 62 00:03:14,880 --> 00:03:17,480 best companies out there from the likes of Open AI to 63 00:03:17,480 --> 00:03:20,400 Microsoft, to give people the most powerful and most safest 64 00:03:20,400 --> 00:03:22,640 tools. I lead the work on speech AI, 65 00:03:22,640 --> 00:03:26,000 looking at opportunities that are either speech to text, 66 00:03:26,000 --> 00:03:28,880 speech to speech, and beyond. Specifically working a lot with 67 00:03:28,880 --> 00:03:32,040 probation offices. I want to help them spend more 68 00:03:32,040 --> 00:03:35,760 time being analytical, more time actively listening, less time 69 00:03:35,760 --> 00:03:38,440 note taking. Taking conversations into 70 00:03:38,440 --> 00:03:41,240 structured notes for example, is been a super big interest to me 71 00:03:41,240 --> 00:03:43,480 of late. And yeah, anything that helps to 72 00:03:43,480 --> 00:03:45,720 support professional judgement is awesome in my mind. 73 00:03:45,960 --> 00:03:48,040 And helping with the relationships that are truly at 74 00:03:48,040 --> 00:03:49,920 the heart of probation. AI can only support but not 75 00:03:49,920 --> 00:03:50,720 replace. Great. 76 00:03:50,720 --> 00:03:51,360 Thank you. Both. 77 00:03:51,800 --> 00:03:54,640 We've punched AI there and electronic monitoring. 78 00:03:54,640 --> 00:03:57,480 So there are terms that a lot of us might hear, but maybe not 79 00:03:57,480 --> 00:03:59,920 everyone fully understands. Are you both able just to 80 00:03:59,920 --> 00:04:02,000 explain both of those things for our listening? 81 00:04:02,120 --> 00:04:04,440 OK, well I'll, I'll do the electronic monitoring one. 82 00:04:04,440 --> 00:04:08,040 I'll leave the AI 12 LUI. Electronic monitoring is a 83 00:04:08,040 --> 00:04:11,600 technology for in its original form for monitoring the 84 00:04:11,600 --> 00:04:16,600 locations and schedules of remanded, convicted or released 85 00:04:16,600 --> 00:04:19,040 offenders. And the initial version of it, 86 00:04:19,040 --> 00:04:21,399 which like most versions of electronic monitoring have 87 00:04:21,399 --> 00:04:25,640 originated in America, is just radio frequency curfew 88 00:04:25,640 --> 00:04:27,840 technology. You put an ankle bracelet on 89 00:04:27,840 --> 00:04:31,800 someone and you ask them to stay in proximity to a reporting 90 00:04:31,800 --> 00:04:35,280 device that's installed in their own homes so that the monitoring 91 00:04:35,280 --> 00:04:38,680 authority will know whether you're in your house or not by 92 00:04:38,680 --> 00:04:42,400 dint of the connection that has been made by your ankle bracelet 93 00:04:42,800 --> 00:04:44,800 and the device that's installed in your home. 94 00:04:45,040 --> 00:04:48,760 GPS tracking also requires an ankle bracelet, but it means 95 00:04:48,760 --> 00:04:52,040 your movements can be followed. It's GPS tracking because it 96 00:04:52,040 --> 00:04:56,560 makes use of GPS satellites, which were developed for various 97 00:04:56,560 --> 00:05:00,000 navigation purposes, some military, some civilian, as far 98 00:05:00,000 --> 00:05:03,040 back as the 1970s. But they weren't accurate enough 99 00:05:03,040 --> 00:05:07,400 to do the location tracking of individuals until the 1990s. 100 00:05:07,520 --> 00:05:10,320 And ever since then, GPS tracking has been developing 101 00:05:10,320 --> 00:05:12,640 around the world. So you can monitor movements and 102 00:05:12,640 --> 00:05:14,520 you can also create exclusion zones. 103 00:05:14,520 --> 00:05:16,760 You can tell people where they can't go. 104 00:05:16,840 --> 00:05:20,160 And more recently we've had alcohol monitoring, which 105 00:05:20,160 --> 00:05:23,400 doesn't monitoring the locations, but it monitors your 106 00:05:23,440 --> 00:05:26,760 intake of alcohol and that's particularly useful if the court 107 00:05:26,760 --> 00:05:29,960 has banned you from using it. It's time stamped so we know 108 00:05:29,960 --> 00:05:33,280 what from it, what times of day you are drinking alcohol or not 109 00:05:33,280 --> 00:05:36,160 drinking alcohol. Some technologies can combine 110 00:05:36,160 --> 00:05:39,160 the 2 alcohol monitoring and location monitoring. 111 00:05:39,680 --> 00:05:42,320 Thank you, Mike. And just as a follow on, as some 112 00:05:42,320 --> 00:05:45,400 of the public might think those things are quite reliable, maybe 113 00:05:45,400 --> 00:05:47,000 they think they're not that reliable. 114 00:05:47,000 --> 00:05:49,800 I just wondered, as you've done a lot of research in this area, 115 00:05:49,800 --> 00:05:52,600 what's your thoughts on the reliability of those? 116 00:05:53,120 --> 00:05:55,320 Right. Well, I have to say that they 117 00:05:55,440 --> 00:05:59,200 are reliable most of the time, but they're not perfect. 118 00:05:59,280 --> 00:06:03,760 And in those instances when the technology has failed, can that 119 00:06:03,760 --> 00:06:06,400 can have quite devastating and fateful consequences for the 120 00:06:06,400 --> 00:06:09,640 individuals concerned. So we shouldn't go around 121 00:06:09,640 --> 00:06:12,800 thinking, you know, in the manner of Fujitsu and the post 122 00:06:12,800 --> 00:06:14,920 office, that the technology is always right. 123 00:06:15,000 --> 00:06:18,560 And if we have anomalies with offenders who say, but you know, 124 00:06:18,560 --> 00:06:21,440 you're saying I was here when I did this burglary when I'm 125 00:06:21,440 --> 00:06:25,000 saying I was somewhere else, that there must be data to to 126 00:06:25,000 --> 00:06:27,840 confirm my innocence in this, we should listen to that because 127 00:06:27,840 --> 00:06:29,840 the technology doesn't always get it right. 128 00:06:29,840 --> 00:06:33,080 But I don't want to use an argument to resist the any of 129 00:06:33,080 --> 00:06:36,080 these technologies by saying they won't work properly, 130 00:06:36,160 --> 00:06:39,160 because for better or worse, a lot of the time they do work 131 00:06:39,160 --> 00:06:40,600 properly. Thank you, Mike. 132 00:06:40,600 --> 00:06:43,280 That's really helpful. And then, Louis, are you happy 133 00:06:43,280 --> 00:06:46,000 to explain what you mean by AI? Absolutely. 134 00:06:46,040 --> 00:06:49,200 There's two flavours of AI. Typically you'd see things like 135 00:06:49,520 --> 00:06:52,400 machine learning and modelling, and this is that type of AI, 136 00:06:52,400 --> 00:06:54,960 which I think a lot of people wouldn't realise, but they use 137 00:06:54,960 --> 00:06:57,040 it, they stay as is. If you're using the weather 138 00:06:57,040 --> 00:07:00,160 forecast every on Netflix, every on Amazon, it's learning from a 139 00:07:00,160 --> 00:07:02,880 pattern to predict or recommend the next thing. 140 00:07:02,920 --> 00:07:05,560 If you watch a particular show on Netflix, you're likely to 141 00:07:05,560 --> 00:07:08,160 potentially like something else. So that's one type of AI. 142 00:07:08,240 --> 00:07:10,440 And then the one that's caught everyone's attention recently is 143 00:07:10,440 --> 00:07:12,320 gender of AI. And it's gender AI. 144 00:07:12,440 --> 00:07:14,640 And I think the nicest way to frame is we're still learning 145 00:07:14,640 --> 00:07:17,080 from patterns, but it's generating new content. 146 00:07:17,080 --> 00:07:19,720 You don't use Gen AI to create a new type of weather because 147 00:07:19,720 --> 00:07:21,760 that's like can do for chance of me balls. 148 00:07:21,840 --> 00:07:24,600 That's not the case here. So gent of AI think about 149 00:07:24,600 --> 00:07:28,920 drafting text, summarising things, creating content based 150 00:07:28,920 --> 00:07:31,440 on an input itself. And the most popular version of 151 00:07:31,480 --> 00:07:33,880 gent of AI tools would be ChatGPT. 152 00:07:34,400 --> 00:07:36,560 Don't click on the figures now that there was something like 153 00:07:36,560 --> 00:07:40,280 10% of the world's population have used chat for GBT, which is 154 00:07:40,280 --> 00:07:43,120 a super massive number. Yeah, if you hear of ChatGPT Gen 155 00:07:43,120 --> 00:07:45,680 AI, if you're using the weather forecast, that's more classic 156 00:07:45,800 --> 00:07:48,360 predictive modelling, but that's the best way to describe it. 157 00:07:49,080 --> 00:07:51,400 It's worth asking if anyone ever says what do you think about AI? 158 00:07:51,680 --> 00:07:53,880 Might be worth checking. What type of AI do you mean in 159 00:07:53,880 --> 00:07:55,160 this? That's really interesting. 160 00:07:55,160 --> 00:07:59,400 So then, Louis, the use of AI in probation, we're really 161 00:07:59,400 --> 00:08:02,680 interested in hearing what you've got to say about how it's 162 00:08:02,680 --> 00:08:05,720 being developed and rolled out in probation. 163 00:08:06,000 --> 00:08:09,600 So in my mind, I think of like sort of three buckets about what 164 00:08:09,600 --> 00:08:11,400 this actually means for probation. 165 00:08:11,400 --> 00:08:14,160 So firstly, you know, when people join the profession, 166 00:08:14,240 --> 00:08:16,840 they're super excited to build loving, kind relationships. 167 00:08:16,920 --> 00:08:20,080 They're super excited to help create a safer community, 168 00:08:20,200 --> 00:08:22,640 prevent risk, etcetera. And yet a lot of people get 169 00:08:22,640 --> 00:08:25,800 shocked by the amount of admin. So Gentive AI in particular, how 170 00:08:25,800 --> 00:08:29,080 can we help provide assistance that can do things like help you 171 00:08:29,080 --> 00:08:32,159 spend more time listening to the personal probation and less time 172 00:08:32,200 --> 00:08:35,240 manually writing up notes, more time analysing information, 173 00:08:35,320 --> 00:08:38,600 joining the dots together less time having to reread the same 174 00:08:38,600 --> 00:08:41,919 information again. So one bucket giving people and 175 00:08:41,919 --> 00:08:45,040 retain them time to the things that are most interesting to 176 00:08:45,040 --> 00:08:46,960 them and most useful to them. More time with the people of 177 00:08:46,960 --> 00:08:49,760 creation in the meeting directly looking at them actually 178 00:08:49,760 --> 00:08:52,280 listening. Second bucket is more about like 179 00:08:52,800 --> 00:08:55,480 faster great ways of like finding information. 180 00:08:55,480 --> 00:08:58,360 I think it was last week I met someone who was telling me it's 181 00:08:58,360 --> 00:09:01,040 crazy hard sometimes because you know, we've got people who have 182 00:09:01,040 --> 00:09:03,160 had 20, fourish offences, they've got so much 183 00:09:03,160 --> 00:09:05,320 documentation about them, so much information. 184 00:09:05,320 --> 00:09:07,640 It's really hard to find the things that are most important 185 00:09:07,640 --> 00:09:10,160 to me. So sometimes a thing that's 186 00:09:10,160 --> 00:09:12,360 really useful in this context is using things like semantic 187 00:09:12,360 --> 00:09:15,320 search, using AI to say you didn't search the direct word 188 00:09:15,360 --> 00:09:20,000 of, you know, drug offence, but by searching drugs you might 189 00:09:20,000 --> 00:09:22,400 have noticed all of the information about this person 190 00:09:22,480 --> 00:09:25,160 and their relationship with drugs called etcetera. 191 00:09:25,240 --> 00:09:28,800 So supporting with consistent school decisions, it's super 192 00:09:28,800 --> 00:09:31,400 useful, especially in the context of using semantic search 193 00:09:31,560 --> 00:09:34,040 and a bit that's touches upon electronic monitoring as well. 194 00:09:34,320 --> 00:09:37,320 It's not gentive AI, but more about classic machine learning, 195 00:09:37,840 --> 00:09:40,600 predictive models. We can use analytics to support 196 00:09:40,600 --> 00:09:44,080 patterns, risk patterns, vulnerability signals, emerging 197 00:09:44,080 --> 00:09:46,160 risks, etcetera. We've got people who work with 198 00:09:46,160 --> 00:09:49,360 probation staff and they're so busy that sometimes happen them 199 00:09:49,560 --> 00:09:53,040 with nudges or information that can inform a decision can be 200 00:09:53,040 --> 00:09:55,320 super useful. But again, it's a decision that 201 00:09:55,320 --> 00:09:57,480 the person who's doing the job are building. 202 00:09:57,480 --> 00:09:59,360 Probation officers, they're the ones who are most informed. 203 00:09:59,720 --> 00:10:02,600 This technology is only as supportive in our context. 204 00:10:02,640 --> 00:10:05,120 Admin time, back to people I'm living and bringing that with 205 00:10:05,120 --> 00:10:07,920 speech AI for sure. Faster, clearer information 206 00:10:07,920 --> 00:10:09,560 handling we're tackling with semantic search. 207 00:10:09,600 --> 00:10:12,200 Smarter triage I think is more about the broader stuff that 208 00:10:12,200 --> 00:10:14,480 happens as is in regards to data modelling work. 209 00:10:14,520 --> 00:10:17,000 Some of these solutions come in the form of bespoke builds. 210 00:10:17,000 --> 00:10:19,720 Some of these solutions, we're finding, are covered by some of 211 00:10:19,720 --> 00:10:22,800 the more common generative AI tools which can be used 212 00:10:22,800 --> 00:10:25,040 effectively and safely with appropriate training and 213 00:10:25,040 --> 00:10:27,400 adoption support. Yeah, you've given a really 214 00:10:27,400 --> 00:10:29,880 interesting summary there, and a lot of this information is set 215 00:10:29,880 --> 00:10:34,320 out in the AI Plan for Justice that was published over the 216 00:10:34,320 --> 00:10:37,120 summer, and it really captures some of that key information. 217 00:10:37,120 --> 00:10:41,040 But what really speaks to me is that recognition of how busy 218 00:10:41,040 --> 00:10:45,800 probation officers are, how much they're juggling, and how you're 219 00:10:45,800 --> 00:10:49,600 working in those offices and you work with those individuals and 220 00:10:49,600 --> 00:10:51,480 you learn from them about what they need. 221 00:10:51,560 --> 00:10:54,760 And I think that's really important to learn their job and 222 00:10:54,760 --> 00:10:59,200 understand how they can reduce time spent on the admin and 223 00:10:59,200 --> 00:11:02,680 develop time for professional judgement and those sorts of 224 00:11:02,680 --> 00:11:04,520 skills. And then I think that might 225 00:11:04,520 --> 00:11:07,960 speak to some of Mike's concerns as well about rollout of AI. 226 00:11:07,960 --> 00:11:12,440 Well, I mean, I think the relational core of probation is 227 00:11:12,440 --> 00:11:14,440 something that is worth preserving. 228 00:11:14,440 --> 00:11:17,480 And if one didn't preserve the relational core of it, I guess 229 00:11:17,480 --> 00:11:20,680 one wouldn't want to be calling it probation, you know, just 230 00:11:20,680 --> 00:11:23,280 calling it Offender Management rather than using the word 231 00:11:23,280 --> 00:11:26,960 probation with all the the moral aspects of the probation 232 00:11:26,960 --> 00:11:30,320 tradition that the term itself incorporates. 233 00:11:30,320 --> 00:11:33,880 And with electronic monitoring over the years, the 20 or 30 234 00:11:33,880 --> 00:11:36,800 years that I was involved with that I was the person who was 235 00:11:36,800 --> 00:11:40,640 championing uses of electronic monitoring in probation. 236 00:11:40,640 --> 00:11:43,400 And certainly in the early days I didn't make myself very 237 00:11:43,400 --> 00:11:45,960 popular in doing that. But I, I was always aware that 238 00:11:45,960 --> 00:11:49,480 electronic monitoring could be used in better or worse ways. 239 00:11:49,520 --> 00:11:52,840 And my belief that electronic monitoring could be used to 240 00:11:52,840 --> 00:11:55,960 support relational work, support, rehabilitative work, 241 00:11:55,960 --> 00:11:59,960 support resistance was the basis on which I thought probation 242 00:11:59,960 --> 00:12:03,320 should take it seriously. But given that for a long time 243 00:12:03,320 --> 00:12:05,920 the Home Office and then the Ministry of Justice wanted to 244 00:12:05,920 --> 00:12:09,080 use electronic monitoring in more punitive ways, I could see 245 00:12:09,080 --> 00:12:13,560 that the I could see why the probation service resisted it 246 00:12:13,560 --> 00:12:16,960 when there was such a punitive rhetoric around electronic 247 00:12:16,960 --> 00:12:21,960 monitoring, which didn't prevail to the same extent in a number 248 00:12:21,960 --> 00:12:25,200 of European probation services who I think finished up using 249 00:12:25,200 --> 00:12:27,680 electronic monitoring in much more interesting ways. 250 00:12:27,680 --> 00:12:31,040 And I guess the core thing there is that the European probation 251 00:12:31,040 --> 00:12:34,120 services, take the Scandinavian ones or the Netherlands, they 252 00:12:34,120 --> 00:12:38,400 did not farm out the delivery of electronic monitoring to an 253 00:12:38,400 --> 00:12:42,320 external private contractor. Obviously you work with the 254 00:12:42,320 --> 00:12:45,640 private sector to get the tech from them and to get technical 255 00:12:45,640 --> 00:12:47,760 assistance with them. In Scandinavia and the 256 00:12:47,760 --> 00:12:50,080 Netherlands, electronic monitoring was always in the 257 00:12:50,080 --> 00:12:53,000 hands of the probation service and they had to learn how to use 258 00:12:53,000 --> 00:12:56,280 it wisely and well. In fact, so split was electronic 259 00:12:56,280 --> 00:12:59,040 monitoring probation in England and Wales that there was never 260 00:12:59,040 --> 00:13:01,880 the opportunity to learn how to use it wisely and well. 261 00:13:01,880 --> 00:13:05,000 Because the Ministry of Justice's view until relatively 262 00:13:05,000 --> 00:13:08,280 recently was that these things should be kept set separate. 263 00:13:08,360 --> 00:13:11,760 And I always feared that if there was a separate development 264 00:13:11,760 --> 00:13:14,280 of electronic monitoring, it would be used to advance a 265 00:13:14,280 --> 00:13:18,120 technological agenda which got further and further away from 266 00:13:18,120 --> 00:13:20,280 the practise ideals of the probation service. 267 00:13:20,280 --> 00:13:23,280 That I think is the situation that we're in at the moment. 268 00:13:23,320 --> 00:13:28,200 And that worries me a little bit about AI because having seen 269 00:13:28,200 --> 00:13:31,200 that the Ministry of Justice did not use electronic monitoring 270 00:13:31,440 --> 00:13:35,080 wisely and well, I fear despite all the rhetoric and all the 271 00:13:35,080 --> 00:13:38,400 dedication of people like Louis, the Ministry of Justice might 272 00:13:38,400 --> 00:13:43,160 still finish up using AI badly rather than using it to support 273 00:13:43,160 --> 00:13:46,160 the things that I could see the point of if that's what it was 274 00:13:46,160 --> 00:13:48,160 used for. Really great points there. 275 00:13:48,440 --> 00:13:50,480 I was going to segue straight to Louis. 276 00:13:50,480 --> 00:13:52,080 I thought you might have a response to that. 277 00:13:52,280 --> 00:13:55,760 I think in our context, in the Justice AI unit and gent of AI 278 00:13:55,760 --> 00:13:58,880 tools, if someone has been given access to a tool and they don't 279 00:13:58,880 --> 00:14:01,520 use it, I've literally just been saying to people throughout my 280 00:14:01,520 --> 00:14:03,680 time here, we won't take a punitive measure. 281 00:14:03,720 --> 00:14:05,520 There's no like slapping on the wrist. 282 00:14:05,520 --> 00:14:06,880 Louis, why didn't you use your tool? 283 00:14:06,880 --> 00:14:08,080 You're supposed to use it no matter what. 284 00:14:08,320 --> 00:14:11,040 Instead, our mantra is more like, oh, interesting, we saw 285 00:14:11,080 --> 00:14:12,560 you got access, but you haven't used it. 286 00:14:12,760 --> 00:14:15,240 How can we help you? So I found myself reaching out 287 00:14:15,240 --> 00:14:17,720 to people and sometimes they go say, oh, sorry, I haven't had 288 00:14:17,720 --> 00:14:20,440 time just to open it. I'm so busy at my, I'm really 289 00:14:20,440 --> 00:14:21,480 tired. My head's full. 290 00:14:21,800 --> 00:14:23,840 We go, no worries. If you want some help, let us 291 00:14:23,840 --> 00:14:26,840 know, reach out. If not, when you're ready, give 292 00:14:26,840 --> 00:14:28,600 it a go. And here's how to get started. 293 00:14:28,720 --> 00:14:31,240 We've also found things like if there is a particular reason you 294 00:14:31,240 --> 00:14:33,800 didn't get started, did you want better training? 295 00:14:33,800 --> 00:14:36,200 Did you want in person training? Thank you for giving us 296 00:14:36,200 --> 00:14:38,120 information. For me, it's like a really cool 297 00:14:38,120 --> 00:14:40,720 learning to say, if you're not using a thing, why not? 298 00:14:40,920 --> 00:14:42,040 How do we make it better for you? 299 00:14:42,440 --> 00:14:44,360 And how do we ultimately live and breathe the goal of 300 00:14:44,440 --> 00:14:46,560 improving your working life no matter what it takes. 301 00:14:47,080 --> 00:14:48,840 And I think like I've been blessed to work with a lot of 302 00:14:48,840 --> 00:14:52,200 probation officers because these people are fantastic, very 303 00:14:52,200 --> 00:14:54,000 honest, upfront. If they like something they'll 304 00:14:54,000 --> 00:14:55,720 tell you they like it, if they hate something they'll tell you 305 00:14:55,720 --> 00:14:58,200 they hate it and they're open minded to try something new. 306 00:14:58,200 --> 00:15:01,320 So I've been lucky enough to spend two days a week at the 307 00:15:01,320 --> 00:15:02,800 moment. It's been busy but it's so 308 00:15:02,800 --> 00:15:06,760 useful in probation offices. Being at the front line with our 309 00:15:06,760 --> 00:15:09,980 brilliant staff has helped me work closely with them, my Co 310 00:15:09,980 --> 00:15:12,960 worker, give them quick things to learn from and then build 311 00:15:13,160 --> 00:15:16,200 better for them. The Justice AR unit is really 312 00:15:16,280 --> 00:15:19,360 big on the fact that we need to get as close as possible to 313 00:15:19,360 --> 00:15:22,160 probation staff in order to build them winning products or 314 00:15:22,760 --> 00:15:25,520 help them with adoption. Any sport or products. 315 00:15:25,520 --> 00:15:28,280 We see their working life, we feel their pain. 316 00:15:28,280 --> 00:15:31,080 I think my favourite example was one time someone said to me, if 317 00:15:31,080 --> 00:15:33,520 you want to know how hard it is, do you want to stay with me 318 00:15:33,520 --> 00:15:35,360 until I finish work? I'm on duty today. 319 00:15:35,760 --> 00:15:38,240 I'm finishing at 8:30. I said yes, it was great. 320 00:15:38,240 --> 00:15:40,320 It was so useful. Like they had 10 meetings 321 00:15:40,320 --> 00:15:42,800 back-to-back in a day. Like that's crazy. 322 00:15:42,840 --> 00:15:45,360 And the amount of admin that has done between it was insane, 323 00:15:45,840 --> 00:15:46,680 Yeah. Thank you. 324 00:15:46,680 --> 00:15:49,400 That's that's a really important overview and what I'm really 325 00:15:49,400 --> 00:15:52,000 hearing is that you're understanding the human element 326 00:15:52,000 --> 00:15:54,440 of probation officers and the role that they do I. 327 00:15:54,840 --> 00:15:56,840 Mean this is a common pattern I see this week. 328 00:15:56,880 --> 00:15:59,720 I got a call up later in the afternoon being like, I've got 329 00:15:59,720 --> 00:16:01,680 an idea for you. This is also the probation 330 00:16:01,680 --> 00:16:02,880 officer. And it's like, I've got an idea 331 00:16:02,880 --> 00:16:04,320 for you. Can you, can you, can you build 332 00:16:04,320 --> 00:16:05,480 this? And what was great is they 333 00:16:05,480 --> 00:16:07,640 showed me this thing and head on their screen. 334 00:16:07,640 --> 00:16:10,800 They're like, this bit takes me ages, but it's really benign. 335 00:16:10,800 --> 00:16:13,000 Like it's really basic. Can you help me with that? 336 00:16:13,240 --> 00:16:15,520 And they always say, I've noticed, but don't do this bit, 337 00:16:15,520 --> 00:16:17,040 don't bother with that, but I'll do the analysis. 338 00:16:17,040 --> 00:16:19,840 So I'm already seeing when people do work with me, they'll 339 00:16:19,840 --> 00:16:22,800 say, oh, here's my idea. But AI, can you help me with 340 00:16:22,800 --> 00:16:25,280 this bit, which I'll describe as like the cognitive routine 341 00:16:25,280 --> 00:16:26,520 stuff. They always tell us this is the 342 00:16:26,520 --> 00:16:27,520 bit that's most important to them. 343 00:16:27,680 --> 00:16:31,080 You know, it's the identifying the things that really inform 344 00:16:31,080 --> 00:16:34,240 your awareness of the protective factors, the risk factor, the 345 00:16:34,240 --> 00:16:36,400 analysis piece itself. They're always super keen on 346 00:16:36,400 --> 00:16:37,520 life. Don't touch that bit. 347 00:16:37,520 --> 00:16:43,160 My concern about introducing AI into probation in this way is 348 00:16:43,160 --> 00:16:46,480 that we're not talking about an ideal or even a healthy 349 00:16:46,480 --> 00:16:49,920 probation service right the moment the Ministry of Justices 350 00:16:49,920 --> 00:16:53,920 plans to vastly expand the use of electronic monitoring and to 351 00:16:53,920 --> 00:16:56,880 introduce AI into all public services. 352 00:16:56,880 --> 00:16:59,560 Because this debate is not just restricted probation. 353 00:16:59,680 --> 00:17:02,640 The probation service is not in good shape and hasn't been for a 354 00:17:02,640 --> 00:17:06,000 long time. It was severely damaged by the 355 00:17:06,000 --> 00:17:10,839 attempt to part privatise it, to privatise 70% of it 10 years 356 00:17:10,839 --> 00:17:12,680 ago. We lost an awful lot of staff 357 00:17:12,680 --> 00:17:15,599 from the probation service who have never been replaced and 358 00:17:15,599 --> 00:17:18,960 we've never managed to recruit sufficient probation officers or 359 00:17:18,960 --> 00:17:22,520 retain sufficient probation officers to do what is required 360 00:17:22,599 --> 00:17:24,800 Now. My reading of that situation is 361 00:17:24,800 --> 00:17:29,160 that a public service that is short of 10,000 staff is rife 362 00:17:29,160 --> 00:17:33,480 for greater or lesser degrees of automation, right? 363 00:17:34,000 --> 00:17:38,320 And my worry, I suppose, is how far that goes under those 364 00:17:38,320 --> 00:17:41,760 circumstances of staff shortage. Because, you know, the Ministry 365 00:17:41,760 --> 00:17:45,080 of Justice can say all it likes that we really believe in a 366 00:17:45,080 --> 00:17:48,440 rehabilitative probation service where people have relational 367 00:17:48,440 --> 00:17:50,840 skills to engage with difficult offenders. 368 00:17:50,880 --> 00:17:53,960 But if you can't find the people, there will be a tipping 369 00:17:53,960 --> 00:17:57,680 point where the technology becomes more important than the 370 00:17:57,680 --> 00:17:59,480 people. And I appreciate that Louis 371 00:17:59,480 --> 00:18:03,080 isn't defending that scenario, but I think that scenario is 372 00:18:03,280 --> 00:18:05,480 possible. And that worries me about the 373 00:18:05,480 --> 00:18:08,280 direction of travel. Engaging with people to find out 374 00:18:08,320 --> 00:18:11,200 what they want under the circumstances that we're in is a 375 00:18:11,200 --> 00:18:14,360 very sensible thing to be doing. But I learned from electronic 376 00:18:14,360 --> 00:18:18,040 monitoring, from all the people who had good ideas about its 377 00:18:18,040 --> 00:18:21,240 potential, including myself, that the road to hell really is 378 00:18:21,240 --> 00:18:24,400 paved with good intentions. And despite your best intentions 379 00:18:24,400 --> 00:18:27,840 as to how a technology might be used, the powers that be can 380 00:18:27,840 --> 00:18:31,520 still shape it in ways that you do not find congenial. 381 00:18:31,560 --> 00:18:33,800 That's my worry. And that's a really important 382 00:18:33,800 --> 00:18:36,480 concern to I'm interested in Louis. 383 00:18:36,520 --> 00:18:38,960 Thoughts on that? Can I just anchor that in 384 00:18:38,960 --> 00:18:42,120 practise a little bit about the about the recording and 385 00:18:42,120 --> 00:18:45,760 transcription technologies that have been used in some probation 386 00:18:45,760 --> 00:18:48,440 services and will be rolled out further? 387 00:18:48,440 --> 00:18:51,600 I was talking to a senior probation officer friend of mine 388 00:18:51,600 --> 00:18:54,600 of long experience who said to me that even if these 389 00:18:54,600 --> 00:18:58,880 technologies do save individual probation services between 30 390 00:18:58,880 --> 00:19:03,600 minutes and 60 minutes a day, and we can all see the value of 391 00:19:03,600 --> 00:19:05,720 that. But then he said I still don't 392 00:19:05,720 --> 00:19:08,160 have enough probation officers to do all the work that needs 393 00:19:08,160 --> 00:19:09,960 doing. And that doesn't vitiate the 394 00:19:09,960 --> 00:19:12,440 technology. It doesn't say don't bother to 395 00:19:12,440 --> 00:19:17,200 do this, but it does mean the technology isn't solving the 396 00:19:17,200 --> 00:19:20,400 problem in a significant way because you just haven't got the 397 00:19:20,400 --> 00:19:22,560 people to do the relational work. 398 00:19:22,600 --> 00:19:26,480 And so his fear about what that the recording and transcription 399 00:19:26,480 --> 00:19:31,360 technology would do would just enable his team to do more 400 00:19:31,400 --> 00:19:34,320 interviews and more transcription, but never get 401 00:19:34,320 --> 00:19:37,880 round to doing the relational, relational work that that 402 00:19:37,880 --> 00:19:40,280 technology is supposed to free them up to do. 403 00:19:40,600 --> 00:19:44,000 So to what Mike said, I think if you think a bigger picture to 404 00:19:44,160 --> 00:19:47,600 ensure that we have one of the, you know, the world class 405 00:19:47,600 --> 00:19:51,280 probation service that's truly focused on, you know, helping 406 00:19:51,280 --> 00:19:53,760 with safer community technology is one part. 407 00:19:54,480 --> 00:19:56,960 But we still need to have a lot of work to ensure that we have 408 00:19:56,960 --> 00:19:59,760 the best people well supported to do a good job. 409 00:19:59,960 --> 00:20:02,320 I think my favourite thing, which I love from our leadership 410 00:20:02,360 --> 00:20:05,360 is the announcement made at the start of this year, which is 411 00:20:05,440 --> 00:20:08,720 that we're going to continue to use AI and other technologies to 412 00:20:08,720 --> 00:20:11,240 explore opportunities. But simultaneously, we're still 413 00:20:11,240 --> 00:20:14,320 committed to recruit 1600 more probation officers. 414 00:20:14,400 --> 00:20:16,560 We know it's not easy. But we know it's also super 415 00:20:16,560 --> 00:20:18,280 important. And I think, Mike, to your point 416 00:20:18,280 --> 00:20:20,840 about a senior probation officer saying, oh, am I just having 417 00:20:20,840 --> 00:20:23,080 more technology solutions to do more work? 418 00:20:23,360 --> 00:20:24,800 I think it's a fair thing to flag. 419 00:20:24,800 --> 00:20:26,960 And that's why in our world we're saying, well, we don't 420 00:20:26,960 --> 00:20:29,440 simply say time savings, time savings, time savings. 421 00:20:29,640 --> 00:20:32,680 That isn't simply the goal. When it talks to if I talk about 422 00:20:32,680 --> 00:20:35,400 transcription summarisation, the things that I noticed landing 423 00:20:35,400 --> 00:20:37,960 better with leadership at the moment is actually like the 424 00:20:37,960 --> 00:20:40,640 conversation and engagement. The people involved in the pilot 425 00:20:40,640 --> 00:20:43,480 so far, they still might take the occasional notes that helps 426 00:20:43,480 --> 00:20:45,320 remember something that's important step, but their 427 00:20:45,320 --> 00:20:47,800 conversations feel more human, more natural. 428 00:20:47,880 --> 00:20:50,080 People probation have been telling us things like, yeah, it 429 00:20:50,440 --> 00:20:52,880 feels like a nicer conversation. You're not looking down all the 430 00:20:52,880 --> 00:20:54,560 time, you're not writing all the time. 431 00:20:54,640 --> 00:20:58,360 So that pure golden outcome of having a better relationship is 432 00:20:58,360 --> 00:21:00,960 super important to me. Also things like, and I found 433 00:21:00,960 --> 00:21:02,640 all the information about you and I don't have to ask you the 434 00:21:02,640 --> 00:21:05,040 same thing again and again, those really important parts in 435 00:21:05,040 --> 00:21:07,320 the crux of the relationship. And that's not something that 436 00:21:07,320 --> 00:21:09,240 would be captured by Tom at time saving. 437 00:21:09,440 --> 00:21:11,560 In that context, that doesn't matter to the personal patient. 438 00:21:11,800 --> 00:21:13,720 It's talking about the fact that they feel like they're being 439 00:21:13,720 --> 00:21:17,120 well supported, that AI can only support to help people get the 440 00:21:17,120 --> 00:21:19,240 best out of working life. You'll be fools to say good 441 00:21:19,240 --> 00:21:21,120 news. You're using AI to simply still 442 00:21:21,120 --> 00:21:23,400 feel busy, but now use AI to feel busy. 443 00:21:23,480 --> 00:21:26,240 We want people to wake up, go to their working life and finish at 444 00:21:26,240 --> 00:21:29,840 5:00 to say, Gee, I've got time to, you know, decompress and 445 00:21:29,840 --> 00:21:31,600 relax. We want people to have time to 446 00:21:31,600 --> 00:21:33,560 take lunch breaks. We want people to have time to 447 00:21:33,560 --> 00:21:36,080 have headspace to go. I've just had a meeting and I've 448 00:21:36,080 --> 00:21:38,120 just remembered, let me speak to social works. 449 00:21:38,120 --> 00:21:40,120 Let me speak to police. I think what we observed today 450 00:21:40,120 --> 00:21:42,800 is people are cognitively so busy, sometimes they don't have 451 00:21:42,800 --> 00:21:45,160 headspace to think of the best decision for the personal 452 00:21:45,160 --> 00:21:46,520 probation for the safer communities. 453 00:21:46,720 --> 00:21:49,040 It's super hard. But Mike, I think your point, we 454 00:21:49,040 --> 00:21:51,360 internally as well as certainly need to find that Trump 455 00:21:51,440 --> 00:21:53,720 retreatment, getting the best people in and sporting our best 456 00:21:53,720 --> 00:21:56,200 people as well as helping them with the best tools. 457 00:21:56,480 --> 00:21:59,920 Tell them to do awesome work. Really important on my radar. 458 00:21:59,920 --> 00:22:01,440 Yeah, that's. Pretty important to recognise 459 00:22:01,440 --> 00:22:03,800 and you know the probation officer training programme does 460 00:22:03,800 --> 00:22:05,960 take time and it's important that it does take time. 461 00:22:05,960 --> 00:22:09,960 So in relation to what's just being talked about there that a 462 00:22:09,960 --> 00:22:13,360 key statement that jumped out to me in the AIA action plan for 463 00:22:13,360 --> 00:22:16,800 justice is a statement around AI should enable human talent, 464 00:22:16,800 --> 00:22:18,560 skill and judgement, not replace it. 465 00:22:18,880 --> 00:22:23,520 So that recognition that we do need a fully staffed probation 466 00:22:23,520 --> 00:22:28,560 service with all those skills and AI is there to enable it and 467 00:22:28,560 --> 00:22:31,160 not to replace it. That's all true, but I suppose 468 00:22:31,160 --> 00:22:34,840 my question is still, what happens if we don't achieve it? 469 00:22:34,920 --> 00:22:37,200 You know? What happens if probation just 470 00:22:37,200 --> 00:22:42,120 ceases to be an attractive profession for enough young 471 00:22:42,120 --> 00:22:46,320 people, or the Home Office or the Ministry of Justice decides 472 00:22:46,680 --> 00:22:50,360 with limited resources they would rather invest in tech than 473 00:22:50,360 --> 00:22:52,680 in people? I don't think either of those 474 00:22:52,680 --> 00:22:56,480 scenarios are entirely out of the picture, although I 475 00:22:56,480 --> 00:22:59,640 understand. I do get that the the rhetoric 476 00:22:59,800 --> 00:23:04,680 about how we should use AI wisely and well just presume 477 00:23:04,680 --> 00:23:06,480 that there are going to be enough people. 478 00:23:06,760 --> 00:23:09,480 But if there aren't, what happens then? 479 00:23:09,480 --> 00:23:14,680 The Council of Europe ethical recommendation on AI and 480 00:23:14,960 --> 00:23:20,040 prisons, which came out in 2024. The work began on it in 2021. 481 00:23:20,360 --> 00:23:25,520 But the point I want to make about it is that even though I 482 00:23:25,600 --> 00:23:30,000 only worked on it for a year and left because I felt that the 483 00:23:30,000 --> 00:23:34,200 whole Council of Europe ethos was a desperation to insist that 484 00:23:34,200 --> 00:23:37,320 there have to be benefits, benefits, benefits across public 485 00:23:37,320 --> 00:23:39,000 service and across criminal justice. 486 00:23:39,040 --> 00:23:42,320 And I didn't like being constrained by that kind of 487 00:23:42,320 --> 00:23:45,600 attitude when I can see what the benefits of AI are going to be 488 00:23:45,600 --> 00:23:49,640 in scientific innovation and military security and maybe 489 00:23:49,640 --> 00:23:53,000 economic growth. But I find the case about AI is 490 00:23:53,000 --> 00:23:56,000 going to transform public service for the better than 491 00:23:56,000 --> 00:23:58,320 that. We may never have enough of the 492 00:23:58,320 --> 00:24:00,800 human element in the public services of the future. 493 00:24:00,800 --> 00:24:03,120 Rather worrying. And what that Council of Europe 494 00:24:03,120 --> 00:24:07,560 report said in an appendix was that you could with emerging and 495 00:24:07,560 --> 00:24:12,400 existing technologies automates a great deal of what probation 496 00:24:12,400 --> 00:24:15,480 officers already do right. That would have included risk 497 00:24:15,480 --> 00:24:18,440 assessments, online rehabilitation courses, 498 00:24:18,440 --> 00:24:21,520 monitoring their location and behaviour in the community, 499 00:24:21,560 --> 00:24:25,000 giving them self help tools to use after release, even using 500 00:24:25,000 --> 00:24:28,480 chat bots for counselling and practise and and finally having 501 00:24:28,480 --> 00:24:31,760 a a technological way of doing the dialogical bits of 502 00:24:31,760 --> 00:24:33,960 probation. Now the Council of Europe 503 00:24:34,000 --> 00:24:38,240 emphatically said don't do that, but it was interesting all the 504 00:24:38,240 --> 00:24:42,840 same that their expert said it could be done and the ethical 505 00:24:42,840 --> 00:24:46,360 recommendation is not enforceable if countries want. 506 00:24:46,360 --> 00:24:49,280 If member countries want to do it differently and push 507 00:24:49,280 --> 00:24:53,400 automation further in public services in general and 508 00:24:53,400 --> 00:24:56,840 probation in particular, there are very few constraints on 509 00:24:56,840 --> 00:25:00,840 them, legal or ethical, to prevent them doing that right? 510 00:25:00,880 --> 00:25:04,880 So without wanting to dispute the technologies that are in use 511 00:25:04,880 --> 00:25:07,840 now and the good intentions behind it all and the 512 00:25:07,840 --> 00:25:11,720 assumptions that we will have a fully human workforce, I just 513 00:25:11,720 --> 00:25:15,280 want to step back and say what if we don't get that good 514 00:25:15,280 --> 00:25:19,000 surrounding human environments? What kind of decisions in a 515 00:25:19,000 --> 00:25:23,240 world of limited resources will government agencies take about 516 00:25:23,240 --> 00:25:26,040 the best way to manage offenders in the community? 517 00:25:26,120 --> 00:25:29,080 And it isn't a foregone conclusion to me that they will 518 00:25:29,080 --> 00:25:31,760 support the humanly grounded model. 519 00:25:31,840 --> 00:25:35,640 They might just want to automate a lot more of probation that 520 00:25:35,640 --> 00:25:37,880 anybody is actually talking about doing at the moment. 521 00:25:38,040 --> 00:25:38,600 Yeah. Thank you. 522 00:25:38,600 --> 00:25:40,760 Mike, I think you raised some valid concerns. 523 00:25:40,840 --> 00:25:42,040 Louis, did you have any thoughts? 524 00:25:42,400 --> 00:25:45,240 I think my take on that is ultimately what we're starting 525 00:25:45,240 --> 00:25:48,720 to see, not simply probation but other sectors, is that we will 526 00:25:48,720 --> 00:25:52,000 need to start leaning into what we define as the best ideal 527 00:25:52,000 --> 00:25:54,040 future of work. What does work look like when 528 00:25:54,040 --> 00:25:57,440 you have tools that can help you do things which are informing, 529 00:25:57,440 --> 00:25:59,240 enhancing, supporting your practise. 530 00:25:59,360 --> 00:26:02,760 But it's a really hard sweet spot to define and I don't think 531 00:26:02,760 --> 00:26:05,280 anyone in the world has defined this, let alone for probation, 532 00:26:05,280 --> 00:26:07,680 but for other contexts. What happens in years in the 533 00:26:07,680 --> 00:26:10,800 future to say we start having tools that change the way we go 534 00:26:10,800 --> 00:26:13,480 about our day-to-day life? How does that change the work? 535 00:26:14,520 --> 00:26:17,000 Obviously we need to make sure we have our best people 536 00:26:17,000 --> 00:26:19,360 supported and recruited. But if we look at the crux of 537 00:26:19,360 --> 00:26:21,960 the relationships, the risk factors, the protective factors, 538 00:26:22,040 --> 00:26:24,080 I think these are the things that we still hold true, the 539 00:26:24,080 --> 00:26:26,120 things that we won't have technology touch. 540 00:26:26,200 --> 00:26:28,720 It would be more looking into what that looks like. 541 00:26:29,080 --> 00:26:30,760 So it's part of the work that we're doing now. 542 00:26:30,800 --> 00:26:34,000 We're giving people helpful tools and we're also making sure 543 00:26:34,000 --> 00:26:37,040 that we work with evaluation leads to see, well, what do we 544 00:26:37,040 --> 00:26:39,240 observe on the field? What do people do differently? 545 00:26:39,320 --> 00:26:41,480 But what do we observe in that practise? 546 00:26:41,480 --> 00:26:43,720 What are the behaviours that we want to promote and what are the 547 00:26:43,720 --> 00:26:46,320 behaviours actually we need to support further. 548 00:26:46,320 --> 00:26:48,840 So one thing that we're doing in our unit, which keeps us very 549 00:26:48,840 --> 00:26:52,400 busy but super important is not simply provide or build 550 00:26:52,600 --> 00:26:55,000 solutions. It's also to evaluate what what 551 00:26:55,000 --> 00:26:57,960 actually happens. What do people do behaviourally 552 00:26:57,960 --> 00:26:59,880 as well as what do they think think about it, what do they say 553 00:26:59,880 --> 00:27:02,000 about it? So I think in the bigger picture 554 00:27:02,000 --> 00:27:05,000 there will be a very different future of work for loads of 555 00:27:05,000 --> 00:27:07,320 context, including probation, which is why it's important that 556 00:27:07,320 --> 00:27:09,200 people like me don't lose track of that. 557 00:27:09,240 --> 00:27:11,560 So those are the things that we're leaning into and it helps 558 00:27:11,560 --> 00:27:15,000 us improve the way that we use technology in the way that it 559 00:27:15,000 --> 00:27:16,440 supports. It doesn't replace people. 560 00:27:16,520 --> 00:27:18,360 Thank you, Louis. Yeah, it's really interesting 561 00:27:18,360 --> 00:27:21,120 about the evaluation and what you observe people doing in 562 00:27:21,120 --> 00:27:23,000 practise, how it might change practise. 563 00:27:23,440 --> 00:27:26,240 And if people are interested in looking at the work and how we 564 00:27:26,240 --> 00:27:28,800 consider evaluation, one thing I'd recommend people looking 565 00:27:28,800 --> 00:27:32,600 into is our safety framework. So it's a framework that was 566 00:27:32,600 --> 00:27:36,400 built in collaboration with our brilliant teams internally and 567 00:27:36,400 --> 00:27:39,360 the Allen Chewing Institute. Essentially it says in the midst 568 00:27:39,400 --> 00:27:42,160 of thinking about an idea, building an idea, or having an 569 00:27:42,160 --> 00:27:45,080 idea being used by real people in a real world environment, we 570 00:27:45,080 --> 00:27:46,760 talk about ethics in divisive ways. 571 00:27:46,760 --> 00:27:51,440 And that looks about from the societal sustainability angles 572 00:27:51,440 --> 00:27:53,600 as well as like behaviour. So one of the things that we are 573 00:27:53,680 --> 00:27:58,000 lucky to benefit from, we've got some data ethics leads who help 574 00:27:58,000 --> 00:28:00,160 us be well informed from an evaluation perspective. 575 00:28:00,160 --> 00:28:01,920 So we'd recommend people have looked if they're interested. 576 00:28:02,040 --> 00:28:02,960 Yeah. Thank you, David. 577 00:28:02,960 --> 00:28:04,240 Yeah. And that's an interesting point 578 00:28:04,240 --> 00:28:06,840 about ethics, because that's your point, Mike, isn't it about 579 00:28:06,840 --> 00:28:10,160 how can we make sure it is still, you know, a service that 580 00:28:10,520 --> 00:28:14,200 isn't all automated, that has that at its heart about people? 581 00:28:14,240 --> 00:28:17,800 So linked to that probation work is rooted in relationships and 582 00:28:17,800 --> 00:28:20,960 professional judgement. So how do you think AI and 583 00:28:20,960 --> 00:28:23,880 technologies like electronic monitoring might support or 584 00:28:23,880 --> 00:28:25,600 challenge this core part of the world? 585 00:28:25,920 --> 00:28:30,240 Well, I've already indicated what some of my fears are about 586 00:28:30,240 --> 00:28:33,680 that, because in a world of limited resources, I'm not sure 587 00:28:33,680 --> 00:28:38,080 what constraints there would be on using technology more than 588 00:28:38,080 --> 00:28:41,160 might be warranted. In the early days of electronic 589 00:28:41,160 --> 00:28:43,960 monitoring, one of the Home Office researchers said about it 590 00:28:44,120 --> 00:28:46,680 that the Home Office wants to use electronic monitoring in 591 00:28:46,680 --> 00:28:49,520 ways that go beyond the evidence of its usefulness. 592 00:28:49,520 --> 00:28:53,280 And there was evidence of its usefulness, but the Home Office 593 00:28:53,280 --> 00:28:55,520 wanted to to take it further than that. 594 00:28:55,520 --> 00:28:58,200 And perhaps that's in in the nature of technological 595 00:28:58,200 --> 00:29:01,880 innovation, one has to invest in uncertain outcomes. 596 00:29:02,080 --> 00:29:04,840 Now it was interesting to me though, and this was one of the 597 00:29:04,840 --> 00:29:08,360 bases on which I was supportive of electronic monitoring, is 598 00:29:08,360 --> 00:29:10,840 that there was some research done in Canada at the turn of 599 00:29:10,840 --> 00:29:13,840 the century that showed how electronic monitoring could be 600 00:29:13,840 --> 00:29:16,000 used to support rehabilitative programmes. 601 00:29:16,040 --> 00:29:19,920 Now that research was very influential in mainland European 602 00:29:19,920 --> 00:29:23,960 probation services. It was of 0 influence in the 603 00:29:23,960 --> 00:29:26,800 England and Wales Probation Service because we didn't 604 00:29:26,800 --> 00:29:29,600 combine the electronic monitoring and probation in the 605 00:29:29,600 --> 00:29:32,920 single organisation. We we created degrees of 606 00:29:32,920 --> 00:29:36,000 tensions between the private company and the probation 607 00:29:36,000 --> 00:29:39,640 service and quite often the government was giving the 608 00:29:39,640 --> 00:29:42,560 impression that it favoured the approach of the private company 609 00:29:42,560 --> 00:29:45,760 more than it favoured the social work approach of the probation 610 00:29:45,760 --> 00:29:48,240 service. So I've yet to be convinced that 611 00:29:48,240 --> 00:29:52,080 a similar dynamic won't kick in with artificial intelligence. 612 00:29:52,080 --> 00:29:55,080 And I am alarmed by the fact that at the moment the Ministry 613 00:29:55,080 --> 00:29:58,640 of Justice is openly saying that it wants to use electronic 614 00:29:58,640 --> 00:30:00,480 monitoring in more punitive ways. 615 00:30:00,760 --> 00:30:04,600 Right the the rhetoric is now prison outside prison or 616 00:30:04,600 --> 00:30:08,840 incarceration without walls, all enabled by electronic 617 00:30:08,840 --> 00:30:11,800 monitoring. Now to a certain probation cast 618 00:30:11,800 --> 00:30:15,360 of mind that is anathema. And it is once again taking 619 00:30:15,360 --> 00:30:18,560 electronic monitoring in an ultra punitive direction, which 620 00:30:18,560 --> 00:30:21,960 is precisely what I didn't want to see and what I at one time 621 00:30:21,960 --> 00:30:26,160 hoped the value base of the probation service would resist. 622 00:30:26,200 --> 00:30:30,160 And in in Europe I see some shaping of electronic monitoring 623 00:30:30,160 --> 00:30:33,120 by probation values, but I haven't seen it here. 624 00:30:33,240 --> 00:30:35,640 It's really interesting the point though, that you make 625 00:30:35,640 --> 00:30:39,640 about the fact that electronic monitoring is delivered by an 626 00:30:39,640 --> 00:30:42,960 external organisation. And I think something to bear in 627 00:30:42,960 --> 00:30:46,000 mind when we're looking at AI and how that's embedded. 628 00:30:46,000 --> 00:30:48,920 And I think that point is a really important one about how 629 00:30:48,920 --> 00:30:52,200 the relationship is managed between who's delivering it and 630 00:30:52,200 --> 00:30:54,960 who's in charge of it. And it's probably worth saying 631 00:30:54,960 --> 00:30:58,120 at this point that it is now the policy of the Probation 632 00:30:58,120 --> 00:31:01,200 Institute to incorporate electronic monitoring within the 633 00:31:01,200 --> 00:31:04,600 probation service. Now that would transform our 634 00:31:04,600 --> 00:31:08,160 probation service and make it more like the European probation 635 00:31:08,160 --> 00:31:11,160 services in ways that we've never done before. 636 00:31:11,160 --> 00:31:14,560 And if we're going to have a more technological service in 637 00:31:14,560 --> 00:31:16,680 general, that makes a great deal of sense. 638 00:31:16,680 --> 00:31:19,800 I mean, the time was when probation didn't want electronic 639 00:31:19,800 --> 00:31:23,720 monitoring within it's boundaries on the principle that 640 00:31:23,720 --> 00:31:26,960 if, if electronic monitoring fails, we can blame the private 641 00:31:26,960 --> 00:31:29,080 sector. But electronic monitoring isn't 642 00:31:29,200 --> 00:31:31,600 going to fail if you use it wisely and well. 643 00:31:31,920 --> 00:31:34,840 And there are no good administrative reasons as to why 644 00:31:34,840 --> 00:31:36,720 in England and Wales we shouldn't do what the 645 00:31:36,720 --> 00:31:38,600 Scandinavian countries have done. 646 00:31:38,920 --> 00:31:42,680 I mean, the Swedish Probation Service actually employs an AI 647 00:31:42,680 --> 00:31:45,920 manager who is now in charge of all technology in the Swedish 648 00:31:45,920 --> 00:31:47,920 probation service. He is like you, Louis. 649 00:31:47,920 --> 00:31:51,760 He's an AI engineer and he gets probation, he gets the Swedish 650 00:31:51,760 --> 00:31:54,160 probation service. And that has worked out for 651 00:31:54,160 --> 00:31:56,120 them. But it's worked out for them in 652 00:31:56,120 --> 00:31:59,760 my view, because they are not as relentlessly punitive in their 653 00:31:59,760 --> 00:32:03,360 approach to offenders in the in the Nordic countries as we are. 654 00:32:03,560 --> 00:32:07,000 And my fear that in the political and penal culture that 655 00:32:07,000 --> 00:32:11,760 we have, that technology will be used badly, despite the best 656 00:32:11,760 --> 00:32:14,960 intentions of the people who are developing it, is very real. 657 00:32:15,400 --> 00:32:18,440 And Louis, if you've got any thoughts about the roll out of 658 00:32:18,600 --> 00:32:23,240 AI and how it can be managed to support that rebuild rather than 659 00:32:23,240 --> 00:32:25,960 risk focused approach? In provision, yeah, for sure. 660 00:32:26,360 --> 00:32:29,280 Some of our staff have joined because they're excited to make 661 00:32:29,280 --> 00:32:31,200 a difference. They might have dyslexia, might 662 00:32:31,200 --> 00:32:33,200 have English as an additional language, they might have 663 00:32:33,200 --> 00:32:36,480 dyspraxia, they might find it difficult to get through the 664 00:32:36,480 --> 00:32:38,720 amount of paperwork on a day-to-day basis. 665 00:32:38,720 --> 00:32:40,760 So one of the benefits we're observing right now, which is 666 00:32:40,760 --> 00:32:43,640 coming through the training as well as the actual feedback, is 667 00:32:43,640 --> 00:32:46,160 that it helps people with a lower barrier of entry. 668 00:32:46,240 --> 00:32:48,720 I've had some staff come to me personally and just say thank 669 00:32:48,720 --> 00:32:50,560 you because PPC, I was close to quitting. 670 00:32:50,560 --> 00:32:52,720 I felt really struggling with the amount of paperwork. 671 00:32:52,880 --> 00:32:54,600 They love the nature of the work. 672 00:32:54,600 --> 00:32:57,760 It is hard, it is complex, and it was struggling with just 673 00:32:57,960 --> 00:33:00,600 keeping track of everything. And they're helping themselves 674 00:33:00,600 --> 00:33:03,520 by using dentive AI in a way that helps them understand 675 00:33:03,520 --> 00:33:05,840 information, easily identify actions. 676 00:33:05,840 --> 00:33:08,760 Again, only as an assistant. The other things that I've seen 677 00:33:08,760 --> 00:33:11,840 so far, so helping people focus on the more human parts of work. 678 00:33:11,920 --> 00:33:15,000 I lose to it earlier, but transcription summarisation work 679 00:33:15,000 --> 00:33:17,280 is my favourite example. We've been seeing people just 680 00:33:17,280 --> 00:33:20,560 say, look, it's not simply about saving me time, it's actually 681 00:33:20,560 --> 00:33:23,360 about me feeling more engaged. Like wow, you observe things, 682 00:33:23,360 --> 00:33:25,440 this is something subtle. I didn't appreciate it really. 683 00:33:25,440 --> 00:33:28,240 But we've had people say, I didn't realise it before, but 684 00:33:28,320 --> 00:33:30,560 when this person would speak about a certain topic, hands 685 00:33:30,560 --> 00:33:34,480 would trench the body language changes and previously I had to 686 00:33:34,480 --> 00:33:36,720 look down and now I'm seeing these subtle changes. 687 00:33:37,080 --> 00:33:39,040 But again, that's what technology, yes, it helps you 688 00:33:39,040 --> 00:33:41,120 focus on the human parts, but it's only because it's 689 00:33:41,120 --> 00:33:44,920 deliberately been built in a way that gives the people meaningful 690 00:33:44,920 --> 00:33:46,880 control. So focusing more human elements 691 00:33:47,320 --> 00:33:49,360 and finding information, that's really important sooner not 692 00:33:49,360 --> 00:33:51,560 later. So finding information and 693 00:33:51,560 --> 00:33:54,920 understanding it as well. For example, I saw recently was 694 00:33:55,280 --> 00:33:57,960 a case where someone was saying, oh, I've just had like 695 00:33:57,960 --> 00:34:00,720 information about drug supports and I can't remember the top of 696 00:34:00,720 --> 00:34:04,240 my head. What is the nanograms to warrant 697 00:34:04,240 --> 00:34:06,560 it being that they've actually consumed it or been in the 698 00:34:06,560 --> 00:34:09,239 facility of it. Oh, let me find what is the 699 00:34:09,239 --> 00:34:11,520 actual guidelines. What is the policy currently? 700 00:34:11,520 --> 00:34:14,679 I found the policy quickly. I found information and now I 701 00:34:14,679 --> 00:34:16,480 can remind myself against the actual results. 702 00:34:16,639 --> 00:34:18,040 Oh, a few. That means that they might have 703 00:34:18,040 --> 00:34:19,480 been in proximity but not taking it. 704 00:34:19,480 --> 00:34:21,360 That's good. It's helped me find information, 705 00:34:21,360 --> 00:34:24,040 find policies, understand them as well as find information 706 00:34:24,040 --> 00:34:26,960 about the person themselves. And it's like AI as an 707 00:34:26,960 --> 00:34:30,040 opportunity to also just be more supportive for new staff. 708 00:34:30,040 --> 00:34:33,679 But we're piloting tools like Dot Lee from the Taught by Human 709 00:34:33,679 --> 00:34:35,639 company. We're helping them with AI 710 00:34:35,639 --> 00:34:39,480 assistance to help learn about topic, collect on a topic, but 711 00:34:39,480 --> 00:34:42,320 simply regurgitate an answer or you're always correct. 712 00:34:42,480 --> 00:34:45,080 You're fantastic. Some of that parent like Dynamic 713 00:34:45,080 --> 00:34:49,120 been excited to use AI to help staff have better reflective 714 00:34:49,159 --> 00:34:51,239 processes to go about learning a thing. 715 00:34:51,480 --> 00:34:53,639 But there are risks. So I think it's worth lagging. 716 00:34:53,880 --> 00:34:55,760 Ultimately, this is a new technology. 717 00:34:55,880 --> 00:34:58,000 This is something that people have never seen before and 718 00:34:58,080 --> 00:35:00,160 people are having to be understanding like that this is 719 00:35:00,160 --> 00:35:03,640 a nondeterministic technology. It's about helping me, but it 720 00:35:03,640 --> 00:35:05,920 can't just give me a binary right or wrong thing. 721 00:35:05,920 --> 00:35:08,480 Sometimes you got to reflect on the thing, critique the thing, 722 00:35:08,520 --> 00:35:09,600 own the thing. The fact that it's 723 00:35:09,600 --> 00:35:11,160 conversational is really alien to people. 724 00:35:11,160 --> 00:35:14,560 I think whenever we do in the past, say I communicate with my 725 00:35:14,560 --> 00:35:16,840 tools, I converse with them to get the best answer. 726 00:35:17,120 --> 00:35:19,000 That never happened. So the example of the drug 727 00:35:19,000 --> 00:35:21,800 report, it wasn't a one way conversation. 728 00:35:22,000 --> 00:35:24,120 That's so different. It's part of that risk. 729 00:35:24,200 --> 00:35:27,920 How do we give our brilliant staff bit of headspace to safely 730 00:35:27,920 --> 00:35:30,360 try something and feel comfortable being empowered? 731 00:35:30,440 --> 00:35:32,600 And the other thing we're interested in is this is a tool 732 00:35:32,600 --> 00:35:34,000 that supports professional judgement. 733 00:35:34,080 --> 00:35:35,960 It can't replace the really important skills. 734 00:35:35,960 --> 00:35:38,480 So we don't want people to become overlined by accident. 735 00:35:38,600 --> 00:35:41,520 We need to encourage and promote those really important human 736 00:35:41,520 --> 00:35:45,480 skills, the motivational interviewing, analysis, etc. 737 00:35:45,600 --> 00:35:48,120 But this is a tool that supports, doesn't take over. 738 00:35:48,120 --> 00:35:50,880 So that's baked into not simply the things we build, but how we 739 00:35:50,880 --> 00:35:52,600 train people and support them with using it. 740 00:35:52,720 --> 00:35:55,440 When we give people new tools, we have to think of it as the 741 00:35:55,440 --> 00:35:57,520 bigger picture. What type of help do they 742 00:35:57,960 --> 00:35:59,920 potentially need? We wanted to be really clear. 743 00:36:00,040 --> 00:36:02,320 This is how you get started, this is how you get help, and 744 00:36:02,320 --> 00:36:04,440 this is how to use it safely in a way that promotes your skills, 745 00:36:04,440 --> 00:36:07,040 not replacing. So just for context, like the 746 00:36:07,040 --> 00:36:09,920 Justice AI unit, we're not simply recruiting people like me 747 00:36:09,960 --> 00:36:12,160 who are building things, engineering things. 748 00:36:12,320 --> 00:36:15,200 We're also recruiting quite a lot of adoption support people, 749 00:36:15,200 --> 00:36:18,440 customer success people who just really care about understanding 750 00:36:18,440 --> 00:36:20,840 the problem space and helping someone get the best out of a 751 00:36:20,840 --> 00:36:22,440 tool. It's really interesting point 752 00:36:22,440 --> 00:36:26,240 about AI could be used for. So if you get given a case from 753 00:36:26,240 --> 00:36:28,320 someone else, for example, and actually could to quickly 754 00:36:28,320 --> 00:36:30,760 summarise some of the key points about that individual. 755 00:36:30,800 --> 00:36:33,720 If you are time short to get up to date with the case before you 756 00:36:33,720 --> 00:36:36,680 meet them, because then you know the individual know that you've 757 00:36:36,680 --> 00:36:38,560 just remembering. When I met some of them, they 758 00:36:38,600 --> 00:36:40,600 said, oh, you've done your homework Kerry, like you know 759 00:36:40,600 --> 00:36:42,320 about me. And I was like, yeah, of course 760 00:36:42,320 --> 00:36:44,640 I'm going to try and find out a bit about you before. 761 00:36:44,640 --> 00:36:47,040 But I suppose it's just some people might have concerns about 762 00:36:47,120 --> 00:36:50,200 the information that it's using because if there's ever any 763 00:36:50,200 --> 00:36:52,320 mistakes in there or anything like that, maybe you wouldn't 764 00:36:52,320 --> 00:36:54,160 know. So I suppose it's always relying 765 00:36:54,160 --> 00:36:56,240 on the data that goes in perhaps. 766 00:36:56,240 --> 00:36:59,920 But I can see how some of those things might help reduce time 767 00:36:59,920 --> 00:37:02,400 but also help when you're time pressured as well. 768 00:37:02,880 --> 00:37:05,000 That could have a good with individuals. 769 00:37:05,280 --> 00:37:06,760 Absolutely. And I think there's a nice way 770 00:37:06,760 --> 00:37:10,720 to frame it going back to the AI action plan, which is this next 771 00:37:10,720 --> 00:37:13,200 year we're really focusing on investing in our foundations 772 00:37:13,200 --> 00:37:15,600 that includes the data. Most of our data isn't as good a 773 00:37:15,600 --> 00:37:18,000 quality as we need it to be to truly leverage emerging 774 00:37:18,000 --> 00:37:19,960 technology. So one part is looking at that 775 00:37:19,960 --> 00:37:23,240 data set and going, actually this might need improving 776 00:37:23,280 --> 00:37:26,120 through XY and Z processes because otherwise we might be 777 00:37:26,120 --> 00:37:29,880 regurgitating things which won't be both useful in the outcome 778 00:37:29,880 --> 00:37:32,200 that we want, but investing in our foundations, which is 779 00:37:32,240 --> 00:37:33,120 something I'm really proud about. 780 00:37:33,120 --> 00:37:35,720 As part of our plan, we don't simply say foundations means 781 00:37:35,720 --> 00:37:38,880 data and quality of data. It also means the foundations of 782 00:37:38,880 --> 00:37:42,040 helping our people feel they are empowered to know how to use it 783 00:37:42,040 --> 00:37:43,920 all, when to use it all, and now to do so safely. 784 00:37:44,080 --> 00:37:45,240 But no, that's a really fairpoint. 785 00:37:45,400 --> 00:37:47,960 Thank you, Kate. And that's something that I 786 00:37:47,960 --> 00:37:49,280 don't know how Mike feels about it. 787 00:37:49,280 --> 00:37:50,800 I'll be interested to hear shortly. 788 00:37:50,800 --> 00:37:54,000 But hopefully is a bit different to the how the electronic 789 00:37:54,000 --> 00:37:57,040 monitoring was rolled out because it's not like there's a 790 00:37:57,160 --> 00:38:00,080 completely different private company that is in charge of the 791 00:38:00,520 --> 00:38:04,280 AI and is saying, look, we're in charge of this, we're running AI 792 00:38:04,280 --> 00:38:07,640 within probation. We have ownership of this. 793 00:38:07,640 --> 00:38:11,640 It sounds like it's certainly being developed, developed and 794 00:38:11,640 --> 00:38:15,000 delivered with the skills and the approaches of the probation 795 00:38:15,000 --> 00:38:19,320 staff in mind and recognises all of that rather than an external 796 00:38:19,320 --> 00:38:22,360 organisation not taking that into consideration. 797 00:38:22,360 --> 00:38:25,680 And then who could potentially be blamed if it goes wrong or, 798 00:38:25,680 --> 00:38:28,680 but I don't know if that's what you're hearing as well, Mike or 799 00:38:28,680 --> 00:38:32,120 if that addresses any of your concerns about the difference or 800 00:38:32,120 --> 00:38:35,080 similarity with electronic? Yes, I mean there are 801 00:38:35,080 --> 00:38:37,880 differences in the electronic monitoring was a much more 802 00:38:37,880 --> 00:38:40,720 limited intervention than the kind of things that have been 803 00:38:40,720 --> 00:38:44,200 envisaged by AI. And potentially it was a tool 804 00:38:44,200 --> 00:38:47,000 that probation officers could easily have got their heads 805 00:38:47,000 --> 00:38:49,720 round as they did in the Netherlands, as they did in in 806 00:38:49,720 --> 00:38:52,480 Scandinavia. And within a certain political 807 00:38:52,480 --> 00:38:55,560 culture they could have used that technology wisely and well. 808 00:38:56,000 --> 00:38:59,840 But I keep coming back to the the issue about it's, it's all 809 00:38:59,840 --> 00:39:05,280 very well to to talk up the micro uses of AI technologies in 810 00:39:05,320 --> 00:39:08,240 the broadest sense in terms of the individual ways in which 811 00:39:08,240 --> 00:39:11,440 they might help people to to process their cases. 812 00:39:11,440 --> 00:39:14,400 And I agree completely with what Louis says about these 813 00:39:14,400 --> 00:39:17,640 technologies are found useful by neurodivergent people. 814 00:39:17,640 --> 00:39:20,560 I get that. I think the the neurodivergent 815 00:39:20,720 --> 00:39:24,400 argument, while true, I think there are limits as to how far 816 00:39:24,400 --> 00:39:27,760 you can use the argument to normalise the development of 817 00:39:27,760 --> 00:39:31,640 this tech in probation. But it is true and the argument 818 00:39:31,720 --> 00:39:34,360 will be used. It's also true, true that all 819 00:39:34,360 --> 00:39:38,440 our lives, and you know, my, the younger members of my family all 820 00:39:38,440 --> 00:39:40,480 embody this. There will be more technology in 821 00:39:40,480 --> 00:39:42,520 everybody's life. We will all have to get used to 822 00:39:42,520 --> 00:39:45,320 more digital infusion into our lives. 823 00:39:45,320 --> 00:39:48,240 We will become an even more hyper connected world. 824 00:39:48,240 --> 00:39:51,040 And in that sense, probation might look odd if it was trying 825 00:39:51,040 --> 00:39:54,920 to stand outside that. But nonetheless, I still want to 826 00:39:54,920 --> 00:39:59,720 register that the things that we go enthusiastically into in the 827 00:39:59,720 --> 00:40:03,520 short term may not bring about the consequences we envision. 828 00:40:03,640 --> 00:40:07,920 In the long term, right? I am reminded of Lord Timpson 829 00:40:07,920 --> 00:40:10,880 when he gave the McWilliams probation lecturer. 830 00:40:10,880 --> 00:40:14,680 He boldly quoted the the line that is attributed to Bill 831 00:40:14,680 --> 00:40:17,200 Gates, although apparently it originated to somebody else. 832 00:40:17,200 --> 00:40:20,920 We overestimate what technology can do in the short term and 833 00:40:20,920 --> 00:40:23,520 underestimate what it will do in the long term. 834 00:40:23,600 --> 00:40:26,920 Now, when Bill Gates says that, it's meant to evoke the idea 835 00:40:26,920 --> 00:40:30,040 that in the long term it's going to be fabulous, right? 836 00:40:30,400 --> 00:40:33,240 But you can turn that round and think in the long term it's 837 00:40:33,240 --> 00:40:36,120 going to be awful. But in the short term we just 838 00:40:36,120 --> 00:40:39,760 can't work that out because we're enthusiastic about the 839 00:40:39,760 --> 00:40:43,280 short term possibilities and we won't see what the long term 840 00:40:43,280 --> 00:40:47,640 disadvantages of it would be in the particular political context 841 00:40:47,640 --> 00:40:50,600 and culture in which we are. We are introducing these 842 00:40:50,600 --> 00:40:54,040 technologies into probation. Really important point and 843 00:40:54,720 --> 00:40:59,120 interesting to reflect back on Lord Simpson's speech there. 844 00:40:59,680 --> 00:41:02,800 There's a a briefing on flag in this world of what I'm working 845 00:41:02,800 --> 00:41:05,680 on with other team members. Like for us, we have a massive 846 00:41:05,720 --> 00:41:08,920 obsession with like solving this thing to make things better. 847 00:41:08,960 --> 00:41:11,160 And that means having our finger on the pulse of what is 848 00:41:11,160 --> 00:41:13,800 happening. So if we observe behaviour 849 00:41:13,800 --> 00:41:17,760 change that isn't what we want change that we want to maybe 850 00:41:17,760 --> 00:41:20,400 support better in other ways, then that's some people like me 851 00:41:20,400 --> 00:41:22,720 to change the way that we build the product. 852 00:41:22,720 --> 00:41:25,600 It's shouldn't be a case where we say the tool of today, 853 00:41:26,640 --> 00:41:29,160 whatever that tool is, it goes out and it's used by lots of 854 00:41:29,160 --> 00:41:30,640 people. But we don't want to play like a 855 00:41:30,640 --> 00:41:33,640 game of darts in the dark. We want to quickly observe and 856 00:41:33,640 --> 00:41:36,040 learn what do people love, what do people hate and what do we 857 00:41:36,040 --> 00:41:39,120 observe as the actual tangible difference in day-to-day working 858 00:41:39,120 --> 00:41:40,640 life. It just means accountability 859 00:41:40,640 --> 00:41:41,840 wise. I don't want to finish my 860 00:41:41,840 --> 00:41:43,880 working day and go sweet, I'll just wait for someone to pick 861 00:41:43,880 --> 00:41:45,640 up. It's like this is on me to make 862 00:41:45,640 --> 00:41:47,200 better, this is on me to improve. 863 00:41:47,200 --> 00:41:50,080 Yes, I'll happily do however much work is needed to make this 864 00:41:50,080 --> 00:41:52,320 come to life in a way that's most useful for our brilliant 865 00:41:52,320 --> 00:41:53,960 probation staff. Yeah, thank you. 866 00:41:53,960 --> 00:41:57,200 So therefore the the concerns that Mike's airing and and I'm 867 00:41:57,200 --> 00:41:59,200 sure you're hearing from them from other people as well, 868 00:41:59,200 --> 00:42:01,920 really important for you to hear and and hold in mind as you're 869 00:42:02,000 --> 00:42:05,280 moving forward and all of this. Thank you for our P quip 870 00:42:05,280 --> 00:42:08,320 listeners who are at the start of their career is in probation, 871 00:42:08,320 --> 00:42:10,720 what do you think they should prioritise in their training and 872 00:42:10,720 --> 00:42:13,800 practise when they're working in a probation service that we do 873 00:42:13,800 --> 00:42:16,520 recognise is increasingly shaped by technology? 874 00:42:16,640 --> 00:42:19,480 Well, that's a that is a that is a fair question. 875 00:42:19,480 --> 00:42:23,200 And if electronic monitoring is going to be more commonplace, 876 00:42:23,200 --> 00:42:27,320 then I think probation officers should try and work more 877 00:42:27,320 --> 00:42:30,400 creatively with it. It seems to me now that the era 878 00:42:30,400 --> 00:42:33,360 where probation is going to resist electronic monitoring is 879 00:42:33,600 --> 00:42:37,560 more or less over, although there will be resistance to an 880 00:42:37,560 --> 00:42:39,840 excess of it. Like use it with everybody all 881 00:42:39,840 --> 00:42:42,040 the time, even if you use it in different ways. 882 00:42:42,040 --> 00:42:45,640 Any attempt to normalise it across the whole of supervising 883 00:42:45,640 --> 00:42:48,560 offenders in the community, whether that's sentenced or or 884 00:42:48,560 --> 00:42:51,640 or released from prison, I think there would be resistance to 885 00:42:51,640 --> 00:42:54,320 that. But the idea that you can use it 886 00:42:54,320 --> 00:42:57,200 constructively in the way you manage offenders in the 887 00:42:57,200 --> 00:43:01,040 community, I think there is an evidence base that helps you to 888 00:43:01,040 --> 00:43:03,160 do that. And I think there are examples 889 00:43:03,160 --> 00:43:05,560 of good practise from other countries which would help 890 00:43:05,560 --> 00:43:07,840 people do that. I mean, I've used this story 891 00:43:07,960 --> 00:43:12,840 before, but back in 2012, a long time ago, I went to the 892 00:43:12,840 --> 00:43:16,920 retirement party of a Dutch probation officer who had had 10 893 00:43:16,920 --> 00:43:21,400 years experience of working with electronic monitoring and 894 00:43:21,400 --> 00:43:24,960 probation in an integrated way. That was 13 years ago. 895 00:43:25,280 --> 00:43:29,400 Now, to this day, there is no probation officer in England who 896 00:43:29,400 --> 00:43:32,640 could even begin to match that kind of level of experience 897 00:43:32,640 --> 00:43:35,520 about how to use electronic monitoring wisely and well. 898 00:43:35,560 --> 00:43:38,920 But such a thing is possible, so I would encourage people to do 899 00:43:38,920 --> 00:43:40,880 that. Our electronic monitoring is 900 00:43:40,880 --> 00:43:43,720 changing. There is more interesting in 901 00:43:43,720 --> 00:43:49,440 using the data that GPS tracking develops to do data analysis 902 00:43:49,440 --> 00:43:53,160 that might lead you to enable a predictive approach to 903 00:43:53,160 --> 00:43:56,400 reoffending. And it's no, it's no you saying 904 00:43:56,400 --> 00:43:58,000 that that's not a good thing to do. 905 00:43:58,320 --> 00:44:01,440 There are techniques for predictivity now that didn't 906 00:44:01,440 --> 00:44:04,920 exist 10 or 15 years ago and that may become an important 907 00:44:04,920 --> 00:44:07,160 part of it. And with higher risk offenders 908 00:44:07,160 --> 00:44:11,400 who wouldn't want to know about the possibility of predicting 909 00:44:11,600 --> 00:44:15,000 when re offending might occur. My sense is that electronic 910 00:44:15,000 --> 00:44:18,120 monitoring, locational monitoring by itself does not 911 00:44:18,120 --> 00:44:21,680 generate sufficient data to make those kind of predictions. 912 00:44:21,680 --> 00:44:25,000 But if you combine electronic monitoring data with other 913 00:44:25,000 --> 00:44:29,600 sources of data about that person's life, you can probably 914 00:44:29,920 --> 00:44:33,600 do predictions. So I was speaking to a Korean 915 00:44:33,600 --> 00:44:38,320 researcher just recently who is in the process of working on an 916 00:44:38,360 --> 00:44:41,640 AI prediction model for sex offenders. 917 00:44:41,640 --> 00:44:46,560 And they are going to study 5000 offenders over five years with 918 00:44:46,560 --> 00:44:49,640 200 variables attached to each individual. 919 00:44:49,960 --> 00:44:53,280 And out of that, they hope to have a predictive model. 920 00:44:53,280 --> 00:44:58,120 And I won't say that that is a bad way to go with if you were 921 00:44:58,120 --> 00:45:01,480 doing it with all offenders, that would be so Ross resource 922 00:45:01,480 --> 00:45:03,760 intensive. I would worry about the level of 923 00:45:03,760 --> 00:45:05,480 resource that was been committed to it. 924 00:45:05,800 --> 00:45:09,320 But for high risk sex offenders, I can see the point of do of 925 00:45:09,320 --> 00:45:12,080 doing that sort of thing. And the alcohol monitoring is 926 00:45:12,080 --> 00:45:15,200 going reasonably successfully in England and Wales at the moment. 927 00:45:15,200 --> 00:45:18,400 And there is much talk about having a monetary device which 928 00:45:18,400 --> 00:45:21,440 would tell you when a person is using drugs that they have been 929 00:45:21,560 --> 00:45:24,560 prescribed from using. Well, that's probably further 930 00:45:24,560 --> 00:45:27,280 down the line than some of it's champions think it is. 931 00:45:27,280 --> 00:45:29,600 But I don't think there's a probation officer anywhere who 932 00:45:29,600 --> 00:45:33,360 wouldn't welcome that element, that contribution to the way 933 00:45:33,360 --> 00:45:36,720 they manage offenders, who's offending is related to the drug 934 00:45:36,720 --> 00:45:38,840 use. So all of this, I think, is up 935 00:45:38,840 --> 00:45:42,120 for grabs in our more technologically inclined age. 936 00:45:42,240 --> 00:45:45,280 So advice for P quips is learning about the technology 937 00:45:45,280 --> 00:45:48,440 that's available and using it to help them develop their 938 00:45:48,440 --> 00:45:51,480 professional judgements, their understanding of the people that 939 00:45:51,480 --> 00:45:54,040 they're working with, but using it wisely and well. 940 00:45:54,040 --> 00:45:58,240 I like that term. So then Louis Technology, what 941 00:45:58,240 --> 00:46:02,320 advice would you give for new training probation officers? 942 00:46:02,440 --> 00:46:04,240 I think the first thing I'd say is if you've been newly 943 00:46:04,240 --> 00:46:05,880 recruited in this space, congratulations. 944 00:46:05,880 --> 00:46:07,760 It's an awesome like really important place. 945 00:46:08,200 --> 00:46:10,680 Super not easy. Broadly speaking, I would I 946 00:46:10,680 --> 00:46:13,720 would recommend people continue to prioritise the most human 947 00:46:13,720 --> 00:46:15,520 parts of the role. How do you build loving care 948 00:46:15,520 --> 00:46:17,760 relationships? How might you use things like 949 00:46:17,920 --> 00:46:21,240 motivational interviewing? How might you have a trauma 950 00:46:21,240 --> 00:46:24,000 informed practise? How do you observe and spot 951 00:46:24,000 --> 00:46:26,320 particular factors that can form your analysis? 952 00:46:26,320 --> 00:46:30,040 Technology can amplify that craft, but it can't replace it. 953 00:46:30,040 --> 00:46:33,840 So being ready to do the things that are most important anyway 954 00:46:33,840 --> 00:46:36,360 records decisions clearly what you did and why. 955 00:46:36,440 --> 00:46:39,320 Maybe in that context you want to record what tools you used. 956 00:46:39,320 --> 00:46:41,520 Did you use an AI tool there? Maybe it's worth flagging it. 957 00:46:41,680 --> 00:46:44,960 But equally, it's important to start being really good at 958 00:46:44,960 --> 00:46:46,480 sharpening that analytical skills. 959 00:46:46,560 --> 00:46:49,120 You're going to be able to find information more easily. 960 00:46:49,200 --> 00:46:53,160 You're able to observe people in the actual meetings and see a 961 00:46:53,160 --> 00:46:55,800 bigger picture of how they communicate, so you'll need to 962 00:46:55,800 --> 00:46:58,920 be able to extract what really matters from a large amount of 963 00:46:58,920 --> 00:47:02,640 information, and sometimes AI can surface it, but your 964 00:47:02,640 --> 00:47:05,280 judgement alone will be the thing that determines the things 965 00:47:05,400 --> 00:47:06,680 that are most important. OK. 966 00:47:06,920 --> 00:47:10,320 So then within the training probation officer programme, a 967 00:47:10,320 --> 00:47:13,280 really important element of it is reflection and reflective 968 00:47:13,280 --> 00:47:16,480 practise, and it's a key part of being a probation officer as 969 00:47:16,480 --> 00:47:19,120 well as they move forward in a professional development. 970 00:47:19,120 --> 00:47:22,320 So a question to both of you, how can reflection help 971 00:47:22,320 --> 00:47:26,320 probation officers critically engage with tools like AI and 972 00:47:26,320 --> 00:47:27,880 electronic monitoring? Mike. 973 00:47:28,240 --> 00:47:32,440 I think that is important and I guess that I experienced with 974 00:47:32,440 --> 00:47:35,240 electronic monitoring and reflective practise was that 975 00:47:35,240 --> 00:47:38,960 there wasn't much reflection on it because it wasn't something 976 00:47:38,960 --> 00:47:41,760 that probation officers have traditionally had to do. 977 00:47:42,240 --> 00:47:45,960 But I remember meeting a team of Canadian probation officers who 978 00:47:45,960 --> 00:47:49,040 had been required to work with the police to develop a 979 00:47:49,040 --> 00:47:52,160 programme for dealing with young adult car thieves, right? 980 00:47:52,200 --> 00:47:56,440 And they devised a plan and then for political reasons they had 981 00:47:56,440 --> 00:48:00,040 electronic monitoring, GPS tracking foisted upon them 982 00:48:00,040 --> 00:48:03,400 against their will, right? But they found that it was 983 00:48:03,400 --> 00:48:06,280 useful, but they had to work out how it was useful. 984 00:48:06,320 --> 00:48:10,040 And in their particular context, they had a kind of free hand as 985 00:48:10,040 --> 00:48:12,600 to how to do that. And that experience with that 986 00:48:12,600 --> 00:48:15,400 team of Canadian probation officers has marked all my 987 00:48:15,400 --> 00:48:20,160 understanding of how reflection on the possibilities here, even 988 00:48:20,160 --> 00:48:22,520 when it's something you didn't initially want to do, can 989 00:48:22,520 --> 00:48:25,480 actually give you some good results, right. 990 00:48:25,480 --> 00:48:29,560 More broadly, I think that if we are encouraging people to 991 00:48:29,560 --> 00:48:34,120 reflect on ethics, well there's a whole field called the ethics 992 00:48:34,120 --> 00:48:37,720 of punishment, which I hope is still taught to probation 993 00:48:37,720 --> 00:48:40,720 trainees, as it was in the days when I was involved in it. 994 00:48:41,000 --> 00:48:44,200 But alongside the ethics of punishment these days, because 995 00:48:44,200 --> 00:48:48,080 technology is so ubiquitous, you need part of your training 996 00:48:48,080 --> 00:48:51,960 programme on reflective practise to equip people with the ethics 997 00:48:51,960 --> 00:48:53,840 of technology. And it's interesting that 998 00:48:53,840 --> 00:48:56,720 proportionality, proportionality of punishment and 999 00:48:56,720 --> 00:49:00,920 proportionality of technology proportionality is a concept 1000 00:49:00,920 --> 00:49:03,400 that comes up in both of those fields, right? 1001 00:49:03,520 --> 00:49:07,440 And I think that because we are all immersed in our hyper 1002 00:49:07,440 --> 00:49:11,120 connected world these days and we have very little time to 1003 00:49:11,120 --> 00:49:14,840 think about the ethics of that in general, although it does 1004 00:49:14,840 --> 00:49:16,960 happen. You know, think about the debate 1005 00:49:16,960 --> 00:49:19,280 we're having about smartphones and school children. 1006 00:49:19,320 --> 00:49:23,040 So doing reflection on the the way we use technology, 1007 00:49:23,040 --> 00:49:26,320 specifically in probation would seem to me to be have become a 1008 00:49:26,320 --> 00:49:29,960 very necessary part of the way we train probation officers. 1009 00:49:30,280 --> 00:49:33,880 Thoughts from you, Louis? So in regards to helping, 1010 00:49:33,880 --> 00:49:37,960 consider what reflection can do to help with engaging with AI 1011 00:49:38,280 --> 00:49:40,160 and the likes. Keeping it simple is key. 1012 00:49:40,200 --> 00:49:43,960 What has this AI tool produced? What would you find most useful? 1013 00:49:44,000 --> 00:49:46,560 What do you want to accept, modify, and reject and why? 1014 00:49:46,640 --> 00:49:49,200 You might find you've used AI technology to service 1015 00:49:49,200 --> 00:49:51,880 information, but you've noticed it's surfaced information that 1016 00:49:51,880 --> 00:49:54,880 might have been 20 years out of date and this person has changed 1017 00:49:54,880 --> 00:49:56,320 since then. So use your professional 1018 00:49:56,320 --> 00:49:58,280 judgement. You might observe, for example, 1019 00:49:58,280 --> 00:50:01,920 this personal probation had these certain situations occur 1020 00:50:01,960 --> 00:50:03,640 and these were the behaviours observed. 1021 00:50:03,920 --> 00:50:05,200 But I was in a different context. 1022 00:50:05,240 --> 00:50:07,640 That was maybe when they didn't live with their family, they 1023 00:50:07,640 --> 00:50:09,320 lived alone. It's in a different city. 1024 00:50:09,360 --> 00:50:13,000 So use information that you find, but be ready to ground in 1025 00:50:13,000 --> 00:50:15,360 the truth of today. That can mean as well. 1026 00:50:15,440 --> 00:50:18,480 You know, this is something that's there's an assistant, but 1027 00:50:19,120 --> 00:50:23,560 you're the person as a person to see the bigger picture, to know 1028 00:50:23,560 --> 00:50:25,120 what's really most important right now. 1029 00:50:25,560 --> 00:50:27,960 And you might have further information up here in your mind 1030 00:50:28,520 --> 00:50:30,560 that would definitely change the decision. 1031 00:50:30,560 --> 00:50:33,320 The information analysis and AI isn't perfect. 1032 00:50:33,400 --> 00:50:36,840 There will be times when it's not going to accurately do a 1033 00:50:36,840 --> 00:50:38,640 thing. For example, with transcription 1034 00:50:38,640 --> 00:50:40,920 summarization, need to bear in mind that it won't be perfect 1035 00:50:40,920 --> 00:50:42,960 for things like names of people, places. 1036 00:50:42,960 --> 00:50:45,680 So we've got to be very mindful that things like transcription 1037 00:50:45,680 --> 00:50:47,920 summarization, it only hears what's been said in the room, 1038 00:50:48,040 --> 00:50:50,280 doesn't see the person, it doesn't know the bigger picture. 1039 00:50:50,360 --> 00:50:53,280 And that's just an example of many where with the person using 1040 00:50:53,280 --> 00:50:55,240 it, the human is the one of meaningful control. 1041 00:50:55,240 --> 00:50:57,920 They need to be ready to say it's not going to be perfect all 1042 00:50:57,920 --> 00:50:59,200 the time. And for them, it's something to 1043 00:50:59,200 --> 00:51:01,000 assist. They can extract the things that 1044 00:51:01,000 --> 00:51:03,520 most useful, move things that aren't useful, whatever suits 1045 00:51:03,520 --> 00:51:05,520 them best. So yeah, identify what's been 1046 00:51:05,680 --> 00:51:06,880 produced, what you would use and why. 1047 00:51:07,000 --> 00:51:09,600 Make sure you're ready to change the output to be analytical and 1048 00:51:09,600 --> 00:51:11,400 recognise that it will have limitations. 1049 00:51:11,560 --> 00:51:13,080 Thank you, Lou. Yeah, that's really helpful. 1050 00:51:13,080 --> 00:51:16,280 And Kyle actually earlier point about ongoing training, but also 1051 00:51:17,040 --> 00:51:21,040 the pilot study that you talked about, whereas it's not AI tool 1052 00:51:21,040 --> 00:51:24,120 that just says yes, that's a good idea like ChatGPT does that 1053 00:51:24,200 --> 00:51:26,880 encourages more thought provoking and questioning the 1054 00:51:26,880 --> 00:51:28,720 data. So that's helpful. 1055 00:51:28,720 --> 00:51:30,760 Thank you. This is the point in the podcast 1056 00:51:30,760 --> 00:51:34,400 where we give you a magic wand and say if you could implement 1057 00:51:34,400 --> 00:51:38,440 one big change in how technology is used or developed in 1058 00:51:38,440 --> 00:51:41,520 probation, what would it be? Louie, do you want to go first? 1059 00:51:41,520 --> 00:51:41,840 Sure. So. 1060 00:51:42,960 --> 00:51:44,960 I think the most important thing for me is about how it's 1061 00:51:44,960 --> 00:51:46,800 developed. I think for all teams doing 1062 00:51:46,800 --> 00:51:49,360 anything of technology, you've got to fall in love with the 1063 00:51:49,360 --> 00:51:52,000 problem you're trying to solve and obsess about it, build with 1064 00:51:52,000 --> 00:51:55,120 the people who have the problem. If I didn't spend half my time 1065 00:51:55,200 --> 00:51:58,200 working alongside probation officers, then how do I truly 1066 00:51:58,200 --> 00:52:01,480 know that what I'm building is truly a useful, winning product? 1067 00:52:01,600 --> 00:52:06,200 So for us, it's really key that we work with people like me who 1068 00:52:06,200 --> 00:52:10,000 share that vision to learn fast in a Safeway, to learn directly 1069 00:52:10,000 --> 00:52:12,320 from the frontline probation officer have been super candid 1070 00:52:12,320 --> 00:52:14,240 about what works but doesn't, and that opens to try new 1071 00:52:14,240 --> 00:52:15,800 things. But whenever we develop a 1072 00:52:15,800 --> 00:52:18,160 solution, even beyond technology, make sure we fall in 1073 00:52:18,160 --> 00:52:21,280 love with the promise of and build something that people want 1074 00:52:21,280 --> 00:52:22,840 in the 1st. Place, Yeah, Thank you, Leo. 1075 00:52:22,840 --> 00:52:24,720 And I think it is really encouraging to hear that you're 1076 00:52:24,720 --> 00:52:28,480 going into probation officers, working with frontline staff and 1077 00:52:28,480 --> 00:52:32,320 finding out really about the job and what might be helpful in the 1078 00:52:32,320 --> 00:52:33,640 future and their role. Thank you. 1079 00:52:34,000 --> 00:52:37,240 How about you, Mike? So if I had a magic wand, the 1080 00:52:37,240 --> 00:52:40,400 one thing that I would push for at the present moment in time, 1081 00:52:40,440 --> 00:52:44,040 as I've said earlier on, is to incorporate electronic 1082 00:52:44,040 --> 00:52:48,120 monitoring into the statutory probation service and treat the 1083 00:52:48,120 --> 00:52:51,400 probation service as a potentially technologically 1084 00:52:51,600 --> 00:52:56,080 sophisticated organisation which can use technologies wisely and 1085 00:52:56,080 --> 00:52:59,840 well with all the safeguards and constraints that Louis has 1086 00:52:59,840 --> 00:53:01,720 talked about. And that would be such a sea 1087 00:53:01,720 --> 00:53:05,400 change for the England and Wales Probation Service, given that 1088 00:53:05,760 --> 00:53:07,600 it's commonplace nearly everywhere else. 1089 00:53:07,920 --> 00:53:11,800 But that would seem to me to be a good base on which you could 1090 00:53:11,800 --> 00:53:15,480 then think about incorporating other technologies into 1091 00:53:15,480 --> 00:53:19,400 probation practise as well. And if my magic wand can just be 1092 00:53:19,400 --> 00:53:22,720 spun once again, it is learned from the mistakes of 1093 00:53:22,720 --> 00:53:25,560 implementing electronic monitoring in the way we've done 1094 00:53:25,560 --> 00:53:28,600 it up to now, and make sure we don't make the same mistakes 1095 00:53:28,600 --> 00:53:32,560 with the much more complicated and sophisticated technology 1096 00:53:32,880 --> 00:53:35,160 that is embodied in. Thank you, Mike. 1097 00:53:35,240 --> 00:53:39,440 And linked to that, really, do you have any articles or a book 1098 00:53:39,440 --> 00:53:42,960 or any resource that you would recommend our P quip learners 1099 00:53:43,000 --> 00:53:46,840 read or anyone interested in the probation service, what one or 1100 00:53:46,840 --> 00:53:48,480 two key articles would you recommend? 1101 00:53:48,720 --> 00:53:52,800 Well, there's lots of articles about electronic monitoring and 1102 00:53:52,800 --> 00:53:54,720 the way it can be used well in probation. 1103 00:53:54,720 --> 00:53:56,920 I mean, some of those I've written myself, but lots of 1104 00:53:56,920 --> 00:53:58,600 people have written those things. 1105 00:53:58,920 --> 00:54:02,280 But to help and to encourage probation officers to think more 1106 00:54:02,280 --> 00:54:06,040 generally about the the changing world in which we all live in 1107 00:54:06,040 --> 00:54:08,120 the way that this might impact on probation. 1108 00:54:08,600 --> 00:54:13,440 The book that I have found most useful as a layperson is by a 1109 00:54:13,440 --> 00:54:17,560 barrister called Susie Allegre, and it's called Human Rights and 1110 00:54:17,560 --> 00:54:21,080 Robot Wrongs, Being Human in the Age of AI. 1111 00:54:21,080 --> 00:54:24,880 And that addresses very directly the question of overselling AI 1112 00:54:24,880 --> 00:54:27,800 proportionality in AI. You know, what does it mean to 1113 00:54:27,800 --> 00:54:32,520 preserve human skills and values in a context where technology 1114 00:54:32,520 --> 00:54:35,240 can be presented as as such an easy solution? 1115 00:54:35,320 --> 00:54:36,160 Right. Thank you, Mike. 1116 00:54:36,160 --> 00:54:39,400 That sounds really interesting. And Louis, any recommendations 1117 00:54:39,400 --> 00:54:41,200 from? You as a recommendation for 1118 00:54:41,200 --> 00:54:47,120 those who are decrypt, if you do a simple Google AI action plan 1119 00:54:47,120 --> 00:54:49,960 for justice, I would, you know, recommend maybe we can hold 1120 00:54:49,960 --> 00:54:52,880 plan, but you've also got a publicly available website and 1121 00:54:52,880 --> 00:54:56,240 it talks to the plan in a more visual way and it also showcases 1122 00:54:56,240 --> 00:54:58,560 the stuff that's actually happening in the in realities. 1123 00:54:58,600 --> 00:55:01,600 So I'll say get started there. It gives you a sense of what we 1124 00:55:01,600 --> 00:55:04,760 are looking to support and progress in the context of their 1125 00:55:04,760 --> 00:55:05,920 working lives. Thank you. 1126 00:55:05,920 --> 00:55:10,360 We'll also make sure both the book and that web link is 1127 00:55:10,360 --> 00:55:13,200 included in our show notes. Thank you very much to both of 1128 00:55:13,200 --> 00:55:16,280 you for those recommendations. So we've, we've come to the end 1129 00:55:16,280 --> 00:55:17,760 of the podcast. We've had some really 1130 00:55:17,760 --> 00:55:19,840 interesting discussions. Thank you for everything that 1131 00:55:19,840 --> 00:55:21,800 you've shared. But is there anything important 1132 00:55:21,840 --> 00:55:24,400 that you came here to say that you haven't had a chance to say 1133 00:55:24,400 --> 00:55:27,160 yet? We haven't addressed the buzz 1134 00:55:27,160 --> 00:55:30,640 phrase human in the loop, which is very beloved of the Council 1135 00:55:30,640 --> 00:55:32,760 of Europe and not just the Council of Europe. 1136 00:55:33,120 --> 00:55:35,880 The human in the loop is supposed to be the person who 1137 00:55:36,160 --> 00:55:41,240 makes sure that any knowledge that is generated by AI in terms 1138 00:55:41,240 --> 00:55:44,520 of risk assessments or whatever, there is always a human 1139 00:55:44,520 --> 00:55:46,920 assessment of its value and viability. 1140 00:55:47,080 --> 00:55:50,360 And that seems like a good idea. That's a major feature of all 1141 00:55:50,360 --> 00:55:52,440 the Council of Europe approaches to it. 1142 00:55:52,440 --> 00:55:55,760 But I'm interested in who the human in the loop would be in a 1143 00:55:55,760 --> 00:55:59,120 future probation service. Because to manage this 1144 00:55:59,120 --> 00:56:02,960 technology, you need to know a lot about this technology. 1145 00:56:02,960 --> 00:56:05,560 And I'm not sure that there seems to me to be a tension 1146 00:56:05,560 --> 00:56:08,400 between the kind of experience and knowledge that you would 1147 00:56:08,400 --> 00:56:12,480 want the human in the loop to bring in a probation context and 1148 00:56:12,480 --> 00:56:15,840 the kind of person who really understands how the technology 1149 00:56:15,840 --> 00:56:17,800 is working. Who is the human in the loop? 1150 00:56:18,200 --> 00:56:21,040 That's a good question, I think. It's a good, good question to 1151 00:56:21,040 --> 00:56:24,320 almost end on. So really important point if you 1152 00:56:24,320 --> 00:56:27,200 want to speak to that, Louis or if there's anything else that 1153 00:56:27,200 --> 00:56:29,240 you came here to say that we haven't managed to say. 1154 00:56:29,440 --> 00:56:32,280 Yeah, sweet. So 2 points in the Justice AI 1155 00:56:32,280 --> 00:56:35,480 unit, we've started going beyond the language of human and loop. 1156 00:56:35,560 --> 00:56:38,160 It came across a bit like implying there's automation and 1157 00:56:38,160 --> 00:56:41,360 you have visibility of it. But we start using language 1158 00:56:41,360 --> 00:56:43,240 which you would hopefully have heard me say a lot in the 1159 00:56:43,240 --> 00:56:45,280 discussion today, is meaningful human control. 1160 00:56:45,480 --> 00:56:48,120 Like we don't want people to simply have in loop. 1161 00:56:48,200 --> 00:56:50,120 We don't want people to simply be able to observe. 1162 00:56:50,160 --> 00:56:52,400 It's more like, wow, I'm in a driving seat here. 1163 00:56:52,520 --> 00:56:55,720 I'm choosing more AI tools or whatever M tech tools to use and 1164 00:56:55,720 --> 00:56:57,400 what context and how to use invest. 1165 00:56:57,440 --> 00:56:59,920 It isn't easy to simply pick up a load of tools and know how to 1166 00:56:59,920 --> 00:57:01,640 use it like that. It's something that needs to be 1167 00:57:01,640 --> 00:57:03,800 baked in as part of that product, as part of that 1168 00:57:03,800 --> 00:57:05,680 experience to sign up in the 1st place. 1169 00:57:05,760 --> 00:57:07,480 So yeah, that's a fair thing to flag. 1170 00:57:07,920 --> 00:57:09,760 And we will hear a lot of people say human loop. 1171 00:57:09,760 --> 00:57:11,800 And my challenge then was we actually want to give them 1172 00:57:11,880 --> 00:57:14,560 something bigger, older, meaningful human control on my 1173 00:57:14,560 --> 00:57:15,960 mind. Secondly, the only thing that 1174 00:57:15,960 --> 00:57:19,320 wants to flag again was just like a massive thanks to all of 1175 00:57:19,320 --> 00:57:22,960 the awesome probation officers, P Crips, PSOSSPOS that I've 1176 00:57:22,960 --> 00:57:26,200 worked with in Wales, Kent. So in Sussex, so far it's been 1177 00:57:26,200 --> 00:57:27,800 awesome. Keep them open minded, they're 1178 00:57:27,800 --> 00:57:30,240 willing to try something new and give us their honest feedback. 1179 00:57:30,320 --> 00:57:32,760 So I'm really grateful for that. The last thing that I'll leave 1180 00:57:32,760 --> 00:57:36,640 with is to say at the Justice AI units, we're still quite young 1181 00:57:36,680 --> 00:57:39,280 and we are recruiting. So if anyone's listening and 1182 00:57:39,280 --> 00:57:42,720 they're super interested to get involved, be it as an engineer 1183 00:57:42,880 --> 00:57:46,120 or a product person, an adoption person, get in touch. 1184 00:57:46,360 --> 00:57:47,680 They'll be super interested to hear from. 1185 00:57:47,760 --> 00:57:48,640 You. Thank you, Louie. 1186 00:57:48,640 --> 00:57:50,440 Thank you, Mike. It's been really great having 1187 00:57:50,440 --> 00:57:52,880 you on the podcast and really interesting for Addison. 1188 00:57:53,200 --> 00:57:54,640 Thank you. It's been great. 1189 00:57:57,320 --> 00:57:59,920 Thanks for listening to Probation Perspectives Insights 1190 00:57:59,920 --> 00:58:02,840 into Community justice. We hope our conversation with 1191 00:58:02,840 --> 00:58:06,560 Professor Mike Nellis and Louis All Good sparked new thinking 1192 00:58:06,560 --> 00:58:09,040 around the role of technology and probation practise. 1193 00:58:09,760 --> 00:58:12,880 From the history of electronic monitoring to the emerging use 1194 00:58:12,880 --> 00:58:16,560 of AI, we trust this episode offered insights into both the 1195 00:58:16,560 --> 00:58:19,920 opportunities and the challenges that come with tech driven 1196 00:58:19,920 --> 00:58:22,200 justice. If you found this episode 1197 00:58:22,200 --> 00:58:25,080 valuable, please subscribe, leave a review and share it with 1198 00:58:25,080 --> 00:58:27,520 colleagues or whoever you think would benefit. 1199 00:58:27,840 --> 00:58:30,360 And do let us know your thoughts or who you'd like to hear from 1200 00:58:30,360 --> 00:58:32,800 next. A big thank you to Mike and 1201 00:58:32,800 --> 00:58:35,680 Louis for sharing their expertise and to the team behind 1202 00:58:35,680 --> 00:58:37,280 Probation Perspectives for making. 1203 00:58:37,280 --> 00:58:40,160 This podcast possible. We'll be back soon with more 1204 00:58:40,160 --> 00:58:43,040 conversations that shine a light on the practise, purpose and 1205 00:58:43,040 --> 00:58:45,360 people behind probation and community justice. 1206 00:58:45,600 --> 00:58:47,360 Until next time, take care.