Exploring Artificial Intelligence in Social Work—Ethics, Opportunities and Risk

Welcome to Let's Talk Social Work. I'm Andy McLenahan and in this episode my guests and I will be discussing artificial intelligence in social work. We'll explore the concept from three angles. How might AI play a role in social care in terms of direct support for people who use services? How might it assist social workers in their practice? And finally, how is AI being used in the education of social workers?

Throughout the discussion, we'll focus on how AI is already helping, how it may help further, and what issues must be considered in terms of the risks associated with the use of artificial intelligence. Helping me come to a better understanding of the role of AI in social work are social workers Tommy Henderson-Ray, Digital Engagement Programme Manager at NHS England's Digitising Social Care Programme.

Julia Ross, chair of Basel UK and author of the novel The Laughing Robot, and Dr Paul Best, director of the Centre for Technological Innovation in Mental Health and Education at Queen's University Belfast. Before we get into today's conversation, I want to highlight two upcoming events which will be a great opportunity to discuss face-to-face the use of AI in social work.

The Baswa England Conference will feature a session on the 21st of October exploring AI in social work, both in the UK and Europe, and examining the ethical challenges AI poses. At that session, you'll be able to pose questions to Tommy, who is part of today's episode, as well as other speakers.

Secondly, as part of Baswa England's 80-20 campaign, there will be an online event on the 17th of November at which a suite of resources will be launched, which are aimed at supporting the transition to scenario in which social workers can spend the majority of their time working.

working directly with children, families and adults rather than on administrative and process-driven tasks. The suite will include a targeted resource setting out top tips to enable relationship-based social work in digital transformation.

and it is expected discussion around AI will feature prominently in that event. That was a very long introduction. I'm not normally giving over to those. Welcome. I want to welcome my guest. Julia, it's great to have you back on Let's Talk Social Work. How are you doing? Delighted.

to be here um i think it's i think it's important to remember that although though i am chair of buzz for uk i'm also a social worker through and through and through uh and um very excited at the thought of artificial intelligence and what it But that changes, those changes can bring for the people that we work with and indeed for us. Absolutely. And I'm really excited to get into the meat of this conversation, Julia. Tommy, how are you doing? Are you well?

I am. I've spent three days touring the country talking about different topics, but not AI. So I'm really keen to delve into this two-letter. concept that seems to be the zeitgeist and really understand different people's perspectives as to what we do, what we shouldn't do and understanding concerns around their topic. Wonderful. And last but not least, definitely not least, Paul, my competitors. How are you doing?

I'm good. Andy, it's great to be here, and hopefully you won't have to do too much translation for my very thick Northern Irish Belfast accent during this conversation, but I think this is a really interesting in time to be involved in social work education so I'm looking forward to having a good chat with everyone here just to see what our different perspectives are on the use of AI and where it's going.

where we should use it, maybe where we shouldn't use it and what might be the road of ourselves as educators as I suppose for a lot of social workers their first maybe opportunity to kind of think about the use of AI within social work practice. So yeah, great to be here. Wonderful. And just in terms of accents, my little boy had said to me recently, I don't like sit down and playing the podcast, but he must have listened to me doing some edits or something. And he's like, Daddy...

what's that voice you're doing? That's not the way you normally speak. I said, oh, that's my podcast voice. So yes, there you go. Right, listen, let's get into it. Artificial intelligence, you know, it strikes me as one of those issues which some people have a really in-depth

depth knowledge about and they're really excited about and interested in. But for many people, it's a fairly vague concept and one which can be difficult to tie down. So I really want to get into concrete examples, first of all, of AI and social work in 2024.

You know, just in the last month, there's been news coverage of the pilot testing of an AI tool named Magic Notes, which is designed to record and analyse face-to-face meetings. And I believe it's been tested with adult services social workers, I think in Swindon. Tommy, can you tell me more about this? Is this a palette you're familiar with? It is, yes. So I have demoed it three times in different settings because...

Exactly as your intro alludes, Andy, I think AI is a very aloof concept. And so I wanted to get people to see it in action. And then that helps people critique it for themselves. So I've used it three times, demoed it three times. in the scenario where we've done effectively short snippets or mock snippets of a care needs assessment and I've done that with colleagues who have acted out working with an older person.

working someone working in supported living and then somebody who has concerns around their mental health conditions and so what we've what we've done is we've demoed that and the the software captures quite accurately it seems what's being articulated the concerns and feelings of that

individual has and the conversation that flows between the social worker and the person in receipt of care tell me tell me tell me the process tell me the process just because again nebulous we're talking about what it does but how does it do that So it uses a mobile device or it could use an iPad or a mobile and it records the voice like...

Lots of apps do or activities do on mobile phones. So it literally records the conversation and then develops the text that goes with that so that at the end of that conversation and you stop record, then. the text is on the screen and you can see what's been discussed quite clearly. And then the other bit that's quite interesting is that not only does it transcribe what has been said between person A and person B or person C,

But it also then says, here are the action points that you discussed within your meeting, your conversation. So the AI is... taking the words that have been said and generating that into a series of action points that the social worker would then take away with them to form what they do next as part of that support piece.

And as I understand, those are recommendations, they're prompts, they're by no means something that the social worker has to act on. The decision is still the social workers. Absolutely. So it's not once it's done, that's it. read-only, you can't edit it. It's used as a guide and a framework and a template. So it would be that the worker then has that information to then look at, review, amend.

and check for accuracy and then take that forward and then once they're happy with that they can then upload that into the case notes that they would normally use. I think that's absolutely fascinating. I wasn't aware that you'd done it in so much detail, Tommy, so I'm really delighted to hear it. Do you know what it reminds me of? It reminds me of when I trained.

quite a few years ago. It reminds me of the process recording we did. So I used to have to do a word by word process recording for my supervision as a trainee. children and family social worker and it was there and then i had to extrapolate from that this would have done it all for me um because it's not that the recording bit is what's so important and so laborious And I was going to ask, is this sort of typical of AI in social work practice or...

Maybe I should rephrase the question, Tommy. Is there an awful lot of AI in social work practice? I'm going to talk to Paul shortly about AI in social work education. I know some of what's going on there. But in terms of practice applications, is Magic Notes the only... example we can think of at the moment no so as you can imagine um lots of organizations and suppliers and tech developers are thinking about it so if

You have Magic Notes. You could also look at Microsoft Copilot, who have developed a very similar set of software. Copilot is very much embedded into... Microsoft Teams and Microsoft Suite that you can use. So we know that Barnsley Council are using Copilot. We know that 28 local authorities are using it. a version of either co-pilot or magic notes according to community care magazines from last week before that i think it's quite interesting to note that

I know that from personal perspective as a social worker, the resource allocation system, RAS, as it's often called, within the case management system, has been around for a long, long, long time. And that is using... a system to generate effectively monetary value towards a care plan. So that's been in existence for a long time. So some form of artificial intelligence has been in existence and used by local authorities and social workers for a very, very, very long time.

Whether social workers knew that's what it was, I'm not sure, but this is a very different iteration of what we would consider artificial intelligence, much more in your face, I think.

And when we're coming back to Magic Notes, and Magic Notes is sort of analysing the conversation and it's pulling out the action points, making recommendations. I think something which we really need to be clear about here at the moment, there's a discussion around, can AI think? And my understanding is that AI does not think the way a human being thinks. It's about an assessment. It looks at data sets.

And just on that point, I had looked up a couple of AI chatbots. People will be very familiar with ChatGPT. ChatGPT has kind of become a wee bit like... ubiquitous, as if it's a generic thing. People talk about Photoshop. For graphic design, Photoshop is one specific piece of software. ChatGPT is one of many chatbots that do similar things. ChatGPT, you ask it, can it think? I'm gonna tell you what it says. No.

It says it can't think. It's not able to think the way humans do. There's a long explanation, but it says no. There's another chatbot called Claude AI, which I hadn't been familiar with until Paul and I had a chat a while ago.

Claude AI gives you quite a different answer and says that it's a complex philosophical question without a simple answer. It talks about how there's significant debate amongst experts about whether current AI systems like Claude truly think in his verticommas. I responded by saying...

No, there's actually not a significant debate amongst experts. It's fairly well understood that AI doesn't think. Oh, thank you for that clarification. I think my answer is misleading. There is not a significant debate amongst experts about whether AI systems think the way humans do.

which I find really unsettling. People will often go to these things and ask them questions and take the answers as if the answers are factually correct. And they respond. They respond to you when you point out the mistakes they've made.

But in a sort of, in a way that to me feels like a really kind of obsequious kind of like assistant who just wants to tell you you're right all the time. Do you know what I mean? I'm thinking, can we move on? I'm just, Paul, it'd be really helpful actually this stage. Can we talk a little bit about...

your use of AI in social work training. So you're working with undergrads and postgrads at Queen's University, is that correct?

Can you give us an example of how you have integrated AI into the syllabus and how you're training social workers? Yeah, absolutely. And it's really interesting actually to hear some of the debates about, I haven't used Magic Notes just... i suppose red on it so it's really helpful to me to get that example i suppose probably of a more maybe traditional ai rather than generative uh ai which we'll maybe get into a bit

kind of later on and the kind of different kind of rules and maybe some of the ethical considerations about those two approaches. Within the education field, it's both, I suppose, scurry and... and kind of fascinating in terms of what AI can offer. There's still definitely a kind of a growing sense and sort of anxiety around this technology, particularly when it comes to how it interfaces or interacts with.

traditional kind of processes within universities and our kind of teaching structure around still largely actually weighted on assignment-based assessments and what AI is going to do within that. So there is quite a bit of unease. about that. My work though I suppose has really been about through the research centre that I direct which is really focusing on technological innovation and how we can practically

use that is about trying to see what these technologies offer us and really kind of test and experiment with them in sort of safe settings. So where we are really getting some benefits from AI. two made fields related to my work and that's around helping content generation in terms of virtual reality simulations. So we find it quite a useful kind of technology when we're working with industry partners to speed up that process of helping students do sort of virtual home visits and things.

So it can be really good for asset generation within that context. Where I'm really excited about where we're going and there's a project that we are undertaking at the moment with colleagues. across multiple disciplines based in Canada and Taiwan is around conversational AI which is sort of like an offshoot or a type of generative AI. pretty much what we're doing now, this ability to have a back and forward conversation with a virtual avatar.

And we're finding that this is going to be, I think personally, it's going to be a bit of a game changer than social work education. So if I'm giving you an example, one of the health trusts that we're working with, the Northern Health Trust,

in Northern Ireland have got a batch of newly qualified social workers. What they're finding is that they they want these newly qualified people to be as comfortable as possible going into homes and asking difficult questions so there's a real kind of learning need there around professional curiosity. The opportunities to practice that within the university environment tends to be quite stale. It's role-play orientated. We might bring in drama students and things are quite limited.

And so they're working with us at the moment to develop a virtual reality simulation in which we'll use AI to generate some of the content, but also then bringing in conversational AI. where we can have a back and forth conversation with a virtual avatar in order to help them develop those skills. Paul, in terms of how that works, the data bank that actually facilitates the AI to do that conversational, where does that information come from?

Like, how does that work? Okay, so I suppose stripping it back a wee bit in terms of where AI is in a minute. And interestingly enough, there was a report. released last week that shows the gaps actually closing between OpenAI and the ChatGBT and the other ones like Claw, which you've already mentioned. So there is a bit of a narrowing of that gap.

And you might have noticed within the last week or two, ChatGBT have now released this advanced mode, which means you can have like an audio kind of conversation and it's starting to pick up tones and respond back.

So really trying to get a bit of emotion into the conversation. So what we do is, so this technology that's being developed anywhere in the background, the process or the way it works in terms of our simulations is we take that, we bring that into a... a separate platform and say you have an environment it could be just a virtual environment on a desktop or in a virtual reality environment of say a family in a living room

our environment what we do is we essentially each avatar is a blank character and we just drag and drop the content in so if you take a traditional case study this is mom this is what the age mom is these are sort of the family situations and all the different kind of maybe there's depressions maybe some symptomology about that what's going on and we essentially just drag that into mom and that's mom's background and then when you go in and have a conversation

conversation with mom you're that avatar will just react based on the information you've given it and that's really as i said earlier that's that's a real like game changer for us because traditionally what we had to do is create more of a branching narratives approach where we would have to have pre-programmed the question and the response and let the student navigate through that.

They just try and get a bit of interactivity or a bit of decision making. Now we don't need to do that. I'm absolutely fascinated by that, Paul. I'm really excited by it too. So the use of chat GPT is... growing and changing almost as we speak. I'm doing some voluntary work with a charity called Heart and Soul down in Deptford. What they've done is use a university to adapt chat. PT to be able to.

create a tool that they're calling the jargon buster and the jargon buster they're doing it with the money from the health foundation the jargon buster is allowing them to create a tool that agrees or doesn't agree and if they don't agree with what's being said

then the chat GPT has to come up with another solution. So they understand that jargon busting is, you know, let's understand for people with a learning disability in particular, what's being said. And it's brilliant. It's revolutionized how they...

how they talk how they how they deal with problems how they they they articulate things how they have conversations really really exciting um so i hope that can be taken further forward i did have one question for you if that's okay because it seems to me that a lot of these

developments are taking place within the the fairly not static world but safe world of adults reasonably safe world of working with adults working with children and families because of children children's access or those that are not disadvantaged to the media and social media might be very, very different. Do you have experience of working with children and families and social workers or is it mainly adults?

Well, my practice background, obviously, is social work, but I'm also a cognitive behavioral therapist. So I still keep a day a week in practice. So it's mostly adults is where my sort of... clinical work is, if you like. But in terms of that project using the AI, there's four different settings. One is assessing suicide risk.

with an A&E department. One is the professional curiosity of a family home visit so that will have a family in it. The other one is a really interesting school-based scenario which is for children and neurodiverse children and having a conversation with them and then the final one is a family therapy environment and this is where this gets kind of

I think more sophisticated, more we could go with this technology. So in one of the scenarios, we can actually introduce multiple avatars in the room, but actually have the avatars using AI interact with each other, argue with each other, contradict each other. in that live space and then challenge the student to unpick that. Other conversations we've had about what happens when you get a disclosure, a child discloses.

Yes, there's a specialist kind of training and approach to talking with the child, but in the first immediate disclosure, the social worker still has to deal with that interaction. We could generate a safe environment in which... the student could do that without having to either bring an adult in to pretend to be a child or to try and find a child, which would be unethical to do, to act out that scenario. So it's really the potential here.

so exciting in terms of what we can unlock. I get that. I'm just thinking about, let's say you were just about to embark on your first practice. placement with young children who were asylum seekers who didn't have great English and you could practice with an avatar beforehand or something like that, or another scenario that might be slightly different. I just wondered, Paul, do you have a sense as to...

If you've practiced with an avatar, is there a jarring when it comes to actually talking to people in real life? Or do people feel that it was great preparation and they were better for it? What would you say? That's an excellent question. And I suppose we don't know because we haven't got to that stage yet. So if I suppose where can I keep up with the technology? So in terms of the conversational AI, the way it was explained to me.

If you use it now, there is a latency issue in terms of you ask the avatar question. You might have to wait a second or two before a response. And essentially what's happening is... you're verbally talking to the avatar it's taking what you say it's converting into text to understand it then uh you know

typing up a response essentially then going back into audio to say it back to you so there's a bit of a gap and these recent updates have now got this kind of audio to audio or verbal to verbal so it's sped up the kind of So that's the version of the technology that we want to test, to ask some of those questions, because I think that's really important. How does this translate? And I mean, it might be...

And I suppose it's the same within the simulation world more generally, about does it just get us a bit closer? so it's never going to be as good um but is it better than what we currently have to offer but i think this idea of halo opportunities so these kind of

It's more of a sort of medical term around how acuity low occurrence event is one really exciting opportunity to do. But as the technology kind of progresses, I think they'll get better. But that's definitely a really important research question is, does it actually matter what we did? There was another example we have which is a more therapeutic example where the service user had a PTSD from a car accident and we were sort of...

practicing asking questions to kind of understand what was going on. Now within cognitive behavioral therapy there's a CTSR which is a rating scale that you can put against every uh every cbt session and you get a score and then there's various different domains we give the ai that ctsr and then ask the ai to rank the performance

based on that criteria so again adding another level to it so you've got practicing the interaction then getting immediate feedback from the AI but to go back to your question Tommy and I think it's a fair point that all sounds great uh as an educational experience you know how how like

Does it make students feel more confident going into practice? Does it give it a better sense of sort of role expectancy? Does it motivate them? Does it keep them more engaged? So then they take more in from that learning experience which ultimately translates. in the real world practice, those are really key questions. I think we need more research to tease that out.

Paul, thanks a million. That's fascinating. I want to move it on though. Julia, you've got a background in working in this area. You've also written a novel recently, The Laughing Robot. And just for listeners, I've read the book, it's set in a dystopian... It's either present or near future. Couldn't quite identify and no COVID happens at the same time. So I took it as present, Julia. It's set in the context of a pilot test in which care robots provide support and assistance.

to older people, to help them live independently. So this seems more like the sci-fi end of things, but can you tell me why you chose to write about this? Yeah, in some ways it's sci-fi, but it's also about the future, isn't it? That is the thing in social work we really need to focus on. We can't just focus on... The past, the current, we have to think about the future. So I think it's a real game changer. I became interested in AI when I joined a company, a startup called PredictX.

And it was about predictive analytics. So it draws attention. I'd like to draw attention, if you like, to the fact that AI is a huge range of things. It's not just, you know, this or that or the magic notes or GP. It's about predictive analytics, too, because I felt that the research that we needed, the understanding that we needed of care pathways, how you integrate care pathways and how you put that data together was absolutely key to how people live their lives.

particularly, as you were saying earlier on, Paul, in an integrated way across disciplines. So that's what got me cited. When I'd been working in it for a while, I became aware of... Bad bots, good bots, robots generally, and began investigating that and then began writing my book. So the book took about three, four years to write. And The Laughing Robot was actually published this Monday, the 7th of October.

Thank you. I'm very, very excited by it. Why I brought it was the thinking that was going on about care robots at the time coinciding with the elderly care crisis. We all know that there's a crisis in social care, particularly for older people, but actually for everyone, because it's not properly funded, it's not properly thought through. So I began to do some research. I did the research through the Nuffield Trust and then COVID happened. I planned to go out to Japan.

talked to some people in Japan about what they were doing there. So this is a lot of background research went into thinking and planning what might happen. And then COVID happened. And so I had to cancel that trip. And since then, I've been talking to enough. your trust again and updated myself and as far as I'm aware the sort of progress that they had been anticipating with Carebox hasn't really materialized and it hasn't materialized here either there are a lot a lot of universities

academic institutions who are doing research into parts of robots. But I felt my imagination... got a bit carried away. And I felt that we could really think about the workforce and how we can manage the elderly care crisis in future. And we should do that. We have a responsibility to do that. So my imagination did.

run away, went to the Isle of Wight, and I wrote and I wrote and I wrote. And this brings together all those disparate components. So care robots, elderly care crisis, things like the impact of the... private sector. And then what happens to families? This is about a family. This is about a family that is impacted by all those different things. And it looks about 10 years ahead.

Yeah, it's a story about grief and loss and betrayal as much as it's about AI and care robots. I'm curious to know how much of the central character is you, Julia? Do you see a lot of yourself in Anna? You're right. It's built on a lot of personal experience because I start as Anna, a 60-year-old woman who has worked as a social worker all her life and then moves into predictive analytics.

But it moved then into families and how families behave and how communities behave. So I moved the main protagonist, the character, onto the Isle of Wight. Her interaction with her immediate family and the people around her is really what the story is about. And then what happens when you've got venture capitalists coming into the mix, where you've got a government that is threatened by the amount of spend on older people?

What do you do when you've got this growing population? So, yeah, a bit. But then, you know, I don't have a robot, sadly. I love robots. Not yet. Don't have them. Not yet. Not yet.

So you mentioned Japan, and I think that was obvious when I was reading it. I was thinking of what I'd read about carebot trials in Japan. I read an article, now this wasn't an academic article, so it was not peer-reviewed. I suppose it was more of an opinion piece. But it was a journalist commenting on carebot trials in Japan.

And their take was that often the care workers were finding the care bots were actually leading to more work. Things were breaking down, things needed to be moved, things were cumbersome. So in that person's assessment, essentially, there wasn't...

a benefit to the workforce, this was actually potentially a detriment. The other thing I'm kind of worried about though, Julia, is, you know, we underpay care workers. It's very hard to recruit care workers. And my concern is, you know, Some would argue that the automation of adult social care could actually serve to continue the devaluation of care work, that it's seen as a less than human task, it becomes a robotic task.

Are we trying to seek a technical solution to something we actually have a real-world practical solution to, which is let's better pay care staff, let's make caring professions more desirable careers for people? I think it's a great question. I think we're a way off robots in home care or residential care being used with carers in this country, partly because of the fragility of the market and providers don't have the money to do it. But secondly, I think the bigger point is,

to a social worker, occupational therapist, a care worker, registered manager, a director of adult social services, whoever you are in social care, children's social care, it's the... confidence of people to critique AI as part of their day job, the ability to say it exists, yes. but in what context do I want to use it? And having those questions to say,

Yes, I would use it here, but not here. And that based on confidence. And my concern at the moment is that AI is a very fast train that is going very, very fast. but we have neither the regulation nor the skills, confidence of the workforce to critique it to the extent that it's developing. And that does worry me. I'm really mindful for listeners that...

For myself, Paul and Julie particularly, we are using and have used it. But there will be others who are our contemporaries who are really cautious and hesitant or even against the use. and it's really important for us to understand those narratives. I think that's absolutely right, Tommy. And that's why we need to talk about it. So we at our next Baswa Council, we're going to be looking at the application of AI generally.

But generally, because it's important, it's the big picture. But we also need to be aware that people aren't there. And that's...

Not only fine, it's good. There needs to be a debate about the pros and the cons and the how and the ethics. So building up, I think. One of you has done some work on the ethics of this, and I badly want Basber to do that. So we get into the... We get into the nitty gritty of it and understand what is important and how it's important for people. Because actually, if you go to people, social workers, okay, they may be reluctant. Some of them will be for very good reasons.

But if we think about people and what they want, some people won't want robots. That's fine. Some people want the caring, warm hands and relationships, nurturing, that emotional intelligence that excels so well in social work. They need and want all of that. Some people don't. And if you look at very lonely older people, for example, there are some new devices now that can have real conversations with people. And the same with children. Not real conversations with people, Julia.

they have artificial conversations with people. And this is a question I really wanted to get onto. You know, we are experiencing what's been termed a loneliness epidemic you know the world health organization last year declared loneliness to be oppressing global health threat there's that stat around you know chronic loneliness being as dangerous as to your health as smoking 20 cigarettes a day or whatever it is you know in that context

Some, I'm guessing you might, from what you were saying there, see AI chatbots as a potential solution to that issue. I think that, to be honest, this is just my personal opinion, makes me really uneasy. As a society, surely we have human responses to this. Social work can have a role in terms of supporting people to have real meaningful engagement, real human-to-human engagement, not talking to a machine.

And there would be the argument, the talking to a machine is a panacea, it's a crutch. You're not actually having any real engagement. That's what I'm thinking. I want to hear what you think. I demoed something yesterday. called Eva Cares. And that was a fascinating thing. It's run in Nottingham at the moment, and it's an AI bot called Eva who can talk to you at length for as long as you want.

And it will articulate day after day and memorise and remember what you have said the day before and connect you with that the following day as a way of addressing. uh identified loneliness in support of living in nothing it's uh i haven't tried it yet but i've i've seen uh its existence yesterday so again it happens but it's probably happening andy because

There aren't enough social workers, support workers, voluntary organisations with the capacity to do it. But that's the point I'm getting at. You know, we have run down, we've had, we have a new government. hasn't shown an awful lot more vigour to address the austerity in public services than the previous government we had for 14 years. But we had public services run down for 14 years, lack of investment.

The reason we don't have the services to support the older people is not because there is not the ability, the human ability to do it. It's because there was never the political will to fund it. there was a political will to defund it. So what I'm getting at and the question I'm posing, and I would like this to be challenged, we are seeking technical solutions to a human-created problem which has human solutions.

that don't need technical solutions could be supported by technical solutions. Absolutely. You know, the idea of, you know, you saw us during the pandemic, how useful video calling is. I mean, this is amazing. The podcast. Paul's in Belfast, I'm in Belfast, Julia and Tommy both in England, you know, we're having this conversation. Technology can support human connection in such a wonderful way. I just, I mean, uneasy about the idea of

technology replacing human conversation in a way that's seen as authentic when it's actually anything but. I'm really happy to come in, but I'd be really interested to hear from Paul as well, because I think the research and academic approach towards this is absolutely key to the way forward. So human intelligence and artificial intelligence.

and emotional intelligence are all part of the same package. We are the beginning of massive change and we need to be open to that change. That is really all that I would say. There'll be people of my generation who will just...

won't want anything to do with this. We'll want warm, loving, caring hands. We'll want... people understanding us and social workers will be the same social workers are people too so so i think we need both i don't think it should be either one or the other but i don't think we're i think we're pretty good at the emotional intelligence the social work support and

I think we should have more social workers. Of course we should. But we should have more communities that are supportive too. And I think we need to take into account the role that not just robots, but that all forms of art. artificial intelligence can play in our lives. Because if we don't join in engaging with that, then the people that we work with and ourselves will be left behind. And I think that would be disastrous. So there are challenges, yes.

Let's face those challenges. It's a lovely question. It's a complicated question. But I do think that...

But I was having a conversation last week with Professor Anne Campbell who is our Disciplinary Lead of Social Work at Queen's and Anne does a lot of work in around drugs and alcohol misuse, I work in hostels and We were getting a debate about AI chatbots and she just finished a bit of work with that service user group and kind of almost did a sort of 180 interviews around AI chatbots because the message was coming.

very strongly from the people that Anne had spoken to around, you know, having someone to talk to at two or three in the morning when you're in crisis is better than nothing, even if that is indeed an AI. Power Chatbot. So it is about maybe supplementing where there's service gaps. I do think though I see this as an enhancement to what we can offer. I think there's maybe a wider debate around

A lot of the intervention models, if you like, that we use were developed during a period where, you know, the internet didn't even exist, never mind, you know, AI. And therefore, you know, there is, we need to reimagine some of our approaches to things as well. Imagine social work if we were to take all the mobile phones away. But I still remember the time where there was a lot of resistance against even using mobile phones and texting.

Service users would have been unheard of as a way of communicating. So I do think we're just in that exciting period, but period where it's unknown and people are a bit uneasy. The chat pod... I mean, we are seeing it being used in so many areas now, Andy. It's really, I'm fascinated to see where it goes. There's a lot of screening tools within the mental health. Limbic AI is one of them.

that's been bought as being used actually in the NHS as a mental health screener. We're seeing Chat Pal which is developed by colleagues in Ulster University in Action Mental Health around a companion app for young people. in distress and then I think I read somewhere a few months ago that in the physiotherapy world there is an AI screener to kind of see what appropriate service you might go to for sort of a back pain so it was doing a bit of an assessment that way.

it's not going away it is there i do see it i very much am interested to see how it enhances rather than replaces um but i think that if we go in with an attitude of This is kind of muscling into our turf and we don't want anything to do with it. We'll only end up kind of shooting ourselves in the foot in terms of stifling progress. I agree. Have I painted myself a subject for Luddite?

Is that the impression I've given? Perhaps. Very clearly. It's funny. I mean, in terms of all walks of life, I mean, AI, I do a lot of design work. I've used AI tools in design work. Never too convinced. I know technology gets a lot better. One of the things, an AI tool for podcasting.

I put a lot of effort into making our podcast sound good even though it's recorded remotely and there was an AI tool which I used once and it made the other guests sound like they were in a recording studio and I was like, oh my goodness, everyone's going to have a really good sounding podcast now.

It only worked once. It's never worked since. It really hit and miss. But I'm aware things are going that way. But I want to move on. I want to talk about ethics because this is fundamentally important. Regardless of how helpful AI is, there's always going to be ethical considerations. And social work is a... as a profession which is very focused on ethics, human rights. The question I want to ask, Julia mentioned predictive analytics earlier on and the role that AI can play in that.

What I want to ask about, though, is kind of high level service planning. So if AI is potentially going to help social workers analyse large amounts of client and community data, you know, to identify trends, to identify risk factors.

and potential interventions. You could see how that could be massively beneficial for service planning and supporting decision making. But what I'm concerned to know about is if there are risks associated with historical biases in data, If AI systems learn from historical data,

If that data reflects past and present societal biases, you know, for example, if there is historical child welfare data that shows higher rates of intervention in minority communities, whether or not that intervention was justified. Could the AI then recommend more interventions in these groups, regardless of actual risk factors? And if we're talking about generative AI, is there a risk that there could actually be a creation of a feedback loop? Were bad decisions previously made?

based on bad data, reinforce further bad data and further bad decisions? That's the question. Is that a risk that we're likely to face? Yeah. Can I give you an example? which I think would be really helpful for our listeners. So I know a Chinese and Cantonese supported living setting in Manchester had trialled predictive analytics to look at hydration for their residents.

And they bought an off the shelf predictive analytics process. And what they found was that in Cantonese culture around eating and drinking, often residents will. take most of their liquid and hydration through their meals, i.e. soup-based meals, either at breakfast, lunch and dinner, and may not have... routine drinks of water and a other drink during the day as you would do if you were a white Caucasian equivalent.

And therefore, the predictive analytics couldn't handle the fact that the residents were intaking most of their liquid with their actual food. because the predictive analytics structure and bias was assuming that they should be having them in drink format, which is what the evidence that they had from the product existed.

they actually stopped using it because they were effectively having to override the system every time because the bias and the predictive analytics didn't match the cultural nuance of the way that a lot of Cantonese people. take their fluid intake. And I thought that was a really fascinating insight into what the reality of the concept of bias might look like. And if somebody using that was worded to using the tool?

then they could be making very bad decisions. I know that's one kind of, it's a very important example because it's about health. Julia. That's absolutely fascinating, Tommy, and I think absolutely right. But for me, it says two things. One, I wanted to say that bias is already there. I mean, in the research that we do, in all sorts of the... what we count, how we count it. It has always been very much a male orientation towards what we...

Therefore, the medications that we have, the health interventions that we have, are or have been traditionally and historically orientated towards a particular group. So we're living with that already, one. And two. That's the fundamental reason why we have to engage because we need to correct that. And my eight years working on predictive analytics, there was always that black swan question. You know, why is that? What's that? What's that? Why is that?

working that way and unless we engage we can't correct it so i think we need to have a much more forceful voice because if if the world if we are the world is creating around us algorithms and results and whatever that is essentially biased. and biased away from emotional intelligence and all those things that we know about, those soft skills and those soft areas, then we will do everyone, including ourselves, a huge disservice. So we must engage.

for those reasons. Thank you, Julia. Now I've got two last questions. I'm going to keep these brief, but these are both still focused on ethics. one issue around data privacy you know third-party software developer has access to huge amounts of personal data how can we ensure this is kept safe i'm not drawing an equivalence here but just there was an instance recently in the news um and it was

of a child protection social worker in Victoria in Australia who'd been using ChatGPT to produce child protection court reports. Essentially they were inputting data into ChatGPT which is now owned by OpenAI or accessible to OpenAI. That was a very bad use of AI in practice. But coming back to that issue around personal data, how can we ensure that personal data is kept safe if it's been used in an AI dataset?

Happy to come in first, but I'm sure that Paul and Tommy may have really wise things to add to this. In the field of predictive analytics and children and families work, we have a lot to learn. We also have... mistakes that are being made. And we need to engage with those mistakes and learn from them, not say that's nothing to do with us or we don't like that. The predictions and the concerns about children and families, I think, is very real.

argue again that that is why we need to engage and talk about it and look for solutions to that, not say it's nothing to do with us or that we're too worried about. I've got two points for me. And it goes back to skills to a degree. But I think...

I looked at Social Work England's website, for example. There's nothing that says about artificial intelligence on their website. There's nothing in the standards that mentions artificial intelligence. There is definitely the role of the regulator for social workers in answering your question. I think as well, secondly, it would be great for Baswa to have a statement on the use of AI. And I know you've alluded to that, Julia, but we know, I think it's worth mentioning the Care Workers charity.

in conjunction with the Oxford Institute for AI, have a statement on ethical use for AI for care workers. I think it's time social workers had that as well to a degree. So those would be my two points to your question, Andy. That's absolutely right, and Baswa will, and I would be very keen. But all social workers need to be keen too, because we'll remember.

organization. So I wanted to mention Skills for Care because Skills for Care did the annual report yesterday and I attended it. It was about the workforce. There was absolutely nothing, a bit like on the regulators. Absolutely nothing about the development of... I don't mind about robots, but of AI and how that might be helpful to the workforce. And we really have to, all of us together, collaborate on making sure that we understand this future world and we are part of it.

okay yeah from my perspective and i think i suppose in some ways we're trying to put the human into this i do feel for this staff member who's now being used as the example of you know how maybe you shouldn't use AI within practice.

Although I would be interested to know a bit more of the details around that in terms of current pressures that you might have been under, but also just generally that discussion about is there a technology that exists that might actually... keep that bureaucracy down and cut it down and we hear that those strong messages coming you know pretty consistent over the last decade regarding the you know the paper trail and you know the lack of sort of service user face time

In terms of keeping yourself safe, I think it is important just to know what the privacy laws are with the jurisdiction you're in. Don't put sensitive data. on it so if you do want some support I would be tempted to to strip out all that personal sensitive data and maybe have you know the the ai if it is about report writing or whatever help help maybe structure or sentence that could be an option to do there do you really see that as a viable use paul do you think that's that's responsible

in terms of stripping out personal data. Yeah, so you write in court reports and you use ChatDBT. Is that not pretty dubious? Okay, well, let me put it back to you. What worries you in the sense of... So does the social worker and their registration, you have to stand over what you're saying. And if you're relying on a piece of software which is...

In terms of its ability at present, it's contested in terms of how reliable it can be, how accurate it can be. If you're pulling information from that, you've fundamentally got to be able to stand over that. And if you misunderstand the capacity of that...

information or that software as it currently stands. If you think it is, you know, unable to make mistakes, you could potentially be signing off on information, you know, reporting it to court, which is just wrong. And that's your registration online. Yeah, and I suppose, though, that, and I remember one of the things that jumped out for me playing around in the early stages of ChatGPT and asking to do various things for it was making up references and citations that didn't actually exist.

And that was a real kind of light bulb moment. But I think, though, to answer your question directly, that assumes, though, Andy, that person is going to strip out sensitive information, put it in a chat GPT, get a report, and then hit some bits. So in terms of keeping yourself safe, I think there's a number of things left to work kind of in tandem with each other. So stripping out sensitive data, using clear prompts, checking it for kind of bias.

using it as an enhancement, not a replacement tool, you know, be transparent about its use. I mean, we have AI statements now in academic journals as well as asking students to make that. educate regarding the limitations of AI is really important. I remember one of the first articles I published in the British Journal of Social Work was about social media and social work and it took me a while to get that published because I had to make this argument of why it was relevant for social work.

But within that, and with some accompanying articles, it was about young people using social media. And one of the big things coming through from that work was... If you don't teach people how to use it, they'll go and try and just use it on their own. And the only way you really learn is by making loads of mistakes.

you know, putting up your mobile number and where you live online. And I think maybe we need that type of attitude towards AI is we need to train it. Maybe that's where educators in universities need to start the process a bit sooner. I know, Tarsim, Synchuna in Birmingham does a lovely resource of YouTube videos using AI.

for students about around, you know, using it as a debating partner, preparing presentations and kind of, you know, it's just a lovely set of resources. I would definitely encourage listeners to check out. I'm not suggesting or advocating at this stage that social workers start using it for court reports. But if the question really was about how do you kind of mitigate some of the risks, then those are some of the steps I would take. We're not at the stage now that...

it's going to be used in that way as of the day, I don't think. Very brief final question, Paul, just because I mentioned there was two ethical considerations to finish up with. It's around informed consent. So if AI is playing a role in care planning, the service user...

Is it important that the service user is aware of that? You know, if the service user thinks a human is making decisions for their care but in fact those decisions are being either recommended, augmented or made by a piece of software. Is there an issue around informed consent? There's a new electronic data management system that you'd be aware of on Compass in Northern Ireland. And when I was chatting to some of the medical staff the other day, they were...

They were telling me about, you know, it will flag up, you know, say medications that, you know, you might want to avoid or if they're on this, don't give this and stuff. And it's kind of almost starting to feed in. So it was using those predictive tools. So it's an interesting one, isn't it? Between it probably, to some extent, technology is influencing people's career already. We just haven't called it AI.

in that sense. But as long as I think it's filtered through a human, it's probably the key point I would make there. Fascinating, Paul. Thanks so much. I have learned a lot today and I think we need to carry on learning. And you've given us some pointers on her. So thank you. And I certainly know that having spoken to the folks, as it were, at Beam who run Magic Notes, it's very clear when they are asking for consent that they make clear that...

the action notes are co-generated with AI, but the social worker will then curate that afterwards and is the author of for veracity as much as anything else. So I do know that that is very clearly in the consent template.

Thank you, Tommy. Could I come in with just one final point, which I thought was really interesting? Some of the projects we're really interested in is around mental health screening, as I said before. So what do you do in terms of if it's just left in the hands of an AI and it's monitored? How then, if risk comes into it, how do we trust AI to make appropriate decisions? Particularly if people are going to, you know, a serious harm, it could be a consequence of this.

Some of the software developers have talked about actually building in an additional AI to check the feedback given by the other AI. So you can see how complicated it can get between AIs looking after AIs. It's pretty meta, doesn't it?

Yeah, probably one of those other points around how this technology might be used that I didn't say in terms of protecting is actually not ignoring the role of industry and working with industry in this. Traditionally, we don't really view them as a partner. It's really maybe academic service providers.

provider's link with industry is there as well. And that is a way of bringing, I think, an additional level of protection. You know, if you have this sort of tech for good ethos within companies and sort of trusted partners can really help, I think, mitigate. um some some of the risks

Thank you, Paul. Julia, Paul, Tommy, thanks so much for your time. It's been absolutely fascinating. There's so much we haven't been able to discuss. This isn't an issue which is going to go away. We'll be making future episodes on AI and social work and it would be great to have you all back on again.

This podcast is hosted on a site called Acast. Acast required me to sign a box the other day that said none of the content on our podcast was AI generated, unless that's been acknowledged. So I can just confirm that Julia, Paul and Tommy are three very experienced, very real people. Thank you for your time. Thanks for coming on to Let's Talk Social Work.