http://www.cam.ac.uk/taxonomy/term/115/feed

Rebalancing the planet must happen faster. Cambridge researchers are using AI to help.  

With the right development and application, AI could become a transformative force for good. What's missing in current technologies is human insight, says Anna Korhonen.

In the battle against climate disinformation, native advertising is a fierce foe. A study published in the journal npj Climate Action by researchers from Boston University (BU) and the University of Cambridge, evaluates two promising tools to fight misleading native advertising campaigns put forth by big oil companies.

Many major news organisations now offer corporations the opportunity to pay for articles that mimic in tone and format the publication’s regular reported content. These ‘native advertisements’ are designed to camouflage seamlessly into their surroundings, containing only subtle disclosure messages often overlooked or misunderstood by readers. Fossil fuel companies are spending tens of millions of dollars to shape public perceptions of the climate crisis.

“Because these ads appear on reputable, trusted news platforms, and are formatted like reported pieces, they often come across to readers as genuine journalism,” said lead author Michelle Amazeen from BU’s College of Communication. “Research has shown native ads are really effective at swaying readers’ opinions.”

The study is the first to investigate how two mitigation strategies — disclosures and inoculations — may reduce climate misperceptions caused by exposure to native advertising from the fossil fuel industry. The authors found that when participants were shown a real native ad from ExxonMobil, disclosure messages helped them recognise advertising, while inoculations helped reduce their susceptibility to misleading claims.

“As fossil fuel companies invest in disguising their advertisements, this study furthers our understanding of how to help readers recognise when commercial content is masquerading as news and spreading climate misperceptions,” said co-author Benjamin Sovacool, also from BU.

“Our study showed that communication-led climate action is possible and scalable by countering covert greenwashing campaigns, such as native advertising, at the source,” said co-author Dr Ramit Debnath from Cambridge’s Department of Architecture. “The insights we’ve gained from this work will help us design better interventions for climate misinformation.”

The research builds on a growing body of work assessing how people recognise and respond to covert misinformation campaigns. By better understanding these processes, the researchers hope that they can prevent misinformation from taking root and changing people’s beliefs and actions on important issues like climate change.

‘The Future of Energy’ ad

Starting in 2018, readers of The New York Times website encountered what appeared to be an article, titled “The Future of Energy,” describing efforts by oil and gas giant ExxonMobil to invest in algae-based biofuels. Because it appeared beneath the Times’ masthead, in the outlet’s typical formatting and font, many readers likely missed the small banner at the top of the page mentioning that it was an ad sponsored by ExxonMobil.

The ad, part of a $5-million-dollar campaign, neglected to mention the company’s staggering carbon footprint. It also omitted key context, The Intercept reported, like that the stated goal for algae-based biofuel production would represent only 0.2% of the company’s overall refinery capacity. In a lawsuit against ExxonMobil, Massachusetts cited the ad as evidence of the company’s “false and misleading” communications, with several states pursuing similar cases.

Putting two interventions to the test

The researchers examined how more than a thousand participants responded to “The Future of Energy” ad in a simulated social media feed.

Before viewing the ad, participants saw one, both, or neither of the following intervention messages:

An inoculation message designed to psychologically ‘inoculate’ readers from future influence by broadly warning them of potential exposures to misleading paid content. In this study, the inoculation message was a fictitious social media post from United Nations Secretary-General Antonio Guterres reminding people to be wary of online misinformation.

A disclosure message with a simple line of text appearing on a post. In this study, the text “Paid Post by ExxonMobil” accompanied the piece. Studies have shown that more often than not, when native ads are shared on social media, this disclosure disappears.

Bolstering psychological resilience to native ads

The team found that the ad improved opinions of ExxonMobil’s sustainability across the study’s many participants, regardless of which messages they saw, but that the interventions helped to reduce this effect. Some of the key findings include:

The presence of a disclosure more than doubled the likelihood that a participant recognised the content as an ad. However, the participants who had seen a disclosure and those who had not were equally likely to agree with the statement “companies like ExxonMobil are investing heavily in becoming more environmentally friendly.”

Inoculation messages were much more effective than disclosures at protecting people’s existing beliefs on climate change, decreasing the likelihood that participants would agree with misleading claims presented in the ad.

“Disclosures helped people recognise advertising. However, they didn’t help them recognise that the material was biased and misleading,” said Amazeen. “Inoculation messaging provides general education that can be used to fill in that gap and help people resist its persuasive effects. Increasing general awareness about misinformation strategies used by self-interested actors, combined with clearer labels on sponsored content, will help people distinguish native ads from reported content.”

Reference:
Michelle A. Amazeen et al. ‘The “Future of Energy”? Building resilience to ExxonMobil’s disinformation through disclosures and inoculation.’ npj climate action (2025). DOI: 10.1038/s44168-025-00209-6

Adapted from a Boston University story.

A sneaky form of advertising favoured by oil giants influences public opinion with climate action misperceptions, but researchers are studying potential solutions.

rob dobi vai Getty ImagesFueling the Fire of Misinformation - stock photo

The text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Images, including our videos, are Copyright ©University of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways – on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.

Yes

These are some of the questions being addressed by a new initiative launched today at the University of Cambridge, which seeks to address the urgent challenge of managing the risks of future engineered pandemics.

The Engineered Pandemics Risk Management Programme aims to understand the social and biological factors that might drive an engineered pandemic and to make a major contribution to building the UK’s capability for managing these risks. It will build a network of experts from academia, government, and industry to tackle the problem.

Increased security threats from state and non-state actors, combined with increased urbanisation and global mobility, means the threat of deliberate pathogen release must be taken seriously as must other intertwined aspects of pandemic risk such as mis- and disinformation, the erosion of trust in a number of institutions and an increasingly volatile geopolitical context. Further potential risks are posed by recent developments in gene-editing tools and artificial intelligence, which have rapidly advanced technological capability that may make it easier to engineer potential pandemic pathogens.

Professor Clare Bryant from the Department of Medicine at the University of Cambridge said: “There is a great opportunity to take a joined-up approach to managing the risks posed by engineered pandemics. We need experts and agencies across the spectrum to work together to develop a better understanding of who or what might drive such events and what their likely impact would be. And we need evidence-informed policies and networks in place that would help us respond to – or better still, prevent – such an eventuality.”

• The aims of the Engineered Pandemics Risk Management Programme are:
• To develop the conceptual underpinnings for the risk management of engineered pandemics based on interdisciplinary research
• To support the capability of the UK’s engineered pandemic risk policy and practice, including building and maintaining networks that connect government, academia and industry.
• To strengthen the international networks that will support this work globally

There are four main strands of work:

Social determinants of engineered pandemic threat

This strand will look at the actors who have the potential to engineer harmful pathogens, either deliberately or accidentally. It will ask questions such as: What could motivate bioterrorism in the coming decades? Who might the relevant actors be? What are the kinds of engineered pandemic that someone might want to create?

Dr Rob Doubleday, Executive Director of the Centre for Science and Policy at the University of Cambridge, said: “The common narrative is that there’s a wide range of potential actors out there who want to create bioweapons but don’t yet have the technical means. But in fact, there’s been very little work to really understand who these people might be, and their relationship to emerging technology. To explore these questions, we need a broad network including social scientists, biosecurity researchers, criminologists, experts in geopolitics and counterterrorism.”

The strand will also look at the governance of scientific research in areas that may facilitate an engineered pandemic, whether unwittingly or maliciously, aiming to deliver a policy framework that enables freedom of intellectual research while managing real and apparent risk in infectious disease research.

Professor Bryant said: “As scientists, we’re largely responsible for policing our own work and ensuring integrity, trustworthiness and transparency, and for considering the consequences of new knowledge and how it might be used. But with the rapid progress of genomic technologies and AI, self-regulation becomes more difficult to manage. We need to find governance frameworks that balance essential scientific progress with its potential misapplication.”

Biological determinants of engineered pandemic threat

Recognising that the most likely cause of an engineered pandemic would be the deliberate release of a naturally-occurring pathogen – viral or bacterial, for example – rather than a man-made pathogen, this strand aims to understand what might make a particular pathogen infectious and how our immune systems respond to infection. This knowledge will allow researchers to screen currently available drugs to prevent or treat infection and to design vaccines quickly should a pandemic occur.

Modelling threats and risk management of engineered pandemics

The Covid-19 pandemic highlighted practical problems of dealing with pandemic infections, from the provision of personal protective equipment (PPE) to ensuring a sufficient supply of vaccine doses and availability of key medications. Modelling the potential requirements of a pandemic, how they could be delivered, how ventilation systems could be modified, what biosafety measures could be taken, for example, are all key challenges for managing any form of pandemic. This strand will address how existing modelling approaches would need to be adapted for a range of plausible engineered pandemics.

Policy innovation challenges

Working with the policy community, the Cambridge team will co-create research that directly addresses policy needs and involves policy makers. It will support policy makers in experimenting with more joined-up approaches through testing, learning and adapting solutions developed in partnership.

The Engineered Pandemics Risk Management Programme is supported by a £5.25 million donation to the Centre for Research in the Arts, Humanities and Social Sciences (CRASSH) at the University of Cambridge. The team intends it to form a central component of a future Pandemic Risk Management Centre, for which it is now fundraising.

Professor Joanna Page, Director of CRASSH, said: “Cambridge has strengths across a broad range of disciplines – from genetics and immunology to mathematical modelling to existential risk and policy engagement – that can make a much-needed initiative such as this a success.”

To find out more, visit the Engineered Pandemic Risk Management website.

Covid-19 showed us how vulnerable the world is to pandemics – but what if the next pandemic were somehow engineered? How would the world respond – and could we stop it happening in the first place?

There is a great opportunity to take a joined-up approach to managing the risks posed by engineered pandemicsClare BryantMartin SanchezIllustration showing global pandemic spread

The text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Images, including our videos, are Copyright ©University of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways – on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.

YesLicence type: Public Domain

Meet the winner of the Cambridge Awards 2024 for Research Impact and Engagement and learn more about their projects.

Research led by the University of Cambridge, in collaboration with Nottingham Trent University, raises serious concerns about bias in the UK criminal justice system due to negative stereotyping of accents.

These stereotypes, the researchers argue, can affect all parts of the system from arrest to sentencing, and undermine not only suspects and defendants, but also the testimony of witnesses. The study is particularly concerned about accented speakers being incorrectly selected from voice identification parades.

The findings, published in Frontiers in Communication, suggest that despite progress in equality and diversity in some parts of British life, including ‘working-class’ and regional accents becoming more prominent on television and radio, harmful stereotypes remain.

“Our findings bring into sharp focus the disadvantage that speakers of some accents may still face in the criminal justice system,” said lead author, Alice Paver, from the University of Cambridge’s Phonetics Laboratory and Jesus College, Cambridge.

“Voices play a powerful role in the criminal justice system and police officers, lawyers and juries are all susceptible to judging voices based on stereotypes, whether they're aware of it or not. As things stand, listeners think some accents sound guiltier than others and we should all be concerned about that.”

The test

The researchers, from Cambridge and Nottingham Trent University, asked 180 participants (~50:50 gender split) from across the UK to listen to recordings of ten regionally-accented male voices: Belfast, Birmingham, Bradford, Bristol, Cardiff, Glasgow, Liverpool, London, Newcastle and Standard Southern British English (SSBE), also referred to as RP.  

Participants were then asked to rate the voices on 10 social traits – ‘Educated’, ‘Intelligent’, ‘Rich’, ‘Working class’, ‘Friendly’, ‘Honest’, ‘Kind’, ‘Trustworthy’, ‘Aggressive’ and ‘Confident’; as well as on 10 morally ‘good’, ‘bad’ and ‘ambiguous behaviours’, which included a range of crime types.

These behaviours included: ‘Return a lost wallet to its owner’, ‘Stand up for someone who is being harassed’, ‘Cheat on a romantic partner’, ‘Report a relative to the police for a minor offence’, ‘Drive dangerously’, ‘Physically assault someone’, ‘Shoplift’, ‘Touch someone sexually without consent’, ‘Vandalize a shop front.’

The study used a wider range of recorded accents, behaviours and criminal offences than previous research which has tended to focus on criminal behaviour in general or the binary of white versus blue-collar crime. This study included crimes which are not class stratified, such as a driving offence and a sexual offence, and is the first to identify links between listener perceptions of morality, criminality, and social traits.

To ensure their results would be valid in a criminal justice context, the researchers created voice samples in a similar way to how they are constructed for voice ID parades. The aim was to mimic, as closely as possible, how a juror or earwitness would experience them.  

Findings: Status, class and regions

The results show that people with non-standard accents are more likely to be associated with criminal behaviour but that there is significant variation in perceptions between accents.

The RP-like accent was perceived as the least likely to behave in criminal ways, while the Liverpool and Bradford accents were the most likely.

Alice Paver said: “The strongest connection we found was between people's perceptions of class or status, negative traits such as aggression, and how they think someone is going to behave, particularly when it comes to crime. This is the first time that a concrete link between traits and behaviours has been made in the context of accent judgements.”

Unlike previous findings, the researchers did not observe a relationship between ‘solidarity traits’ (such as kindness and trustworthiness) and any behaviours. Status proved a much more important predictor of behaviours, re-enforcing the link between social class and expectations of behaviour in the UK.

However, non-English accents, in particular Belfast’s and Glasgow’s, were rated significantly less likely to behave in criminal ways than almost all other accents. They were also thought most likely to ‘stand up for someone being harassed’ (‘honourable behaviour’) and least likely to exhibit ‘morally bad’ behaviours.

Alice Paver said: “Our findings show that perceptions of speakers of regional accents and how status, social attractiveness and morality interact are much more complex than previously assumed. We need a much more nuanced understanding of how accents are evaluated when it comes to different crime types.”

Findings: Sexual offences

The London and Liverpool accents were rated most likely to touch someone sexually without consent, but they were very closely followed by the RP accent. Participants thought the RP accent was more likely to commit a sexual assault than any of the other offences tested.

“This finding simultaneously undermines certain traditional stereotypes about both higher status and working-class men,” Alice Paver said. “This may indicate shifting perceptions of the ‘type’ of man who can and does commit sexual offences.”

The Glasgow and Belfast speakers were thought the least likely to commit this sexual offence.

The study found that participants perceived this sexual offence as distinct from other criminal behaviours. Poor ratings for it clustered with those for non-criminal ‘morally bad’ behaviours, namely ‘being unfaithful to a romantic partner’ and ‘lying on a CV’.

Findings: Newcastle and Birmingham

Previous studies have found that the Newcastle accent rates highly for traits such as friendliness, but this study recorded less positive ratings for kindness, honesty, friendliness and trustworthiness.

By contrast, the Birmingham accent, which has rated poorly in previous research across these measures, performed better than Bradford, Bristol, Liverpool, London and Newcastle in this study.

“Although relatively stable over time, language attitudes can change,” Alice Paver said. “This might be the case for the Birmingham and Newcastle accents. But previous studies have often asked people what they think of an accent label whereas we played them an actual voice. That’s a very different stimulus so we’re not surprised people reacted differently.”

Bringing about change

The study contributes to the Improving Voice Identification Procedures project. Its team of researchers is currently drafting revised guidelines for voice identification parades aimed at police officers and legal professionals.

They support the use of pre-tests to screen for bias against foil or suspect voices to make sure that they don't stand out as sounding unduly guilty or untrustworthy.

“Jurors are not currently made aware of or warned against letting voice- or accent-based prejudice sway their decisions,” Paver said. “If we're asked to judge whether someone is guilty or not, and they've got a particular accent, we need to be sure we're not making that judgment because we think they sound like a bad guy.”

The researchers hope that future studies will examine even more offence types; further explore the relationships between perceptions of criminality and other, non-criminal, behaviours; and make use of a broader range of voices for each accent to tease apart the effect of individual voices and the strength of regional accents.

The research was carried out in collaboration with Professor Natalie Braber and Dr David Wright of Nottingham Trent University’s School of Arts and Humanities, and Dr Nikolas Pautz, of NTU's Dept. of Psychology.

 

Funding

This research was supported by the United Kingdom Economic and Social Research Council as part of the project Improving Voice Identification Procedures (IVIP), reference ES/S015965/1. Additional funding was provided by the Isaac Newton Trust.

Reference

A. Paver, D. Wright, N. Braber and N. Pautz, ‘Stereotyped accent judgements in forensic contexts: listener perceptions of social traits and types of behaviour’, Frontiers in Communication (2025). DOI: 10.3389/fcomm.2024.1462013

People who speak with accents perceived as ‘working-class’ including those from Liverpool, Newcastle, Bradford and London risk being stereotyped as more likely to have committed a crime, and becoming victims of injustice, a new study suggests.

Listeners think some accents sound guiltier than others and we should all be concerned about thatAlice PaverArt De Cade via Flikr under a CC licenseThe Old Bailey in London. Photo: Art De Cade via Flikr under a CC license

The text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Images, including our videos, are Copyright ©University of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways – on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.

YesLicence type: Attribution