Written evidence submitted by Josephine Marlow (SMH0012)
Inquiry: Social media, misinformation and harmful algorithms
In my submission I refer to the following terms of reference:
[2.1] As a university politics student, I have encountered a vast range of fascinating topics during my study – however above all I have discovered a sustained interest in the rise of extreme right-wing groups and how the discourse such groups spread translates into violence, as seen in the summer 2024 riots. This forms the basis of my interest in submitting evidence to this inquiry.
[2.2] Although I lack a background in computer science, and therefore cannot share insight into the scientific mechanisms of algorithms; I am hopeful that the knowledge I have gained through social and political science will produce strong theoretical conclusions and recommendations in line with the goals of this inquiry.
[3.1] To understand the role algorithms played in the summer riots, we must first determine the ways algorithms operate on social media platforms and the social harms they have the potential to cause. Social media algorithms are essentially a set of encoded rules that determine how people interact with technology – these rules transform into assumptions about the types of content individual users want to consume, this process produces social theories (Forberg 2022, 294). These theories can then potentially transpire into weapons with the intent to push prejudiced views of minorities or incite radicalised political views if left unchecked by the bodies responsible for their programming (295).
[3.2] Patterns of such weaponisation have been studied under the context of Google’s algorithmic search engine; Nobel’s work on ‘technological redlining’ challenges society’s common but false view that algorithms are ‘neutral’ and ‘objective’, and instead encourages us to consider the humans within the tech industry that facilitate racist/sexist search results (2018, 1). Poor representation of marginal groups in algorithmically driven search results allows prejudiced attitudes to become embedded in the architecture of technology (9), thus reinforcing oppressive attitudes.
[3.3] In defence of algorithms, we may consider the role ‘biased users’ play in spreading false information. A study on fake news on social media sites investigated the rolls of ‘filter bubbles’ and ‘echo chambers’ – filter bubbles are attributed to the personalisation of content that algorithms construct for users, whereas echo chambers focus on a user’s placement in an online group or community which only shares information which coincides or amplifies their pre-existing beliefs (Zimmer et al 2019, 42). The research found that social media algorithms did not independently form communities where fake news was spread, instead the transmission of fake news into social media and beyond concerns itself mainly with individual users and the echo chamber communities that they interact with (49). This interpretation aids our understanding of the issue in that it recognises that responsibility for misinformation lies with both algorithmic faults and human bias.
[3.4] We now turn to a brief explanation of how algorithms have benefitted far-right groups. In Daniels’s case study of American white nationalism, we see how technological innovation has been historically exploited by alt-right groups to push racist and nationalist views into the mainstream media and public discourse (2018, 62). The infrastructure of social media itself is beneficial in this mission as it has a ‘many-to-many’ sharing environment as opposed to the ‘one-to-many’ model used in broadcasting or printing (64). When these factors are instrumentalised, algorithms have the power to disperse fake news and alt-right perspectives into mainstream avenues, expanding the scope of the issue.
[3.5] Due to the contemporary lack of existing scholarship on the 2024 riots, to determine the role algorithms played we must draw on the information gathered above and previous events. Research into the murder of Keira, a 14 year-old German girl, whose death produced vast amounts of misinformation circulating in far-right German groups reveals that the summer riots are not the first instance of a murder proliferating far-right perspectives. Immediately after news of her murder broke in 2018, speculation begun on the origins of the perpetrator with many right-wing actors publishing false assumptions that he had foreign origins or a migratory background (Darmstadt, Prinz & Saal 2019, 155). Targeted misinformation, spread through the mechanisms of algorithms detailed above, serves a purpose in weaving tragic empirical events into the political agendas of discriminatory groups (158).
[3.6] We may note here that speculation was driven via the German police’s refusal to name racial or religious characteristics of the perpetrator after his arrest; the department has guidelines in place recommending that identifying aspects such as race are not released publicly to avoid discriminatory attitudes from growing on the basis of a criminal’s individual wrongdoing (159). Here we see that these guidelines seemed to have the opposite effect and instead encouraged further speculation and misinformation – this is not to say that such guidelines are inappropriate, rather that they demonstrate the limitations of policy in emotional cases such as the murder of a young girl.
[3.7] The event that incited the 2024 riots bares similarities to the murder of Keira in Germany. In July 2024, 3 children were fatally stabbed in Southport – the tragic event sparked floods of misinformation on social media regarding the religion and citizenship status of the perpetrator (Choonara 2024, 3). Alike to the German case, we see a pattern emerge between tragedy and immediate suspicion and villainization of racial/religious minorities, incited through social media platforms. We can assume that algorithmic faults encouraging racism, filter bubbles and user constructed echo chambers contributed to the spread of misinformation which ultimately resulted in the mass mobilisation and violence performed by right-wing protest groups during the UK riots. The theoretical assumptions outlined here form the basis of my subsequent recommendations.
[5.1] We now turn away from an academic research driven basis of evidence and centre our investigation more towards policy. This section will focus on the parliamentary Online Safety Bill (OSB), which reached royal assent in October 2023 thus making it an Act. Both the key features of this legislation as well as areas in which it could have benefitted from more attention are explored to evaluate the legislation’s efficacy.
[5.2] In order to identify the key features of the Bill, and their significance – we must investigate the Bill’s journey through parliament. Three fundamental aspects of the legislation include (1) better protection of children online, (2) protection of users’ rights to freedom of expression and (3) transparency and accountability of service providers (Online Safety Act 2023). Freedom of expression has played a particularly important role in debates on the responsibilities of the Bill; in early debates the new ‘legal but harmful’ categorisation of content was scrutinised on the basis that it may hinder freedom of speech on user-to-user sites. It was replaced instead with a focus on ‘user empowerment’, requiring services to provide functionalities giving individuals more control over the types of potentially harmful content they engaged with – including hate speech (HC Deb 29 November 2022).
[5.3] However, in later debates, the removal of this content category sparked concerns that it may lessen the ability of Ofcom to predict future harms in that it reduces the need for rigorous risk assessments if the responsibility of content consumption is placed partially on the user (HL Deb 19 April 2023). Due to the rapidly progressing nature of social media platforms, the foreseeability of future problems is essential in maximising the efficacy of the Bill, as new legislation often takes a long time before coming into force – this suggests that the decision to draw attention away from the ‘legal but harmful’ categorisation may have been an oversight.
[5.4] Another area in which amendments were made during the journey of the Bill relates to the categorisation of user-to-user services. Previously, the strategy had functioned on a size-based approach where the top categories comprised the biggest platforms in use in the UK. After calls from the opposing party to update this, the strategy has changed in that it now considers categorisation on the grounds of potential risk as well as size. This is important as it makes it harder for smaller, but more saturated, sites such as ‘4chan’ (which regularly platforms far-right and racist content) from escaping the scope of the legislation (HC Deb 12 September 2023). Increased regulation of services based on risk plays an essential role in lessening the impact of online hate speech and groups, and aids in preventing such attitudes from translating into real world events.
[5.5] It is important also to consider the areas of the Bill which may lack sufficient attention and would benefit from being returned to in future debates and legislation. One interesting detail raised in the House of Lords calls into question how the business models of social media platforms themselves may produce harms – the logic of the argument broadly follows arguments made previously in this work detailing how algorithms create personalisation processes that increase user engagement and groups people with common interests together, to increase such engagement. However, here we explore the commercial incentive towards this process – the more engagement a platform has means the more opportunity to increase revenue through advertisements. Personalisation may become harmful in that it can push users into constructed environments where hate speech and extremism is normalised, thus the concern calls for the importance of design features that favour safety beyond engagement (HL Deb 19 April 2023).
[5.6] Another area of the issue which seems to suffer from a lack of attention in the legislation is provisions for misinformation. As raised in parliament; misinformation, among other harms, presents a challenge to democratic processes and therefore must be actioned quickly to prevent future harms. At present, the legislation does not seem to layout concrete guidance concerning misinformation for companies or users and so this part of the Bill is notably week in comparison to other areas (HC Deb 12 September 2023). Given the role misinformation played in the summer riots this year – we see undoubtably how important such provisions are and the contemporary lack of them may undermine the Bill’s efficiency overall to keep people, particularly minority groups, safe.
[5.7] In the original White Papers of the Bill, the rationale for creating internet safety legislation was presented. Two factors that appeared in this rationale included online disinformation and online manipulation. The first was evidenced by a large segment of the public wanting the government to increase measures to differentiate between genuine and fake news. The second signals the government’s feelings that the recent normalisation of AI software could increase the strength and commonality of psychological manipulation (Trengove et al 2022, 4). As demonstrated through the previous paragraphs, both issues seem to have been overshadowed by other focusses within the Bill, all of which remain important. This may signify the need for further legislation that has a more specific focus on the system design of social media platforms in order to maximise efficiency in provisions created to prevent online events translating into real world violence, as they did in the 2024 summer riots.
[Wordcount, excluding citations: 2,182]
16 December 2024
Bibliography
Choonara, J. (2024) Confronting Britain’s far-right problem. International Socialism, 2(184). Pp. 3-22. [Online] Available at: https://www.researchgate.net/publication/384628347_Confronting_Britain's_far-right_problem [Accessed: 02/12/2024]
Daniels, J. (2018) The algorithmic rise of the “alt-right”. Contexts, 17(1). Pp. 60-65. [Online] Available at: https://journals.sagepub.com/doi/full/10.1177/1536504218766547 [Accessed: 04/12/2024]
Darmstadt, A., Prinz, M. & Saal, O. (2019) ‘The Murder of Keira, Misinformation and Hate Speech as Far-Right Online Strategies’ in Fielitz, M. & Thurston, N. (eds) Post-Digital Cultures of the Far Right: Online Actions and Offline Consequences in Europe and the US. Bielefeld: Transcript Publishing. Pp. 155-167.
Dawes, M. (2024) Dame Melanie Dawes to Secretary of State, 22 October. [Letter]. [Online] Available at: https://www.ofcom.org.uk/siteassets/resources/documents/about-ofcom/public-correspondence/2024/letter-from-dame-melanie-dawes-to-the-secretary-of-state-22-october-2024.pdf?v=383693 [Accessed: 05/12/2024]
Forberg, P.L. (2022) From the Fringe to the Fore: An Algorithmic Ethnography of the Far-Right Conspiracy Group QAnon. Journal of Contemporary Ethnography, 51(3). Pp. 291-317. [Online] Available at: https://journals.sagepub.com/doi/full/10.1177/08912416211040560 [Accessed: 02/12/2024]
Nobel, S.U. (2018) ‘Introduction: The Power of Algorithms’ in Algorithms of Oppression: How Search Engines Reinforce Racism. New York: New York University Press. Pp. 1-14.
Online Safety Act 2023, c. 50. [Online] Available at: https://www.legislation.gov.uk/ukpga/2023/50/2024-08-23 [Accessed: 05/12/2024]
HC Deb (29 November 2022). Vol. 723, cols. 31WS-34WS. [Online] Available at: https://hansard.parliament.uk/Commons/2022-11-29/debates/22112948000012/OnlineSafetyBill?highlight=social%20media%20algorithms#contribution-46A51465-BD0A-403B-9CF8-6A4D0397ABC7 [Accessed: 04/12/2024]
HC Deb (12 September 2023). Vol. 737, cols. 799-846. [Online] Available at: https://hansard.parliament.uk/commons/2023-09-12/debates/81853BB7-375E-45C0-8C9D-4169AC36DD12/OnlineSafetyBill#main-content [Accessed: 04/12/2024]
HL Deb (19 April 2023). Vol. 829, cols. 699-728. [Online] Available at: https://hansard.parliament.uk/lords/2023-04-19/debates/0C27A2E7-484D-46A2-A68C-4C6B9FBA671E/OnlineSafetyBill#main-content [Accessed: 04/12/2024]
Trengove, M. et al. (2022) A critical review of the Online Safety Bill. Patterns, 3(8). Pp. 1-10. [Online] Available at: https://www.cell.com/action/showPdf?pii=S2666-3899%2822%2900147-7 [Accessed: 03/12/2024]
Zimmer, F. et al. (2019) Fake News in Social Media: Bad Algorithms or Biased Users? Journal of Information Science Theory and Practice, 7(2). Pp. 40-53. [Online] Available at: https://www.ling.hhu.de/fileadmin/redaktion/Fakultaeten/Philosophische_Fakultaet/Sprache_und_Information/Informationswissenschaft/Dateien/Franziska_Zimmer/2019_JISTaP_7_2__Zimmer_Scheibe_StockM_StockWG.pdf [Accessed: 02/12/2024]