Written evidence submitted by the Internet User Behaviour Lab (SMH0058)

Dear Members of the Committee,

Re: Call for Evidence: Social media, misinformation, and harmful algorithms

The Internet User Behaviour Lab (IUBL), funded by the Internet Society Foundation, is affiliated with the University of Missouri, Kansas City, US. IUBL is a UK-US academic group that employs social and network science research to study the dynamics of social media, enhances Algorithmic Literacy, and develops evidence-based tools that increase user internet agency to tackle misinformation (Boots, B. C., Krause Matlack, A., Richardson-Gool, T. S., 2024).

This submission addresses the committee's concern about how social media companies' business models enable the spreading of harmful online content. We focus on why online social network echo chambers are a problem that can have real-world harm, with algorithmic recommendations contributing to polarisation and social fragmentation. Our research found higher educational attainment is correlated with greater confidence in news sources (possibly due to better filtering skills and lower susceptibility to misinformation), with younger users also showing more trust. We call for Algorithmic Literacy to be integrated into education to give users (particularly those with lower education levels or susceptible to being marginalised online in echo chambers) to improve awareness and agency from an individual, community, and ground-up fashion to reduce the spread of harmful online content.

Problem

Online social networks are integrated into modern communication and driven by business models prioritising engagement and revenue through algorithmic content manipulation. Users do not have access to inspect these algorithms, nor information about how they work, and they have minimal agency to change them. Echo chambers that reinforce existing beliefs and entrench homogeneity of thought and opinion can facilitate the spread of misinformation and disinformation while reducing the information diversity of online users. Online political homophily—where we seek and congregate with like-minded individuals—reinforces ideological segregation driven by algorithmic recommendations and user choices and plays to our cognitive bias. Further, the structure of social media networks fosters echo chambers whereby platforms like Twitter/X and Facebook segregate users into distinct clusters ["filter bubble" effect] that limit interaction between opposing views, thus reducing opportunities for cognitive dissonance.  (Boots, et al., 2024; Matlack et al., 2024). These systems amplify divisive narratives and contribute to ideological polarisation, social fragmentation, and real-world harms, as evidenced during the UK 2024 riots.

Findings

●        Connectivity doesn't necessarily improve trust: Higher Internet connectivity does not consistently increase trust in online social networks such as news sources. We found UK users who comparably have high connectivity reported lower trust in online social networks for news than in less connected countries, such as India and South Africa.

●        Education and trust: Contrary to our initial hypothesis, we found a correlation between higher education level and greater trust in their online social networks. This may reflect better user filtering and lower levels of being targeted by unreliable sources. Younger users also demonstrated higher trust in online social networks than older demographic groups.

●        Echo chambers: Internet users with higher education levels and younger demographic groups reported proactively diversifying their online information consumption.

●        Algorithmic manipulation: Higher education levels are related to an increased awareness of algorithmic manipulation. Frequent online social network users also reported heightened concern.

Gaps in the literature:

●        Education and trust paradox: The relationships identified are not causative. The question of why higher education increases trust in online social networks for news while being more aware of the potential for algorithmic manipulation requires further examination. One possible explanation includes better algorithmic evaluation skills.

●        Diverse Sampling: More research that includes different education backgrounds, more countries, and individuals with lower awareness of echo chambers and algorithmic manipulation

●        Cultural Dimensions: More understanding of how societal and cultural contexts influence digital trust and online behaviour.

Algorithmic Literacy 

We define Algorithmic Literacy as an integrated pedagogical, civic, and cultural understanding of how technological algorithms – particularly on, but not limited to, social media platforms – manipulate and filter information to retain information seekers' attention. Algorithmic Literacy is the ability of everyday citizens (without any training in computer science or data science) to: 

●        Recognise when they are interacting with a technological system that is likely to be driven by algorithms;

●        to Reason (understand, infer) about what sorts of data the algorithm-driven systems may be collecting and have access to about the person interacting with it, the decisions the algorithm-driven systems might be likely to make about the person interacting with the algorithm-driven systems, and how the algorithm-driven systems are likely to make these decisions; 

●        to Respond, or to take decisions and actions about how or if they want to interact with algorithm-driven systems based on their recognition and reasoning.

●        And to Retain, where recognition, reasoning, and response behaviour are sustained and become a habit.

We call for integrating Algorithmic Literacy into digital education programs to empower users to understand how algorithmic systems influence their trust and behaviour. Educating users about platform algorithms could help reduce blind reliance on content and foster informed scepticism, even as physical Internet infrastructure expands.

This recommendation forms part of a broader effort to advance international best practices in online safety and digital governance. As such, we disclose that it was previously submitted to the Australian Standing Committee in November 2024 in response to the Online Safety Amendment (Social Media Minimum Age) Bill 2024.

Five Policy Proposals for Supporting Algorithmic Literacy

We propose embedding Algorithmic Literacy within broader media and digital literacy frameworks to safeguard young users against algorithmic manipulation. Specific actions include: 

Proposition 1: Frame Policy on Algorithmic Literacy using a Capabilities Approach 

The Capabilities Approach, underpinned by Amartya Sen’s Nobel-winning research and Martha Naussbaum's work, emphasises enhancing individuals' freedom and ability to lead lives they value with human dignity (Nussbaum, 2001). It helps provide a way to extend Algorithmic Literacy and a lens to assess what is needed in a digitalised socio-economic environment to advance the individual's basic capabilities (innate potential),  internal capabilities (developed through learning and training), and combined capabilities (enabled by supportive social, economic, and political environments). For example, some of Nussbaum's ten central human capabilities may guide the positioning of research and policy: 

●        Senses, Imagination and Thought: with Algorithmic Literacy aimed at helping information seekers have greater agency to choose pleasurable experiences. 

●        Practical Reason: where information seekers are more engaged in critical reflection and aware of how algorithmic technologies shape their lives. 

●        Affiliation: Algorithmic Literacy is a provision to foster non-discrimination practices to reduce potential biases. 

●        Control Over One's Environment: Algorithmic Literacy can improve political agency by helping individuals discern how algorithms are used to target them. 

Proposition 2: Media and digital literacy should incorporate Algorithmic Literacy

A comprehensive digital and media literacy approach should encompass content evaluation, technical proficiency,  and an acute awareness of algorithmic manipulation and its implications. While pedagogy exists to educate information seekers on evaluating media content, using digital tools, and recognising credibility and bias, a gap exists in understanding algorithms and their impact on information dissemination and user manipulation.

Traditional media literacy education should cover social media and AI algorithmic functions. We need a pedagogical framework that includes:

●        Recognising, reasoning, responding, and retaining algorithmic influences. 

●        Understanding how algorithms influence media content distribution and visibility. 

●        Developing technical skills to interact with and comprehend digital algorithms. 

●        Critically evaluating algorithm-driven media recommendations. 

Proposition 3: Establish pedagogical approaches for imparting Algorithmic Literacy

Attention has been given to digital and media literacy programs in education, community-based initiatives, and public awareness campaigns. However, the next step is integrating specific Algorithmic Literacy pedagogy into these efforts. We recommend prioritising the development of methods and resources to help learners understand how algorithms operate, their impact, and their role in technology. 

Research is needed on the delivery methods and types of tools provided to information seekers to build these  Algorithmic Literacy skills. Additionally, studies should explore how personality traits and psychological factors influence Algorithmic Literacy and compare different strategies and curricula for educating individuals about online social networks and artificial intelligence. 

Proposition 4: Develop an Algorithmic Literacy program using a cross-stakeholder approach

A multi-stakeholder approach is recommended to implement algorithmic literacy education. Collaboration between government, education, and technology sectors is needed and can help to ensure that digital environments are accessible, transparent, and equitable, reducing the risks of exclusion or manipulation.

Proposition 5: Community-focused Algorithmic education

The UK is a highly connected region, yet some communities face digital inequalities. Some need to be more connected, and others may be epistemically marginalised due to different algorithmic literacy levels. Programs might be targeted to help these communities and users recognise when algorithms manipulate content and develop critical reasoning skills to assess reliable online information.

Conclusion

Algorithmic Literacy is an upstream intervention to improve social media and online social network behaviour to equip our generation and the next with the knowledge, skills, and agency to adapt to an algorithmically driven world. We call for integrating Algorithmic Literacy into digital education programs to empower users to understand how algorithmic systems influence their trust and behaviour.

We urge the Committee to support Algorithmic Literacy development to combat echo chambers, reduce harmful misinformation, and build online resilience.

The Internet User Behaviour Lab, affiliated with the University of Missouri Kansas City, US

18 December 2024

Acknowledgements

A grant from the Internet Society Foundation funded the research referenced in this document.

References

●        Boots, Bryan C., Alex Krause Matlack, and Theo S. Richardson-Gool. "A Call for Promoting Algorithmic Literacy." Available at SSRN 4912427 (2024).

●        Matlack, Alex Krause, Bryan C. Boots, and Theo S. Richardson-Gool. "The Impact of Internet Connectivity in Navigating Online Social Networks: A Cross-Country Analysis." Available at SSRN 4913130 (2024).

●        Nussbaum, M. C. (2001). Symposium on Amartya Sen’s philosophy: 5 adaptive preferences and women’s options.  Economics and Philosophy, 17(1), 67–88. 

●        Hobbs, R. (2020). Propaganda in an age of algorithmic personalization: Expanding Literacy Research and Practice.  Reading Research Quarterly, 55(3), 521–533.

●        Palincsar, A. S. (1998). Social Constructivist Perspectives on teaching and learning. Annual Review of Psychology,  49(1), 345–375.