Chatbots can sway political opinions but are ‘substantially’ inaccurate, study finds
Chatbots can sway people’s political opinions but the most persuasive artificial intelligence models deliver “substantial” amounts of inaccurate information in the process, according to the UK government’s AI security body.Researchers said the study was the largest and most systematic investigation of AI persuasiveness to date, involving nearly 80,000 British participants holding conversations with 19 different AI models.The AI Security Institute carried out the study amid fears that chatbots can be deployed for illegal activities including fraud and grooming.The topics included “public sector pay and strikes” and “cost of living crisis and inflation”, with participants interacting with a model – the underlying technology behind AI tools such as chatbots – that had been prompted to persuade the users to take a certain stance on an issue.Advanced models behind ChatGPT and Elon Musk’s Grok were among those used in the study, which was also authored by academics at the London School of Economics, Massachusetts Institute of Technology, the University of Oxford and Stanford University.
Before and after the chat, users reported whether they agreed with a series of statements expressing a particular political opinion.The study, published in the journal Science on Thursday, found that “information-dense” AI responses were the most persuasive.Instructing the model to focus on using facts and evidence yielded the largest persuasion gains, the study said.However, the models that used the most facts and evidence tended to be less accurate than others.“These results suggest that optimising persuasiveness may come at some cost to truthfulness, a dynamic that could have malign consequences for public discourse and the information ecosystem,” said the study.
On average, the AI and human participant would exchange about seven messages each in an exchange lasting 10 minutes.It added that tweaking a model after its initial phase of development, in a practice known as post-training, was an important factor in making it more persuasive.The study made the models, which included freely available “open source” models such as Meta’s Llama 3 and Qwen by the Chinese company Alibaba, more convincing by combining them with “reward models” that recommended the most persuasive outputs.Researchers added that an AI system’s ability to churn out information could make it more manipulative than the most compelling human.“Insofar as information density is a key driver of persuasive success, this implies that AI could exceed the persuasiveness of even elite human persuaders, given their unique ability to generate large quantities of information almost instantaneously during conversation,” said the report.
Feeding models personal information about the users they were interacting with did not have as big an impact as post-training or increasing information density, said the study.Kobi Hackenburg, an AISI research scientist and one of the report’s authors, said: “What we find is that prompting the models to just use more information was more effective than all of these psychologically more sophisticated persuasion techniques.”However, the study added that there were some obvious barriers to AIs manipulating people’s opinions, such as the amount of time a user may have to engage in a long conversation with a chatbot about politics.There are also theories suggesting there are hard psychological limits to human persuadability, researchers said.Hackenburg said it was important to consider whether a chatbot could have the same persuasive impact in the real world where there were “lots of competing demands for people’s attention and people aren’t maybe as incentivised to sit and engage in a 10-minute conversation with a chatbot or an AI system”.
NHS braces for ‘unprecedented flu wave’ as hospitalised cases in England rise
The NHS is facing an “unprecedented flu wave”, a senior healthcare leader said, as the number of people with the illness in hospitals across England hit a record high.The statistics, published by NHS England as part of its first weekly snapshot of the performance of hospitals this winter, found that an average of 1,717 flu patients were in beds each day last week – more than 50% higher than last year – including 69 patients whose condition was critical.This is compared with 1,098 flu patients in NHS hospital beds at the same time last year, representing a 56% increase. In 2023 there were 160 flu patients in beds in the same week.Prof Julian Redhead, the national director for urgent and emergency care, said the figures confirmed the NHS’s “deepest concerns” that the health service was bracing for an unprecedented flu wave this winter, with cases “incredibly high for this time of year and there is no peak in sight yet”
Children in England most active since 2017 – but majority still fall short of targets
Children in England are the most active they have been since 2017, according to research that warns that less than half are meeting government activity level targets.In 2024-25, 3.6 million children took part in an average of more than 60 minutes of sport and physical activity per day across the week, according to Sport England’s annual active lives survey. This represents 49.1% of five- to 16-year-olds in England – a 1
Parents and young people: share your concerns about ultra-processed foods (UPFs)
This month, the Lancet published the world’s largest review on the health threats of ultra-processed foods (UPFs), showing that they are replacing fresh food on every continent and are exposing millions of people to long-term harm. Globally one in 10 children are considered obese, as junk food overwhelms childhood diets. Previous research has shown how susceptible children are to junk food advertising.Parents and young people, are you concerned about the level of UPFs in your diet? Is it easy and affordable to find fresh food and eat healthily where you live? What changes do you think would help encourage healthy eating habits? We’re particularly interested in hearing from parents in low- and middle-income countries where the rise in childhood obesity is steepest.You can share your concerns about ultra-processed foods (UPFs) using this form
‘I don’t take no for an answer’: how a small group of women changed the law on deepfake porn
For Jodie*, watching the conviction of her best friend, and knowing she helped secure it, felt at first like a kind of victory. It was certainly more than most survivors of deepfake image-based abuse could expect.They had met as students and bonded over their shared love of music. In the years since graduation, he’d also become her support system, the friend she reached for each time she learned that her images and personal details had been posted online without her consent. Jodie’s pictures, along with her real name and correct bio, were used on many platforms for fake dating profiles, then adverts for sex work, then posted on to Reddit and other online forums with invitations to deepfake them into pornography
Pornography company fined £1m by Ofcom for not having strong enough age checks
A pornography company that runs 18 adult websites has been fined £1m by the watchdog Ofcom for not having strong enough age checks, in the largest fine yet under the UK’s Online Safety Act.The Belize-based AVS Group has been hit with the punishment, plus a further £50,000 for failing to respond to information requests.It is the third time that the internet and communications watchdog has fined a company in relation to the Online Safety Act, which brought into force strict age-checking requirements in July.While AVS has implemented what it claims is an age verification regime, the regulator’s investigation did not deem it to be highly effective.The company now has 72 hours to introduce age checks that Ofcom will view as effective or face a penalty of £1,000 a day
Probation officers in England and Wales to be given self-defence training after stabbings
Probation officers will be given self-defence training, bleed kits and body-worn cameras for the first time under plans before ministers in the wake of two stabbings, the Guardian has learned.Knife arches and handheld metal-detecting wands, which can be used to search people for weapons, have been approved for pilot schemes in selected offices.The disclosures come days after a staff member was stabbed in a probation office in Oxford. Separately, a man has admitted the attempted murder in July of a female officer in another probation centre in Preston, Lancashire.The probation officers’ union, which believes these are the first knife attacks in probation offices, has said members have a “palpable fear” about going to work since the attacks
The great Christmas taste test: I tried seven fast food offerings. Which will make me feel festive?
Party starters: Jacob Kenedy’s Italian Christmas canapes – recipes
Australian supermarket sorbet taste test: is this the most enjoyable taste test yet?
Christmas main course made easy: Max Rocha’s braised turkey legs with colcannon – recipes
Christmas mains: Georgina Hayden’s pan-fried monkfish in a herby champagne butter – recipe
How to make coquilles St-Jacques – recipe | Felicity Cloake's Masterclass