Written evidence submitted by Sense about Science (SMH0041)

Social media, misinformation and harmful algorithms Inquiry

Sense about Science

Sense about Science is an independent charity that promotes the public interest in sound science and evidence. We work with communities all over the UK to make sense of evidence, and with researchers and policy makers to raise the standard of evidence in public life. We have 20 years’ experience in equipping people with good questions to navigate the most complex and controversial areas of evidence and policy making, from parents of children undergoing heart surgery, to politicians faced with output from simulation modelling.

We welcome the Committee’s inquiry into this neglected area, and in particular its focus on how algorithms determine the online content that users see – a key subject that has been neglected in other reviews. The lack of transparency in how these algorithms operate presents a significant risk not only of people being shown biased and harmful information, but of receiving lawful but highly polarised material that does not enable them to make a balanced and informed decisions. We cannot hold users solely responsible for what they see and share on social media – why, for example, does a video from Texas, purporting to show there were no Covid cases during the pandemic, get put into the feed of someone in a boxing club in North Kent? Why do we allow the social media feed of young people at risk of self-harm to be filled exclusively with content that enforces their beliefs, and provides no counterstatements?

1. To what extent do the business models of social media companies, search engines and others encourage the spread of harmful content, and contribute to wider social norms?

Social media algorithms are designed to serve the attention economy by holding the user’s attention, promoting engagement and ensuring they stay on the site for as long as possible. They act as ‘negligent curators’ choosing what to promote for their own ends rather than that of users. This means misinformation or harmful content that is presented as entertaining has an advantage on social media over optimal information[1]. Such content is often emotive and confirms our pre-existing biases, thus holding our attention and leading to further promotion of similar content by algorithms.

Algorithms are deciding more and more what content users can see[2], with more than half of all content seen on some platform recommended by AI[3] rather than a result of following others’ accounts. Currently companies are not held accountable for how they curate the information people see – and, just as importantly, do not get to see – on their platforms.

Most efforts to combat the spread of misinformation rely on fact-checking services or censorship of specific content. While removal of misinformation is useful, a lack of transparency in process can lead to users being suspicious of what information is hidden from them, driving attempts to subvert official sources. Opaque censorship hampers the ability for people like us to see what others have read, and the ability of society to address misinformation and its consequences. Our work with people minded to conspiracy theories during covid[4] found that dismantling false beliefs requires engagement and opening questioning of all sources, with authoritarian approaches of censorship only entrenching views.

Automated censorship tools employed by companies that are opaque, and with no external oversight also results in the suppression, removal or demotion of good quality information – often undermining important discussions of difficult or contentious issues where there is growing and conflicting evidence. For example, Facebook tagging a credible BMJ article as ‘partly false’ because 'the authors did not express unreserved support for vaccination’, or Instagram ‘shadow banning’ a Cochrane (Harding Prize for Trustworthy Science Communication[5] winning) review for ‘false content’, which was also tagged by Twitter as ‘misleading’[6]. The nature of emerging science and innovative research means that alignment with existing consensus is not a good measure for veracity.

Not only are these methods often counterproductive[7], they fail to address the underlying issue of what content is promoted to users’ feeds by algorithms - the primary solution to misinformation should be ensuring people have access to good information, and recourse to balanced sources, not restricting access to information, or limiting opportunities to ask good questions.

2. How effective is the UK’s regulatory and legislative framework on tackling these issues?

a) How effective will the Online Safety Act be in combatting harmful social media content?

Censorship is not a sustainable long-term solution to address the growing challenges from misinformation and the spread of harmful content. Whilst censorship can lead to the desired outcomes for specific cases in the short-term, it exacerbates the scepticism towards authority that believers of misinformation often already hold[8]. Regulators and government agencies should act as guides to good information, but this legislation works to strengthen the perception that they are seeking to control the narrative, which will likely contribute to feelings of mistrust: people’s attraction to misinformation is sometimes an expression of how they feel about institutions and authority and outlooks associated with them. Therefore, appealing to authority for good information does not always work; trusted sources are only effective on people who already trust them.

The interventions proposed in this legislation do not do enough to address the underlying issues of bad information being promoted at the detriment of good information, or the problem of users’ feeds providing polarising content that inevitably gives them a skewed perspective on an issue. People have become accustomed to blaming other users for putting up poorly judged or ill-informed content, but the real question should be why the content has made it into people’s feed. This hidden curation of information is the most neglected aspect of initiatives to tackle misinformation. If we want platforms, and the effect they have on society, to change, we need to focus on how they curate social media feeds, rather than being diverted into wild goose-chases of what people are posting.

b) What more should be done to combat potentially harmful social media and AI content?

Our work indicates there is strong public appetite for good information and reliable sources. Our 2022 NatCen survey on how people used government information during the pandemic, for example, found that 15% of the public accessed the ONS website in pursuit of reliable information, with statements from experts the most preferred format[9]. This is also demonstrated by public engagement with Evidence Week in Parliament[10], organised by Sense about Science in Partnership with the parliamentary Office for Science and Technology (POST), and support for our call for evidence transparency in government policymaking[11]. Despite the widespread recognition of the threat from misinformation, there is dangerously little recognition of the lack of resources for those who help guide people to good information – librarians, specialists, journalists, editors and research integrity officers. These groups support the seeking of information against a transparent set of public good criteria, which is a distinct contrast to the opaque, for-profit criteria governing social media and search engine algorithms. Public libraries have seen a substantial cut in funding[12] and over 180 have closed or been passed to charities since 2016[13]. Specialist journalists are similarly seeing a distinct lack of funding. In our experience, people benefit from knowing more about how knowledge was generated and how reliable it is.

Currently the focus is on reactive responses to misinformation, but a long-term solution must involve improving the availability of good information and supporting those who guide us to it. This should involve input from established curators who work to guide people to good information, such as librarians and journalists. We urge the committee to consider what good information curation should look like in a digital age.

The best way to counter mis- and disinformation is to create a level playing field by providing good access to useful information. We cannot prevent ‘false narratives’ without providing a full account of ‘true narratives’. To achieve this, good information has to be available and accessible. This requires social media algorithms to be transparent about what information is being surfaced, and why. How much better placed would users be to deal with content if it was honestly labelled e.g. you are being shown this because a) “a friend wrote about it”; or b) “we think it will make you angry and keep you here longer, and we can show you six more adverts”?

Commercial confidentiality need not be a barrier to transparency requirements that safeguard society’s interest, if companies are required to publicly set out the explicit the objectives of their algorithms (whether written by humans or self-generating), even if not every line of code is shared. To ensure that social media algorithm, which are a leading provider of information for many demographic groups, are working as intended, the processes by which information is shared/promoted or suppressed/censored should be open to public scrutiny.

18 December 2024