Paul Murphy warns AI search results could mislead citizens

Yeah, thanks a lot for the presentations. Just to start with citizens' information, I mean, firstly, to congratulate you on your website, you know, it's really... We use it all the time, like, you know, when people are coming to us looking for information, generally the information is, well, it's accurate and it's a bit more simplified than you might get it in the kind of original places. Basically, it's been translated into more readable English for people like me and the public. So it's a little bit concerning that what you're basically saying is people are going online and searching for a query about, let's say, something topical. I'm thinking, you might be in receipt of carers allowance and you're waking up today hearing the news that, oh, you may be subject to tax, and you go online and you search for this. And historically, presumably, what happened is that the first search result is citizen information and you click online to citizen information and you get the accurate information. Whereas now, you're saying that when you go through Google or whatever search engine, there is a decent chance that you'll get an AI, the first thing will be an AI answer, which will have taken potentially some information from citizen information, presumably accurate, potentially information from revenue, also accurate, but may have mangled it together with information that is inaccurate and people are, as a result of Google front-loading and pushing people and other search engines into AI, people may be getting less accurate information than they would have previously through search engines. Is that a fair summation? They're getting information from the search engine maybe for the first time. Whereas before, they would have got the links through and that click-through, we would have been able to track in terms of how are we appearing in search results, what types of things are people searching for, what are they clicking through for. What's really happening to us now is we don't have any real insight into that other than we've appeared in a search result. But we don't have a sense of what people are asking Google and then potentially getting results from us and not clicking through. So we have a bit of an information gap and what we also have, I guess, from the scale of the decrease is the sense that people are evidently satisfied with the AI overview. They're using AI. They want it as a method to interrogate information and that's part of the reason why we've pushed ahead with this project because we think that it's something that people find really useful and we wanted to try to limit the sources and see how would it work if it were just limited to our website. Yeah, but in a sense, it is a different project, the Ask CI versus Google because what you have done by kind of parceling off the AI into the citizen's information website is that you're guaranteeing it's only getting information from sources that are accurate. The problem is your average person who goes online to check about their potential tax implications of having carer's allowance, the first thing they do is go to Google, not citizen's information, in probably 90% plus of cases and there is an increasing danger that people are not being funneled to websites, be it any of your websites, which I trust all have accurate information and instead are getting kind of garbled AI stuff and people may be satisfied because you got the information but nobody tells you the information you got is inaccurate. You may only discover that later on when it turns out you have a tax bill that you didn't expect or I mean is this something that you as an organisation which is about providing people with information is concerned about, not so much in terms of your own AI usage but in terms of the wider information landscape around people's entitlements and so on? We are concerned about it. I think the zero search, zero click search, apologies, is probably a concern for most website owners at this point. I think it's fairly widespread among anyone who's running a public-facing website that traffic is not coming through to the same degree that it did before. Somebody mentioned earlier that we're not a commercial organisation so we're not trying to benefit from people moving through to our website but we do want to be able to track it and we do want to be able to stand over the accuracy of the information that we're providing to people and we're essentially taking what we can control which is our own content and trying to provide a new tool for people to interact with it and we hope that they find that useful. Where they don't, they can keep clicking through to the website or they can go to our phone service or they can go to one of our centres. Just related to that, an issue that came up for me in the general election actually on a number of occasions was people being complimentary about citizen information on the one side but on the other side saying we used to have a kind of local one in our community centre that no longer happens. I dug into a particular case and there have been the closure of some information centres for example in Wrath Mines, I think there have been others. You're saying to us that that process which presumably is linked to funding of a less, it'd be harder for people to go to their local centre again information has nothing to do with this process of going the AI route. I guess people will be concerned about this because I mean the kind of person who goes into a citizen information centre I presume it's either the issue is quite complex and they're kind of people who might come to us looking for assistance or there may be they may have challenges in terms of literacy you know what I mean they can't just go onto the website and accurately get the information from the website so similarly it seems to me the cohort who would be looking to go and speak to people in an office and the cohort who might benefit from an AI tool are, I'm not saying they're not identical but it seems to me they're quite dissimilar cohorts I think if you've got a complex or a nuanced query or a very personalised query you're not going to get an answer from Ask CI and you might spend quite a lot of time on the website trying to personalise it to yourself in which case you are going to need to call our phone service or drop into a centre to get somebody to look through all of the different aspects and as was mentioned earlier sometimes a query that starts in one place moves on to actually turn out to have several other facets to it and that's really the benefit that comes from visiting a centre I mean we absolutely support and promote the face-to-face service we don't see Ask CI as being anything remotely comparable to it it's really very important to go and see somebody who has the expertise and the experience to look at it in the round and to move from facet A to facet B and to realise that actually this is a more nuanced problem than even the person themselves might realise Thanks a lot just a quick question for revenue as I understand from the opening statement I as a customer contact revenue I don't interface directly with AI but there may be AI operating in the background to route me in a certain way can I choose to opt out can I say when I make my first contact I don't want anything to do with I don't want any AI to direct my query So Deputy the current way we use AI to classify your query is through our My Inquiry services which is available just through our Revenue Online service or My Account and allows you to raise a query electronically and then it is dealt with by a revenue officer all of those queries do get routed through AI purely to classify them to say is this something maybe in respect of PAYE and then we will send it to a person who has skills in that area or is it something to do with importing a car and we'll send it to someone in our customs division so all it does is make sure it gets to the correct revenue officer as quickly as possible there isn't an opt out on it And are there any plans to use it for any customer facing services in the future? So we have built a assistant that's trained on our information on our website we have 1400 tax and duty manuals which are guidance on all the Taxes Consolidation Act and obviously we have the primary legislation as well Internally we have deployed a version of that so it's trained on all of that information and it provides allows a user to do a question and answer type scenario We are testing it internally we would hope in the future to be able to release that externally but it's very interesting as you go through the testing process on tools like that when we tested it with our own revenue staff who all have good understanding of Tax and the way they would write their questions is very different to how a member of the public might and when you're testing ICT systems you're generally doing kind of binodal trinodal testing and you know your boundaries there is no boundaries in this so we're currently refining that model but we hope to release it publicly once we are satisfied with the guardrails around it so that there isn't incorrect information provided to a member of the public Thanks a lot Thanks to you