Paul Murphy: Urges Ban on Algorithmic Targeting of Children
Paul Murphy warned that recommender algorithms act as vectors of harmful content and called for banning the targeting of children, arguing that screening is not sufficient and children must be fully safeguarded. He said recommender systems should be off by default or at minimum opt-in, and that social media can exist without engagement-driven algorithms.
Murphy argued recommender systems feed vulnerable young people with suicidal, anti-worth, pro-eating disorder and other harmful material - he said there is no safeguard at home or in schools when such content is pushed by algorithms and that banning targeting of children is necessary rather than mere filtering.
He supported the option to turn recommender systems off as a minimum and said having them turned off fully is the ideal. Murphy framed opt-in controls as a floor, not a solution, and pressed for systems that do not profile or target children by default.
Murphy highlighted that AI and algorithms can effectively out people without intent - AI does not have intent yet can produce harmful outcomes. He raised concerns about behavioural tracking and profiling that determine what will engage users and can expose or exploit vulnerable individuals.
He insisted it is possible to build social platforms without engagement-driven recommenders and argued current models prioritise profit and attention to sell ads over user safety. Murphy said a platform based on explicitly stated interests and opt-in features is feasible and preferable for children.
Representatives from CyberSafe Kids reinforced that engagement-designed systems, including AI chatbots, are engineered to keep users consuming and are especially insidious for children. They argued there is no safe or safer alternative while tracking and profiling are used and called for a model that centres explicit consent, safety checks and protection for minors.
Call to ban algorithmic targeting of children
Murphy argued recommender systems feed vulnerable young people with suicidal, anti-worth, pro-eating disorder and other harmful material - he said there is no safeguard at home or in schools when such content is pushed by algorithms and that banning targeting of children is necessary rather than mere filtering.
Opt-out and off by default
He supported the option to turn recommender systems off as a minimum and said having them turned off fully is the ideal. Murphy framed opt-in controls as a floor, not a solution, and pressed for systems that do not profile or target children by default.
AI outing and lack of intent
Murphy highlighted that AI and algorithms can effectively out people without intent - AI does not have intent yet can produce harmful outcomes. He raised concerns about behavioural tracking and profiling that determine what will engage users and can expose or exploit vulnerable individuals.
Possibility of algorithm-free social media
He insisted it is possible to build social platforms without engagement-driven recommenders and argued current models prioritise profit and attention to sell ads over user safety. Murphy said a platform based on explicitly stated interests and opt-in features is feasible and preferable for children.
CyberSafe Kids warnings and safety concerns
Representatives from CyberSafe Kids reinforced that engagement-designed systems, including AI chatbots, are engineered to keep users consuming and are especially insidious for children. They argued there is no safe or safer alternative while tracking and profiling are used and called for a model that centres explicit consent, safety checks and protection for minors.
We publish thousands of recordings to make Irish politics transparent and resistant to manipulation. Spotted an error? Report it — together we are building a reliable archive of Irish politics.
Other speeches
Tego samego dnia All speeches from this day →
Victor Boyhan
Victor Boyhan thanks staff and calls for debate on housing
Danny Healy-Rae
Danny Healy-Rae Blasts Energy Regulator and Gas Policy
Richard O'Donoghue
Richard O'Donoghue: Condemns ESB's €3.8m Charge on Solar Farm
Gerard P. Craughwell
Gerard P. Craughwell on Tricolour, SAR Pay and 2013 Pension
Richard O'Donoghue
Richard O'Donoghue presses Minister on Garda numbers in Limerick
Sharon Keogan
Sharon Keogan: Says mass migration is worsening housing crisis
Transcript
Thanks, Jim, and it's a good place to start. Thanks for all the presentations. Just I was struck in your presentation, Bernard, about the algorithms acting as vectors of harmful content, reinforcing and amplifying of content that is harmful and so on, and you make a call to pay more attention to recommender systems, and we spoke about this in the earlier session. Would you support the banning of targeting of children with recommender algorithms? Absolutely. I think it's really important when you're starting to look at particular types of information, when you start to feel that you're in a particular vulnerable place and you're starting to look at information and you're now overloaded with information, whether it's kind of suicidal, whether it's mental health, whether it's, you know, you're not worthy and you're looking at all this kind of anti-negative perpetrated information and there's no filter, there's no turn-off, it's in your room, it's in your home, it's not, you know, if it was on the street, there would be some type of safeguard and some type of protection, if it's in schools, but because it's online, there's no safeguard, there's no protection. So I would say that, you know, it's not only about filtering through, but it's actually banning and ensuring that children are absolutely safeguarded, protected, and that screening is not even sufficient enough. Thanks a lot. Yeah, and then also to belong to, you raise many of the same points about, you know, people being force-fed anti-LGBTQ, racist, far-right, pro-eating disorder content because it works for the tech giants and then you have the added issue in terms of AI and the algorithms effectively outing people without the intent of AI to do it in the sense that AI doesn't have intent. Similarly, is that something you've considered, the question of, you know, I think you raise people being able to opt out of recommender algorithms, but, I mean, should they not be off by default? Ideally, yes. Okay. I would say, I suppose, having the option to turn them on and off is kind of minimum viable option. Having them turned off fully is the ideal. Thanks a lot. And just, like, because it came up a little bit earlier, like, it's possible to have social media without these algorithms. It's possible for people to choose themselves what they want. We don't have to have this fed to people. I mean, they don't want to do it because this is the magic that enables them to keep people's attention, then to sell more ads, et cetera. But it's driven by profit. It isn't driven by providing a good social media experience for people who signed up to Facebook to keep in track with their school friends or today signing up for TikTok to follow things or whatever. Like, it is possible to have social media without being driven by these algorithms. I don't know, would CyberSafe Kids want to comment on the... Yeah, sorry. Yeah, I mean, they're designed for engagement entirely, as are AI chatbots. So, yeah, the idea is just to keep you on, keep you consuming. But it is very insidious. So what we would say is there isn't a safe or safer alternative for children. You need to just not use a system of tracking and profiling children and looking at their kind of behavioural signals and then determining what's going to engage them based on that. It should be entirely based on explicitly stated interests and opt-in. So I think the model is entirely inappropriate for children, and we're just rolling this stuff out without the checks and balances. You know, we're considering innovation. We're considering privacy to some degree. We're not considering safety. So the model is just wrong. Thanks a lot.