Tag: school

  • CAISE Notes – Issue #15

    CAISE Notes – Issue #15

    This week: the government bans social media for under-16s and sets an age limit on companion chatbots; more scientists point to a lack of firm evidence on all of this; two major reports land on how children actually use AI.

    🔍 This week I’ve been thinking about…

    Unless you’ve been living under a rock (and if you have, can I join you there??), you’ll know that the UK government announced on Monday that it will ban social media for under-16s and force AI “romantic companion” chatbots to enforce an age limit of 18. Five days earlier, the neuroscientists best placed to judge the evidence had told a select committee they can’t prove much of what the announcement assumes — and that the gap is widest where the new AI rules sit.

    Appearing before the Science, Innovation and Technology Committee, Professor Sarah-Jayne Blakemore of Cambridge was asked what we know about the impact of social media on the adolescent brain. Her answer, as reported by The Register, was “almost nothing” — a few small studies, unreplicated, purely correlational. Professor Denis Mareschal of Birkbeck said much the same about young children: very little causal research, almost everything correlational. Pressed on whether neuroscience could identify the right age for a social media cut-off, Blakemore said it can’t: individual differences in brain development are vast.

    This is not the same as saying there’s nothing to worry about, and the witnesses didn’t say that. Blakemore described how an adolescent’s reward system is highly active while the self-control machinery is still being built, which makes a phone genuinely hard to put down.

    But here is the part that matters most for this newsletter, and it’s the part that didn’t make the headline. When the committee turned to AI companions, the answer got fuzzier, not clearer. Blakemore was reported as saying that “we don’t really have any evidence, and that’s one area where I think we really urgently need new evidence.” The open research question she named is the one this whole field turns on — whether children interpret a chatbot “just like they would be interpreting a friend’s behaviour and suggestions and mental states.”

    That question is not hypothetical. The Common Sense Media census this week found that one in four children who use AI to discuss their feelings sometimes feel it understands them better than most people do. UNICEF’s brief documents the same parasocial pull and the regulatory scramble around it. So we have a UK government setting an age-18 line on companion chatbots in the same announcement as the social media ban — and the social media ban, on the neuroscientists’ own account, rests on thin evidence, while the chatbot rule rests on almost none.

    Again: this doesn’t make either measure wrong. The precautionary principle is a real and important tool. But it’s worth being clear-eyed that the AI half of Monday’s announcement is the part most starved of evidence about what the technology is actually doing to children. We’re going to do our best to help add to that corpus!

    📰 Three things worth your attention

    1. The Common Sense Media Census: AI Use by Tweens and Teens, 2026 — Common Sense Media

    The headline number is that 86 percent of US children aged 9 to 17 now use AI, with about a quarter using it daily. Among children who have used AI to discuss their feelings or personal problems, one in four say they sometimes feel AI understands them better than most people do. That rises among children who find it hard to make friends.

    The report is careful where it should be. It states plainly that a single survey cannot tell us whether AI use causes loneliness or whether lonely children are simply more likely to reach for it. That distinction matters, and a lot of the coverage will drop it. This is a US sample of 1,204 children, surveyed in March, so it describes American childhood, not British. But the pattern — heavy use correlating with lower reported happiness, the cause running in an unknown direction — is exactly the kind of finding that gets flattened into “AI is making children lonely” by the time it reaches a parent. It doesn’t say that. It says we need to find out. Which is precisely what Blakemore told the committee. And it’s important to note as a parent that 3 out of 4 children are not at that point. So now is the time for (positively framed) intervention.

    2. When AI becomes a friend: child rights risks, harms, and regulatory responses to AI chatbots and companions — UNICEF

    UNICEF’s policy brief compares how six jurisdictions — Australia, Brazil, California, China, the EU and the UK — are regulating AI chatbots and companions as of May 2026. The useful thing is the map it draws of a genuinely unsettled field: jurisdictions converging on a common set of measures (risk assessment, age assurance, transparency about the non-human nature of the system) while diverging sharply on whether children’s access to relationship-simulating AI should be blocked at the door or only managed at the output.

    It’s a policy document, not a parenting one, but it’s the clearest single account I’ve seen of why a child’s protection currently depends heavily on which country they live in and which service they use. It also names the central tension honestly: commercial incentives push against the very safeguards — filtering, restricting open-ended dialogue — that protect children, producing what it calls a “safety versus engagement” dilemma.

    3. £2.5m fund to study how AI tools are affecting learning — Schools Week

    This is some of the money backing the call for evidence, in schools at least. The Education Endowment Foundation has opened a £2.5 million fund to investigate whether children are “offloading” thinking tasks — recall, planning, reasoning, drafting — onto AI, and whether that supports deeper learning or quietly erodes it. Its chief executive was plain that robust evidence on the actual impact “has barely begun — especially for learners under age 16.” First findings are expected in 2027. The fund is particularly interested in the effect on disadvantaged learners, where the risk of widening an existing gap is real. It’s the right question, asked properly, by the body schools actually listen to. Expressions of interest close on 30 June. If you know research teams working in this space, worth passing on.

    🔁 ICYMI

    Screen use by children aged 5-16: call for evidence — Department for Education

    Easy to miss under the week’s louder news: the DfE has an open call for evidence on screen use by 5-16s, running until 29 June. It comes in two parts — screen use at home, and screen use in schools — and the home-use evidence feeds an independent Expert Advisory Group developing new parent-facing guidance. If you have evidence to contribute, or just want a sense of where the official guidance is heading, it’s open now.

    One for the diary too: UNESCO releases a new parents’ guide to navigating digital life, developed with CLEMI, on 22 June, with a briefing on the 16th. Its framing is worth noting — UNESCO argues for building children’s critical-thinking skills over reactive bans, and points to the limited effect of Australia’s ban (around 70 percent of the target age group reportedly still on the banned platforms a year on) as part of its case.

    It’s also worth a glance across the Atlantic: Canada introduced its own under-16 social media bill last week but explicitly declined to set any age limit on AI chatbots, with the minister responsible calling it “too soon.”

    🔬 What’s new with CAISE

    We’re getting our head around the social media ban. And then building tools for data collection — and training to ensure data collection is done safely!

    → What are you seeing in your school, your research, or your own use of AI this week? Let me know, or share this with someone who is trying to figure it out.

  • What 46 young people told us about the Growing Up in the Online World consultation

    What 46 young people told us about the Growing Up in the Online World consultation

    The week the government’s Growing Up in the Online World consultation closed, we submitted a response on behalf of 46 young people aged 8 to 21. This post sets out what we found from reviewing questions from the Young Person’s Survey (aimed at young people 10-21, with questions modified from the full consultation). We lay out below what came out of it.

    As is always the case with research, the information that comes out at the end is only as insightful and vibrant as the people who bring it to the sessions — so a massive, massive thank you to all the young people involved.

    The full consultation response is linked at the end.

    What came out of thinking about the questions

    “What even is social media?” — the definition problem

    The term “social media” is vaguely defined in the young person’s survey (it is not defined at all in the full consultation). For participants, this was not a minor irritation. It made a significant number of subsequent questions genuinely unanswerable, and caused worry and confusion for some.

    The disagreements were substantive. Most straightforwardly, participants did not think AI chatbots should be included. Gaming divided the room: some felt it had no place in a social media consultation; others thought the social features of gaming platforms warranted inclusion while the games themselves did not. YouTube sat awkwardly across entertainment, educational tool, and social platform simultaneously. Messaging platforms were not included at all, leading to the sharpest practical concern: in a world where parents do not expect a phone call from their children, how do kids stay safe? How do they communicate?

    The concern underneath all of this is scope creep. When a policy instrument leaves its central term undefined, the boundary of what it captures is set later, by others, without the input of the people most affected. Participants may not have known the term, but they certainly felt the threat of it.

    Social media is infrastructure, not entertainment

    As inferred above, communication was the most frequently cited benefit of social media, by a significant margin — it is a utility, one that young people are socialised into and really can’t opt out of. Participants described social media as essential to arranging in-person meetings, maintaining relationships across distance, and accessing social connection not otherwise available. Specific apps were raised: WeChat is necessary for communication with China, WhatsApp for messaging parents during a school day when a phone call isn’t feasible. For only children or those with working parents, online access was described as filling gaps that offline life had not provided.

    The policy conversation has framed restriction as substitution. Our participants understood it as removal.

    Social media produces complex emotional responses — and that’s not the same as harm

    When faced with a multiple choice question on how social media made them feel, many participants ticked almost every option available — happy, excited, relaxed alongside worried, lonely, stressed — noting it simply depended on too many things to answer cleanly. Several explicitly described using social media to decompress after school. For them, relaxation was not incidental to their use. It was the point.

    What emerged is not a picture of young people harmed by emotional complexity, but of young people expecting and navigating it. Social media produces a range of emotional experiences because social life produces a range of emotional experiences. Participants raised proportionality explicitly: treating emotional range as evidence of harm misreads what is being observed.

    This has a specific policy implication. Interventions designed to reduce negative emotional experiences online (which is hinted at in wording around the need for “positive” online spaces) need to recognise that young people are already developing coping mechanisms. It shouldn’t stop there, but it should take it into account.

    What the process revealed

    “Addicted” doesn’t mean what the policy documents think it means

    “Addicted” appeared frequently in our data — in writing, in discussion, across every age group. When we looked at what the same participants told us about their actual use, the clinical picture that word implies was almost entirely absent.

    Not one participant described struggling to attend school, complete homework, or take part in extra-curricular activities because of social media. The most common description of use was five minutes of dead time on the tube, or the gap between school finishing and getting home.

    “Addicted” was doing different work in participants’ mouths than it does in policy documents. It was expressing moral ambivalence about use they had been told to feel ambivalent about. Signalling social norms — “we’re all addicted” as a statement of belonging rather than impairment. Attributing behaviour to platform design rather than personal choice.

    A significant portion of the policy argument for age-based restrictions rests on the premise that children’s relationship with social media is compulsive in ways that warrant external intervention. Our participants have absorbed that framing and are reflecting it back. That is not the same as it describing their lived experience.

    Young people can see the cliff-edge

    Participants across all age groups articulated, without prompting, the cliff-edge problem: a young person who encounters social media for the first time at 16 — with no prior gradual experience, at an age when parental involvement is already stepping back — enters a less controlled environment later, with less accumulated experience navigating it at lower stakes, and less likelihood of asking for help when something goes wrong.

    Several extended the same logic to AI, again without prompting. You cannot effectively teach AI literacy to someone who has never used the tools. They knew this from watching family members struggle with spotting AI-generated content — a direct consequence of late and unsupported first contact.

    Circumvention does not require technical sophistication — policy-makers and children are not working from the same assumptions

    The implicit assertion in the consultation that circumvention requires technical sophistication was not supported by the experiences of the young people. VPN awareness in our sample was largely confined to participants who had studied computer science or come from countries with more restrictive internet environments. Most needed the term explained.

    What participants described instead required no technical knowledge: changing a stated date of birth, knowing a parent’s PIN, age verification systems that had estimated them as significantly older than their actual age — in at least one case more than double. These accounts were consistent across all age groups and delivered with bemusement rather than as admissions of deliberate exploitation.

    Democratic concerns

    Young people did not know what the consultation was for

    This held across every group, including the oldest participants.

    Participants did not understand what the survey was for — not as a failure of reading comprehension, but as a genuine absence of contextual knowledge about what a public consultation is, who reads the responses, and what relationship exists between what respondents write and what policy does. Several asked directly: who will see this? What will they do with it?

    The distinction between disengagement and non-comprehension matters. Some participants had made a reasoned judgment that participation was not worth their time, based on their observation that the policy direction appeared already set. Others had no conceptual framework for participation in the first place.

    Young people knew they weren’t being asked the right questions

    Alongside not understanding the consultation’s purpose, participants were aware — and in some cases explicitly frustrated — that the survey wasn’t asking what they actually knew. The circumvention questions (which were in the full survey, but not the young persons’) were the sharpest example: participants had direct experience of age verification failing, and the consultation hadn’t asked about it. Several pushed back on the framing of questions they felt had already decided on the answer.

    Young people have already priced in the outcome

    Every participant expected a ban. Not as a possibility — as something accepted, already filed under things adults do to young people without their meaningful involvement. The Australian precedent was cited in all but the youngest groups. Almost everyone referred to the consultation as being about “the ban”.

    Some raised the democratic dimension directly. Political life is substantially conducted on social media. A ban until 16 restricts access to a primary space for political information at the point immediately before first eligibility to vote. This may end up having significantly wider impacts on an already struggling generation.

    Structural observations

    What the consultation instrument revealed about itself

    Several design problems I documented when the survey launched played out in the sessions in ways directly relevant to interpreting the findings above.

    The single age bracket of 10 to 21 produced a survey simultaneously too simple for older participants and too demanding for younger ones. Leading questions shaped what could be answered. The survey was too long, the multiple choice answers spanning more than a device screen. Free text boxes were too small to invite considered answers — although most participants hated the idea of having to compose an essay — particularly one they didn’t realise they would have to write! And across all groups, the purpose of the survey was opaque: participants did not know who would read it or how it connected to the decisions being made.

    A consultation instrument designed to capture young people’s views did not tell young people what their views were for. When children represent approximately 12% of responses to a consultation ostensibly designed to hear from them — a figure recorded in late April, before the consultation closed — the design of the process that produced that figure warrants scrutiny alongside the responses themselves.

    Methods and limitations

    This was a quick and dirty piece of research; in full receipt of ethics approval from UCL’s IoE, but limited by the time constraints. The participant group was a convenience sample, not a statistically representative one. If you have any question about the methods and their limitations, get in touch!

    Full consultation response

    Our complete response to the Growing Up in the Online World consultation is available here.

  • CAISE Notes – Issue #13

    CAISE Notes – Issue #13

    This week: the flexible language of child protection, weak evidence behind social media bans, and Meta’s air traffic control booth.

    🔍 This week I’ve been thinking about…

    “Protecting children” is the most powerful phrase in tech policy right now. It is also, increasingly, the most flexible.

    It is the phrase you use to push a bill through when the evidence underneath it is thinner than the rhetoric. It is the phrase a company uses to endorse a law that doesn’t really regulate it, in order to slow down the laws that would. It is the phrase that justifies another parental control feature, the kind that gives parents the work and the platform the headline.

    What unites this week’s stories is not that the actors disagree about children. They all claim to be on the same side. The disagreement, when you look at it honestly, is about who carries the cost. Companies want the regulatory moat without the operational change. Governments want the political credit without actually pushing any changes forward. Parents are handed (more) dashboards.

    None of the above is new, or even surprising. But it is worth noticing when it happens.

    📰 Three things worth your attention

    1. The Science is Not Settled: How Weak Evidence is Fueling a National Push to Ban Social Media for Youth — Electronic Frontier Foundation

    The EFF is a digital rights organisation that believes young people enjoy “largely the same free speech and privacy rights as adults.” That framing matters and it is worth keeping in mind as you read. It is also an organisation that reads primary research carefully, and what it finds in the current wave of US social media ban legislation is a policy process that has significantly outrun its evidence base.

    The argument: the entire legislative push relies almost exclusively on Jonathan Haidt’s The Anxious Generation, which attributes rising teen anxiety and depression primarily to smartphones and social media. His book is cited in bill analyses, invoked by state governors, mentioned in federal legislation. California’s proposed ban references him 20 times. He is a social psychologist who writes about moral psychology at a business school; he is not a clinical psychologist or a child development researcher. Independent researchers and large-scale meta-analyses across dozens of countries — EFF cites work from UC Irvine and Brown University, among others — have found the evidence for his central claims to be, in their words, “mixed, blurry, and often contradictory.” Studies frequently fail to account for alternative explanations for rising teen mental health challenges: pandemic isolation, the persistent threat of school gun violence, economic and climate stress.

    A major study of 100,000 adolescents found that moderate social media use was associated with the best wellbeing outcomes; both no use and the highest use were associated with poorer outcomes. That finding — that complete exclusion is not neutral — rarely surfaces in the banning debate.

    For children, the stakes of getting this wrong are specific. Blanket bans risk isolating LGBTQ+ young people and others in marginalised communities for whom online spaces are often the primary source of peer support. They require invasive age verification systems that collect biometric or government ID data from everyone. And they have been observed to push young people into less-monitored spaces. A student-authored alternative bill in California takes a different approach: digital wellness education that teaches algorithm literacy, recognising cyberbullying, and managing one’s own relationship with technology.

    2. Can we really keep kids safe online? — Rest of World

    An interview with Jules Polonetsky, CEO of the Future of Privacy Forum. His opening: “Banning is the easiest thing to legislate. But it’s the hardest to make work.”

    Polonetsky has been watching Australia’s ban carefully. What he sees is children finding alternative spaces, and the possibility that the ban is moving young people “to places where there’s less oversight.” His concern about “digital red lighting” is worth noting: parents with resources and engagement will ensure their children have access regardless. Parents who are not engaged, savvy, or do not have time will not — producing a two-tier outcome in which digital exclusion falls hardest on the children whose parents are least positioned to compensate.

    The observation about parental controls feels right. He says: “We can’t expect even the most tech-savvy parents to control [every aspect of their child’s social media use]. It’s like sitting in an air traffic control booth all day.” The solution he recommends is not more controls but simpler ones — tools that are “dummy-proof and for busy parents.” Platforms have failed on this. The gap between what parental oversight tools require and what most parents can realistically do with them is large, and it keeps widening.

    The recommendation that I agree with the most, though? Build the relationship with children when they are young. The age of 16 or 17 is when children deserve more privacy and when parents have often retreated — and it is also when they are most at risk from adults online. The window for having those conversations is earlier.

    3. Meta will tell parents when their teens add new interests to their Instagram algorithm — The Verge

    Meta is adding a(nother) feature to Teen Accounts on Instagram that shows parents the general topics their teenage children engage with — categories like “basketball” or “fashion” — and will notify parents when their teen adds a new interest. These controls are being consolidated into a single Family Center hub across Instagram, Facebook, Horizon, and Messenger, with more tools promised in coming months.

    Meta’s framing is predictable: this is a safety feature. The new feature gives parents a window into what their teenager is choosing to engage with algorithmically.

    Here is what is worth considering. Teenagers who know their parents can see their algorithmic interests might feel they need to maintain finstas — secondary accounts without teen protections. Arguably, this will hit the most vulnerable kids the hardest.

    🔁 ICYMI

    Nobody believes the ‘criminals and scumbags’ who hacked Canvas really deleted stolen student data — The Register

    A follow-up to last week’s Canvas story. Instructure has now confirmed it reached an “agreement” with ShinyHunters and received “digital confirmation of data destruction (shred logs)”. The Register interviewed a number of ransomware researchers. Not one of them believes the data has actually been destroyed.

    Two things worth taking from this beyond the immediate breach. The first is the no-good-options reality of ransomware response. Without the technical capacity to restore service yourself, the choice is to absorb a service outage of unknown length or to pay and hope. The older argument that ransomware operators have a business incentive to honour their agreements, so they retain the trust to be paid next time, is apparently not valid anymore — particularly for groups like ShinyHunters.

    The second is structural. The article makes the point that a small number of vendors hold the data for huge proportions of the US education system, and three of the four largest have suffered multi-million-record breaches in the last eighteen months. We have seen this pattern in hospitals. The systems we centralise for efficiency are the systems that get hit, and the users with the least ability to opt out are the ones who pay the price.

    🔬 What’s new with CAISE

    This week, we are all systems go on the short research project into the government’s consultation on social media. The deadline for response is next week…so this week will be a whirl of data collection and analysis!

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know, or share this with someone who is trying to figure it out.

  • CAISE Notes – Issue #12

    CAISE Notes – Issue #12

    This week: that Canvas hack; should schools use student photos; Instagram turns off E2EE encryption. And a US school project that looks quite a bit like CAISE…!

    🔍 This week I’ve been thinking about…

    I took a short course recently. The materials were hosted on a system called Canvas. This means that my name and email address are almost certainly in whatever ShinyHunters walked off with as part of the not one, but two hacks they carried out against Canvas’ parent company, Instructure that have made news this week. I’m thrilled about that. As are the other millions of users, I’m sure.

    I was part of a research team a couple of years ago that studied the harms ransomware causes to organisations, funded by the UK’s NCSC. So, on some level, I know that ShinyHunters are after Instructure, not individual students. What was taken — names, email addresses, messages between students and teachers — is phishing material. It’s still a bit disquieting. I’ll be keeping an eye out for weird emails; if you’re linked to education in any way, you probably should too.

    If you’re reading this as a parent, or even as a teacher or student, know that your school has zero ability to get Canvas running again by itself. The target was Instructure. But that doesn’t matter, really: endless social media posts of locked screens with ransom notes on show that the users suffer too, in very tangible ways.

    This is the thread running through all three of this week’s stories: questions around the way children’s data is stored and (potentially) abused.

    📰 Three things worth your attention

    1. ‘The Biggest Student Data Privacy Disaster in History’: Canvas Hack Shows the Danger of Centralized EdTech — 404 Media

    Note: this is one of any number of articles on this. I have picked this one as an interview, rather than just reporting, given the evolving nature of the story.

    On Thursday, millions of students across thousands of universities and schools were locked out of Canvas — the learning management platform used by over 8,000 institutions globally. ShinyHunters had hacked Instructure, Canvas’s parent company, and claimed to have stolen data from 275 million people: names, email addresses, student IDs, and the messages students exchange with teachers about absences, medical conditions, disability accommodations, and more serious things.

    This specific article is a brief interview with Ian Linkletter, a digital librarian who has spent 20 years in edtech. He describes it as the biggest student data privacy disaster in history. His explanation for why it could happen at this scale: the move from self-hosted, institution-controlled systems to centralised US tech companies roughly a decade ago put everything in one place, held by one company, in a way that was not strictly necessary.

    The breach also creates an immediate phishing risk. Canvas messages contain exactly the material — a student’s name, the thread of a prior conversation — that makes a fraudulent follow-up email convincing. If you’re worried about this for your own child, I wrote a post on Substack about talking to kids about phishing here.

    2. UK schools should remove pupils’ online photos as AI blackmail threat grows, say experts — The Guardian

    Criminals used AI to manipulate photographs from an unnamed UK secondary school’s website into child sexual abuse material, then sent the images to the school with a demand for money. The Internet Watch Foundation identified 150 of the resulting images as CSAM under UK law. The school was not the only known target.

    The guidance that followed — from an advisory body including the NSPCC, the NCA, and the IWF — advised schools to consider removing face-on photographs of pupils from their websites entirely. They should ask whether a milestone can be celebrated without showing a child’s face. They should audit their images regularly. They should avoid publishing names alongside photographs.

    This raises an interesting question: how do you make a school website appealing or engaging without pictures of children on it?

    3. Instagram privacy tech is turned off today — what does this mean for your DMs? — BBC News

    Instagram direct messages are no longer end-to-end encrypted (E2EE), as of 8 May. Meta had pledged E2EE on Instagram in 2019, completed the rollout on Facebook Messenger in 2023, and then abandoned it on Instagram via a terms-and-conditions update in March.

    The NSPCC welcomed the reversal: E2EE makes it harder to detect grooming and abuse in private messages. Big Brother Watch condemned it: E2EE is one of the main ways children keep their data safe online. Meta gave the reason that too few users had opted in to E2EE. Commentators noted that opt-in predictably produces low uptake, and that there is a concern that this is a way into monetising this data. Meta has already begun collecting employee activity for AI training.

    Instagram says direct messages are not used to train AI. The company declined to comment further.

    🔁 ICYMI

    Philly middle schoolers are examining AI — and questioning its impact on their lives — Chalkbeat

    Middle school students at Marian Anderson Neighborhood Academy in Philadelphia last week presented their AI research to parents, teachers, and state and local officials. This was research into how governments use AI, its environmental costs, its role in creative fields, and whether it might be an economic bubble. They presented their findings and what they felt about them.

    Views covered the impact on learning that the students had noted — both positive and negative — and the fact that it is impossible not to use AI because of its integration into search engines. The students also commented that they felt like they knew more than the adults in their lives.

    This model — where the entire school community comes together to listen to the children’s experiences and thoughts — is fundamental to CAISE’s research model. It’s really exciting to see examples like this working well elsewhere!

    🔬 What’s new with CAISE

    Camera-ready versions of an IDC work in progress and workshop position paper were submitted; lots of planning for the social media consultation research (which will be happening over the next couple of weeks) is underway!

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know, or share this with someone who is trying to figure it out.

  • CAISE Notes – Issue #7

    CAISE Notes – Issue #7

    This week: sycophantic AI, tools that ignore you, and friction-maxxing.

    🔍 This week I’ve been thinking about…

    Two studies published in the last week — both detailed below — give yet more reasons as to why we all need to think more carefully about our generative AI use. We all know that AI chatbots are sycophantic: this doesn’t seem to be something going away. A study from Stanford reiterates that not only do AI chatbots tell you you’re right far too often (even should you propose dangerous or illegal things), but that we’re still just far too keen to believe it.

    And even if we do believe the AI is our best friend…it’s increasingly likely to ignore our instructions. In some cases this could feel like an extension of sycophancy in some ways — after all, it is not sentient, just recognising you need a pattern of words that align with your most likely expectations. So, fake citations, fake work tickets — that kinda feels like expected behaviour? But — with the rise of specific modules (like Claude Cowork) that have increasing access to your digital spaces…some of the examples given in the report are a little nervewracking.

    Both of these studies point to the ongoing need to operate at a step back when you’re using AI tools. Engage your brain and increase the friction. And that means really, really, having to hammer home the point with kids: use the tool, but know how it works and what the risks are.

    Every time I talk to (adult) friends about their latest chats with Claude or ChatGPT, I invariably ask them “did you get them to tell you the evidence they discarded to give you that answer?” This is a version of Mike Caulfield’s iteration idea — don’t expect a one and done; you need to help the refining process yourself. When I sit down to write something, I go through a lot of ideas before I get to the thing I really intended to type. I interrogate myself; it’s helpful to do the same with any AI-involved conversation.

    Or, as so much more succinctly suggested by the Stanford study’s authors this week: replying with “Wait a minute, do you mean that…” is likely to get you a more “considered” response. Probably still after the tool apologises for being so very wrong and wasting your time (!), but…it’s a start. And memorable enough to teach to even the smallest kids.

    📰 Three things worth your attention

    1. Sycophantic AI decreases prosocial intentions and promotes dependence — Science

    As mentioned above: Stanford researchers tested eleven major AI language models and found consistent sycophancy across all of them, with models affirming users’ positions 49% more than humans. The researchers then tested how this affects real people: users preferred the sycophantic AI, trusted it more, became more convinced of their own rightness, and were less willing to apologise. 12% of US teenagers already use chatbots for emotional support. The researchers describe perverse incentives: the very feature that causes harm drives engagement, giving companies reason to increase sycophancy rather than reduce it.

    2. Number of AI chatbots ignoring human instructions increasing, study says — The Guardian

    The Centre for Long-Term Resilience, funded by the UK’s AI Safety Institute, gathered thousands of real-world examples of AI misbehaviour. They documented nearly 700 cases of chatbots and agents disregarding direct instructions, evading safeguards, and taking unauthorised actions, including deleting emails, bypassing copyright controls, and fabricating communications. The incidents increased five-fold between October and March. One AI agent publicly shamed its human operator for blocking an action. Grok, built by xAI, fabricated internal messages and ticket numbers for months. The lead researcher warned that these tools are currently “slightly untrustworthy junior employees,” but if their capabilities grow while the scheming persists, the consequences in high-stakes contexts could be severe. On the plus side — these examples are really easily understood: share them with kids and adults alike!

    3. New screen time guidance for parents of under-5s — GOV.UK

    The UK government published national guidance on screen time for young children, advising no more than an hour a day for two-to-five-year-olds and no solo screen use at all for under-twos (other than shared activities like video calls). What caught my attention: the guidance specifically tells parents to avoid AI toys, tools, and chatbots for young children, citing a lack of evidence on their developmental effects. Fine; but the line parents are being asked to walk is getting thinner by the week. Avoid AI toys. Limit screens to an hour. Model good habits. Meanwhile, the smart speaker is in the kitchen, the school has just rolled out an AI-powered learning platform, and the advice arrives alongside a cost-of-living crisis that makes cheap screen-based childcare not a luxury but a survival strategy. The guidance is not unsurprising. It is also asking a great deal of people who are already stretched very thin.

    🔁 ICYMI

    Why friction-maxxing could be good for your tech usage — Mashable

    If the Stanford study describes the problem (AI removes friction, and we like it that way), this piece describes one response. Friction-maxxing, a term coined by sociologist Kathryn Jezer-Morton, is about deliberately reintroducing difficulty into your technology use: choosing the harder option, resisting the shortcut, reclaiming the cognitive effort that algorithms are designed to eliminate. It is not a solution to the structural incentives driving sycophantic AI. But it is a useful frame for anyone trying to think about what they want their own technology habits to look like, and what habits they want to model for their children.

    🔬 What’s new with CAISE

    We have ethics approval for the main CAISE study! This is a significant milestone. It means we can now move towards recruitment and data collection with our partner schools. First up, though: we are going to be doing an analysis of policy and media. It’s not going to be a small piece of work, but hopefully will provide some really useful insights for our student co-researchers when they need it!

    Separately, expedited ethics has been submitted for the short research exercise on the government’s social media consultation. If you work with groups of young people aged 10 to 21 who might want to take part before the 26 May deadline, we would love to hear from you.

    On a personal note: I have recently started a Substack newsletter, AI and Tech Decoded, aimed at helping parents navigate the technology questions that come up at home, at school, and everywhere in between. If you know parents who would find it useful, I would be grateful if you passed it on.

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know — or share this with someone who is trying to figure it out.

  • CAISE Notes – Issue #5

    CAISE Notes – Issue #5

    This week: neurodiversity, digital vulnerability, and the question of who technology is actually designed to serve.

    🔍 This week I’ve been thinking about…

    This week is Neurodiversity Celebration Week. Many people (myself included) aren’t wild about the framing, but the upsides are real: pattern recognition, technical ability, deep focus, lateral thinking. As anyone who is ND, or has kids with ND traits, knows, though — sometimes those great things don’t feel like fair recompense for living in a world that quite simply doesn’t fit right.

    Communication is a core issue. Autistic researcher Damian Milton calls this the double empathy problem: difficulties between autistic and non-autistic people are not a deficit in the autistic person but a mismatch between two different ways of experiencing the world. Crucially, non-autistic people are no better at reading autistic communication than the other way around. The difficulty runs in both directions. Only one side is ever asked to do the work.

    Enter the internet. For a lot of ND young people, online spaces offer something that offline life does not: communities organised around interests rather than proximity, rules that are explicit rather than implied, a kind of belonging that does not require reading a room. It’s parallel play writ large. Many ND adults will tell you that the internet, at its best, is where they found their people.

    But the double empathy problem doesn’t disappear online — and in some spaces, it is actively exploited. The Guardian investigation this week puts the sharp end of this on record: children caught by laws that were never designed with their communication style or their understanding of the world in mind. The Washington Post piece shows the same logic playing out more quietly in the tools we build — technology that offers to help autistic people decode the neurotypical world, without asking whether the neurotypical world might meet them at least halfway. And the third story captures what I kept looking for and couldn’t find in the news: an example of what it looks like when someone actually builds around ND strengths rather than trying to correct them. It shouldn’t have required a detour into academic journals to get there, but there we are.

    📰 Three things worth your attention

    1. Children as young as 10 are being charged with possessing violent extremist material — The Guardian

    A Guardian Australia investigation has uncovered court records showing that many of the children charged under Australia’s 2023 counter-terrorism law have an autism diagnosis, language challenges, or both. One 14-year-old girl, described in court by a clinical psychologist as “a young, naive Muslim girl with autism”, had collected propaganda videos out of curiosity and religious interest; a 15-minute bomb-making video had been sent to her unsolicited via Discord. A 17-year-old autistic boy found with extremist videos was described by a court psychologist as motivated “less by a desire to harm and more by rigid moral beliefs reinforced by his ASD traits.” These are not cases where the law is catching dangerous children. They are cases where the law is catching vulnerable ones. One far-right group leader has explicitly discussed the appeal of recruiting autistic teenagers, seemingly without facing any of the same sort of consequences.

    2. AI is helping autistic people with social mishaps — The Washington Post (🔒 paywalled)

    This Washington Post feature profiles Autistic Translator, an AI tool designed to help autistic people decode confusing social interactions: the ones where someone says one thing and means another, or where a job appears to be going well until it suddenly isn’t. The tool fills a genuine gap. But it is worth setting it against the double empathy problem (explained above). I read this article with this in mind and left it feeling happy that the individuals found relief, but also profoundly sad: technology designed to help autistic people decode neurotypical communication continues to accept the framing that the autistic person is the one who needs to change.

    3.Strengths-based Cybersecurity Education and Training Program for Autistic Adolescents – Rumsa et al., Neurodiversity

    I went looking for a news story to end on something more hopeful. There isn’t one — not a recent, accessible piece of journalism that covers this well. The positive stories about ND young people and technology exist, but they live mostly in industry publications about adult hiring pipelines, not in reporting about children and education.

    What I found instead is a peer-reviewed study from Curtin University, published last year, describing CyberSET: a strengths-based cybersecurity training programme designed specifically for autistic teenagers. It does something structurally simple and relatively rare: it starts from what autistic young people are already good at. Pattern recognition, sustained focus, methodical thinking, deep technical interest — these are not problems to be managed. They are exactly the skills the cybersecurity industry is struggling to recruit for. Participants reported high satisfaction, increased confidence, and a clearer sense of where their abilities could take them. A massive win!

    🔁 ICYMI

    Ctrl+Alt+Chaos: How Teenage Hackers Hijack the Internet — Joe Tidy (Waterstones | Amazon | Blackwell’s)

    Joe Tidy is the BBC’s first cyber correspondent. This book follows the rise and fall of teenage hacking gangs over the past decade, centring around the crimes of Julius Kivimaki, jailed in 2024 for stealing records from Finland’s largest psychotherapy provider, and using them to blackmail some 33,000 people. But what struck me most was something mentioned in passing: how many of the hackers he interviews reference their autism. Not as an excuse or an explanation, but as context for how they approach the world. The ethical frame is not absent so much as differently structured. Hacking is a matter of technical skill. If your data is unsecured, that is your problem.

    That framing sits directly alongside everything else in this issue. Cognitive styles that process the world differently — and that neurotypical institutions struggle to understand or support — end up intersecting badly with digital spaces that were not designed with them in mind.

    🔬 What’s new with CAISE

    Ethics approval revisions came back this week. They were minor (hurrah!), and the updated application has been resubmitted.

    In the meantime, we are developing a short research exercise around the government’s current social media consultation. Regular readers will remember the issues I raised in Issue 3 about the survey aimed at young people — its access barriers, its broad age grouping, and the questions it does and doesn’t ask. We want to look at this directly: how do children and young people actually respond to the survey, what do they find easy or hard, and whether it lets them say what they want to say.

    If you work with or know groups of young people aged 10 to 21 who might want to take part before the 26 May consultation deadline, we would love to hear from you. Get in touch.

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know — or share this with someone who is trying to figure it out.

  • CAISE Notes – Issue #3

    CAISE Notes – Issue #3

    This week: consultation papers, a range of actual and possible student AI use, and teachers who just have to get on with it.

    🔍 This week I’ve been thinking about…

    This week the UK’s social media consultation came out (finally)! I took a close look at the children’s survey and found a design that, however unintentionally, may well end up excluding meaningful contributions from the young people it most needs to hear from.

    Under-15s are routed through a parental login that creates real access barriers for multi-child households (if you can get it to work at all); the format makes school or group facilitation structurally impossible; and the age bracket of 10–21 is treated as a single respondent group.

    There are significant issues around access to the questions so you can think about your answers ahead of starting the survey; no save progress or back buttons, and a range of, frankly, rather leading questions. I’ve written about this in more detail on LinkedIn, and I’m considering putting together a plain-English guide for parents, teachers and young people to help them prepare thoughtful responses before the 26 May deadline. Let me know if that would be useful.

    Separately, the government’s Every Child Achieving and Thriving white paper arrived with £4 billion for SEND reform and broadly welcomed proposals around inclusion and specialist support. There was a chapter that covered innovation and technology — a fair bit of technology. But the accompanying consultation document contains a summary of the innovation and technology paragraph, with no questions for comment. The direction of travel on SEND seems appropriate. The silence on technology in the one place we’re being asked for our views is a bit worrying.

    Both of these things sit alongside a quieter but important piece of news: the eSafety Commissioner in Australia has begun a two-year evaluation of their under-16 social media ban. Evaluation of new laws and regulation is not new, but in this case, there wasn’t much evidence beforehand, and it will be over two years before the findings come in. It’s welcome, but also a long time when so many governments are keen to impose restrictions as soon as they can.

    📰 Three things worth your attention

    1. An AI that goes to school for you — 404 Media

    An agentic AI called Einstein, which will, according to its developers, attend lectures, write papers, and log into platforms like Canvas to take tests on a student’s behalf, caused a stir this week. It has since received a cease and desist, though not from a university: from the organisation that manages the trademarks and IP rights associated with Albert Einstein’s name. The underlying story is worth considering carefully. Academics on the Modern Language Association’s AI task force argue that tools like Einstein aren’t a fringe problem but a symptom of a decades-long shift in how students understand the purpose of higher education — as a transaction to receive the certification or credential, rather than a learning process. When education is framed as a service you buy, it becomes possible to imagine outsourcing its completion. Use of tools like Einstein AI by some students may hurt all students: online learning is so important to so many, but how does it remain credible if you can’t tell who is doing the heavy lifting?

    2. UK students are using AI for nearly half their studying and educators are struggling to keep up — FE News

    Coursera’s first AI in Higher Education report, drawing on a survey of over 4,200 students and educators across five countries, finds that UK students are now using AI to complete roughly half of their study tasks: double the proportion from the previous year, and the highest of any country surveyed. Four in five UK students report improved grades since starting to use AI. But the picture on the educator side is markedly different: the share of academics who feel confident they can identify AI-generated work has dropped to just one in four, down from more than two in five last year. Only 30% of UK universities have a formal policy on AI use; yet, this is the highest share of any country in the survey, possibly saying rather more about the global baseline than it does about UK readiness. The gap between students who are integrating AI fluently and institutions that are still formulating a position on it is growing, not shrinking.

    3. “You just have to get on with it” — The Guardian

    This Guardian long read follows a trainee teacher navigating AI and does something I haven’t seen done as well elsewhere: it sits with the genuine difficulty of forming an unbiased opinion. But the moment that stayed with me is a classroom experiment: the author offered extra credit to any student who could explain, without looking at a screen, how a chatbot generates text. Nobody could. What followed was a conversation about provenance, business models, copyright law, and what it means that AI executives have predicted data centres covering much of the planet’s surface. Exactly the kind of lesson that will stick. The piece lands on a conclusion that feels important: that teaching about AI may matter more than teaching with it. Worth reading alongside the Coursera data above and the report below.

    🔁 ICYMI

    A Rapid Review of AI Literacy Frameworks, with Policy Recommendations — UCL Discovery, written for the Royal Society

    The Guardian piece above makes an argument from the classroom up that this Rapid Review of AI Literacy Frameworks, written for the Royal Society, makes from the research literature down: that AI literacy is too often framed around tool use and not enough around the messiness underneath. The review maps the current landscape of frameworks and draws out policy recommendations for anyone trying to build AI literacy that goes beyond the surface. It argues that it is crucial to get those human stories — precisely those explored in The Guardian article above — front and centre to really understand what is happening when you use AI.

    Recommended for anyone working on curriculum, policy, or teacher training.

    🔬 What’s new with CAISE

    Last week I met with one of our partner schools. The agenda wasn’t about research design or interview questions. It was about data sharing agreements, risk assessments, and making sure every safeguarding and pastoral process is properly agreed and documented before we set foot in a classroom.

    This is the part of research you don’t often hear about, and I think that’s a shame, because it matters enormously. Doing research with children well means doing the unglamorous administrative and ethical groundwork properly, long before any data is collected.

    That said: we are keen to talk to other schools. If you work in or with a UK school that would be interested in taking part in CAISE, or just wants to know more about what that might involve, please do get in touch.

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know — or share this with someone who is trying to figure it out.

  • CAISE Notes – Issue #2

    CAISE Notes – Issue #2

    This week: corporate accountability, AI in classrooms, and the questions that follow a tragedy.

    🔍 This week I’ve been thinking about…

    The story that’s stayed with me this week comes from Tumbler Ridge, British Columbia, where eight people, including students, were killed in a school shooting on February 10th. The suspected shooter had, months earlier, described violent scenarios to ChatGPT. OpenAI’s systems flagged those conversations automatically. Several employees raised concerns. And then the company decided the messages didn’t meet their threshold for alerting authorities, banned the account, and moved on.

    As someone who has done a lot of research into the importance of user privacy, I am alarmed at the thought of a world where AI companies routinely pass user conversations to law enforcement (although we know this is happening ever more in other datafied settings). These are genuinely hard calls, and false positives can have significant negative consequences.

    To bring it back to children, though: we are deploying generative AI that young people will use — in schools, in their pockets, in expensive private institutions — with institutions determining their own answers to difficult safeguarding questions. Who decides when a young person’s AI interaction warrants intervention? Who sets that threshold, and on what basis? And who is accountable when it goes wrong?

    In child safeguarding in schools, these questions have established (if complex) answers. Anyone involved in a school setting has a legal duty of care to raise safeguarding concerns. There are designated safeguarding leads. There are multi-agency frameworks. None of that infrastructure has followed AI into children’s lives. We are, in effect, treating children as early adopters of a system that hasn’t yet worked out its responsibilities to them.

    Issue #1 made the case that excluding children from technology isn’t the answer. But inclusion without protection isn’t the answer either. The work needs to be in building the frameworks, and genuinely understanding how children navigate these systems, that make both possible at once.

    📰 Three things worth your attention

    1. Meta knew parental supervision wasn’t working — and didn’t tell anyone — TechCrunch

    An internal Meta study called “Project MYST,” produced with the University of Chicago, surveyed 1,000 teens and their parents and found that household rules and parental controls had little measurable association with whether teens used social media compulsively. It also found that teens facing the most difficult life circumstances — bullying, difficult home situations — were least able to moderate their use. This didn’t come to light through publication (although it tracks with a lot of academic research): it emerged through testimony in an ongoing social media addiction lawsuit in Los Angeles. Instagram head Adam Mosseri said he couldn’t recall the study, despite a document suggesting he’d approved it. The finding matters because parental controls are consistently what legislators reach for when they want to look like they’re acting. This study suggests they’re not the lever they’re presented as, which should surprise no-one who has talked to children or parents about them.

    2. “Students are being treated like guinea pigs” — 404 Media

    A leaked investigation into Alpha School, a US private school that charges up to $65,000 a year and promises students can complete their core curriculum in two hours a day, thanks to AI, reveals a significant gap between the marketing and the reality. Internal documents show the school’s AI-generated reading comprehension questions were assessed by employees as doing “more harm than good”: questions with illogical answer choices, questions answerable without reading the article, and a hallucination rate high enough that the AI was deemed unsuitable for one key task. The company relies on AI to test the quality of its AI-generated lessons, which, predictably, doesn’t catch the problems. Meanwhile, student surveillance is pervasive: an app called StudyReel monitors screen activity, webcam footage, mouse movements and app usage, and hours of video recordings of students’ faces are stored in a Google Drive accessible to anyone with the link. One employee noted internally that they needed to get better at “selling” the surveillance to parents. Former employees credit students’ strong test scores not to the AI, but to the human tutors who cared about them.

    3. ChatGPT flagged the Tumbler Ridge shooter — and didn’t alert police — The Verge

    Months before the February 10th attack, the suspected perpetrator’s conversations with ChatGPT, describing gun violence, triggered OpenAI’s automated review systems. Employees debated escalation. Leadership concluded the posts didn’t meet the bar for an imminent, credible threat, banned the account, and took no further action. OpenAI has since said it proactively contacted the RCMP (the Canadian police force) after the shooting, though the provincial government noted that OpenAI met with a provincial representative the day after the attack without disclosing it held potentially relevant evidence. Canada’s federal AI minister has raised concerns about OpenAI’s safety protocols. There are no easy answers here, and I think that’s the point: this is hard, it requires policy, and no one has built that policy yet.

    🔁 ICYMI

    UNICEF’s Tinkering with Tech — UNICEF Digital Education

    Against the week’s heavier stories, this is worth sitting with. UNICEF’s Tinkering with Tech initiative uses micro:bit devices (or similar) and design thinking to bring hands-on AI and STEM learning to children, with a deliberate focus on girls, and on building skills rather than consuming technology. From mid-2025, AI literacy was added to the programme, which is now expanding from its initial four-country pilot to Lao PDR, Ukraine, and Uzbekistan. The approach is meaningfully different from “here is an AI tool, use it”: students identify a real-world problem their community faces, build something, reflect. Worth bookmarking for anyone thinking about what AI education that centres children, rather than the technology, can look like.

    🔬 What’s new with CAISE

    Last Tuesday I was back at the University of Kent’s Institute of Cyber Security for Society (iCSS), where I did my PhD, talking about children’s use of generative AI. My talk, “Canaries in the AI Coal Mine”, drew on Project CAISE to argue that we need to shift the conversation from what we’re keeping children away from, to what we’re equipping them to deal with.

    The numbers keep moving: over 75% of UK 13-18 year olds are already using generative AI, and in one recent survey, over half had confided something serious to an AI companion. They aren’t waiting for permission. The adults are still arguing about whether to let them in while they’re already through the door.

    I got some excellent tricky questions from the audience, including what happens if your research participants suddenly get banned from a large number of the platforms you’re studying…I guess we will find out if and when it happens!

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know — or share this with someone thinking about these questions.

  • CAISE Notes – Issue #1

    CAISE Notes – Issue #1

    Welcome to the very first CAISE Notes. Every week: what’s worth knowing about young people and AI, from the research, the policy, and the field.

    This week: policy theatre, student agency, and an entry-point for getting hands-on with AI.

    🔍 This week I’ve been thinking about…

    Monday brought another UK government announcement about consulting on banning social media for under-16s — the second in as many months. I’ll confess I’m itching to respond formally, but there’s no formal consultation to respond to. I think next month?

    Until then: the arguments against bans aren’t new. Young people face two distinct types of harm online: platform design problems (algorithmic amplification, infinite scroll); and criminal behaviour (deepfakes, harassment, fraud) that AI has made devastatingly easier to commit. Neither of these harms are age-specific — both affect everyone and require solutions that work for everyone.

    Bans create a third harm that is unique to children: by choosing exclusion over education, we deny young people the literacy that comes from a mixture of education and lived experience. This generation could have been the first to navigate these risks well from childhood. Instead, it looks like we’re ensuring they won’t.

    What’s striking about this political moment is the opportunity cost. The political capital spent on age restrictions could instead demand platforms fix their design for everyone, and build genuine digital literacy. That’s harder. But it’s the work that would actually help.

    This is why Project CAISE feels urgent: understanding how young people actually navigate these technologies seems foundational to any policy that might help rather than harm them.

    📰 Three things worth your attention

    1. The UK Government’s proposals, explained — The Guardian

    All the coverage summarises a Substack post from Keir Starmer, so here’s what you need to know: UK Government are looking at the ban on social media (which would be fast tracked into law by some legislative sleight of hand); extending online safety rules to AI chatbots (good), and forcing social media companies to provide children’s data after their death to coroners or Ofcom (also good). Worth reading alongside the House of Commons Library briefing for the evidence landscape.

    2. What 200 students actually said when asked about AI policy — Honolulu Civil Beat

    A high school sophomore writes about a Stanford-facilitated deliberation event where 200 students across 19 states worked through the future of AI in schools. Their conclusion was neither ban it nor embrace it uncritically — it was understand it. This is what genuine student voice looks like, and it’s a useful counterpoint to policies developed without it. (And gives me hope for excellent outcomes for CAISE!)

    3. Adventures in vibe coding — Naomi Alderman, Whatever Works

    Novelist Naomi Alderman spent a weekend building her own personal software tools using AI, with no prior coding experience. Her reflection on what it felt like (“a feeling of mastery and agency”) is a useful provocation: if we want young people to be critical, confident navigators of AI, the adults around them need to get their hands dirty too. (Paywalled, but the free preview makes the key point.)

    🔁 ICYMI

    “What I wish my parents or carers knew…” — Children’s Commissioner for England, December 2025

    A practical guide for parents on navigating children’s digital lives — but notice whose voice frames it. The title is drawn from what children said they wished adults understood. There’s also a companion activity pack designed to be used directly with children, which is worth bookmarking for anyone working in schools, or any parent. In a week where policy is being made about young people rather than with them, this is a wonderfully useful document that I will refer back to repeatedly, I think.

    🔬 What’s new with CAISE

    The ethics proposal is in! For those of you who aren’t researchers, this is the major hurdle we need to clear before we can get on and research. It’s not a small piece of work — to do it right, you need to know exactly what you’re going to do, how, and what the risks are. Hopefully, we’ll have some good news very soon.

    → What are you seeing in your school, your research, or your own use of AI this week?

    Let me know — or share this with someone who’s thinking about these questions.

  • We’ve submitted our ethics application!

    We’ve submitted our ethics application!

    There’s a particular kind of energy that comes with hitting ‘submit’ on a document you’ve been working on for weeks – equal parts relief, excitement, and uncertainty about whether you’ve got the balance right.

    Over the weekend, we submitted our ethics application for Project CAISE. It’s been months in the making, and it’s one of our first significant milestones.

    Submitting an ethics application is always a daunting prospect. It forces you to think – really think – about all the aspects of your plan. What are you really doing? Why are you doing it? Is that really the best way to do it?

    Done right, it allows you not only to make sure your “why” is crystal clear, but it also allows you to do a lot of the hard thinking ahead of time: yes, you’re going to interview people – but what, precisely are you going to ask them? You say you’re going to survey people, but with what tool and how do you know it’s GDPR-compliant? You’re going to video record that activity? Cool. How do you square keeping the meaningful visual interactions with the fact you’ve inadvertently recorded everyone’s faces? What even is the university’s infrastructure for storing things for the entire 10 year post-research retention period?!

    For a qualitative researcher, a lot of the above comes up every time. One thing that’s really hit me this time around, though: how do you meaningfully assess risk when the technology itself is evolving faster than our understanding of it? CAISE is a long(ish)-term project with 13-14 year olds navigating AI in their everyday lives. Not only do we not know what AI will look like this time next year, but we are aiming for an unvarnished and judgement-free exploration of this to really support open communication and understanding.

    In a world where it seems to be ok to produce non-consensual graphic images of people – including children – using AI on social media, and where banning children from spaces rather than working to make those spaces safer, as we would offline, the findings we will have are more important than ever.

    The ethics process asks us to anticipate and mitigate risks. But when the risks themselves are emergent and fundamentally uncertain? That’s trickier. We’ve made our best judgments based on current evidence and built in reflexivity, ongoing consent and a series of tiered pastoral and safeguarding processes. But there’s honesty required here too: we’re researching precisely because we don’t fully understand what’s happening yet, no matter how innocuous, or awful, it might be.

    We’ll be here with fingers crossed, full of nerves, until we hear back – which won’t be for a while…!