Written evidence submitted by Hossein Dabbagh (PhD) (SMH0017)

Introduction

I am Hossein Dabbagh, an Assistant Professor of Philosophy at Northeastern University London and a Philosophy Tutor at the University of Oxford. My academic work focuses on practical ethics and public policy, with a particular interest in the intersection of technology and society. I investigate the ethical and political dimensions of digital platforms, especially their role in exacerbating societal harms. This submission responds to the Committee’s inquiry into the role of social media algorithms and generative artificial intelligence (AI) in spreading harmful content, with specific reference to incidents such as the Southport riots in 2024.

This document examines how platform business models, algorithmic design, and generative AI contribute to the spread of misinformation. It evaluates the effectiveness of current regulatory frameworks and offers evidence-based recommendations for improvement.

Business Models and Social Harms

Social media platforms operate within an economic framework that commodifies user attention, with algorithms designed to maximise engagement through content curation. This profit-driven model incentivises the proliferation of emotionally charged material, often at the expense of truth, likely to provoke strong emotional responses, such as outrage, fear, or anger. Research consistently shows that content with these emotional triggers spreads faster and further than more neutral or factual information (Cinelli et al., 2021; Vosoughi, et al., 2018). Such practices can undermine the public sphere where rational discourse is central to democratic life (Habermas, 1991). The spread of falsehoods—motivated by platform algorithms rather than public reason—is inimical to this ideal, furthering polarisation and societal fragmentation.

This dynamic creates an ecosystem where divisive and misleading content thrives. The Southport riots illustrate the risks of such a system. False claims, including those about the attacker being a Muslim asylum seeker, gained rapid traction on social media, fuelling anti-immigration sentiments and inciting violence. These incidents demonstrate how platform incentives can align with harmful outcomes, as algorithms prioritise reach and engagement over accuracy and social harmony. Additionally, this reflects a structural tension between commercial incentives and the ethical responsibilities of platforms. Although the principle of free expression presupposes access to a marketplace of ideas where truth prevails through open debate, algorithmic amplification skews this marketplace, privileging incendiary content that undermines collective reason.

The Role of Algorithms in Content Amplification

Algorithms are not neutral tools; they encode normative values that influence what information is visible and prioritised. They are central to how content is ranked and disseminated on social media. These systems are designed to predict what content will keep users engaged and then amplify it accordingly. This amplification might disproportionately favour content that aligns with users' biases, reinforces existing beliefs, and generates heightened emotional responses (Bozdag, 2013). Designed to optimise engagement, these systems often exploit cognitive biases such as confirmation bias and negativity bias, leading to the creation of ‘echo chambers’ where users are predominantly exposed to viewpoints that align with their own while opposing perspectives are minimised or excluded (Mosleh, Martel & Rand, 2024). Such dynamics contribute to information cascades, where individuals adopt beliefs based on widespread dissemination rather than evidence (Sunstein, 2007). By amplifying partisan content, this dynamic has significant implications for political discourse, potentially facilitating the spread of right-wing extremism.

In the context of the Southport riots, algorithms amplified divisive narratives that capitalised on existing tensions, demonstrating the ethical failure of platforms to mitigate harm. By privileging virality over veracity, recommender systems acted as misinformation agents, undermining social cohesion. This demonstrates a fundamental flaw in algorithmic design: while optimised for engagement, these systems lack mechanisms to differentiate between beneficial and harmful content. This raises an ethical issue—platforms have built tools that magnify harm yet often deflect responsibility for the consequences—and invites us to ask about the moral agency of algorithmic systems: can platforms absolve themselves of responsibility for their tools’ outcomes? Drawing from Hannah Arendt’s (1963) analysis of systemic evils, one might argue that algorithmic harms arise from a failure to exercise moral judgment at the design stage, leading to the erosion of democratic norms.

Generative AI and the Proliferation of Misinformation

Generative AI technologies such as large language models (LLMs) present unique challenges to information integrity. These systems, capable of producing realistic but fabricated narratives, amplify the epistemic crisis of the digital age (Zellers et al., 2019). Generative AI technologies introduce a new and significant risk by enabling the rapid production of false information. These tools can create convincing narratives, images, and videos that are difficult to distinguish from genuine content. Their scalability makes them particularly dangerous during crises when misinformation can escalate tensions and incite violence. This issue raises a critical philosophical concern about truthfulness as a precondition for trust in public discourse (Williams, 2002). By eroding the ability to distinguish between truth and fabrication, AI-generated content undermines this essential trust as a basis for informed decision-making.

While no direct evidence links generative AI to the Southport riots, the potential for such tools to exacerbate misinformation crises is undeniable. The ability to create large volumes of misleading content at scale undermines public trust in democratic institutions and complicates moderating harmful narratives effectively. But perhaps more importantly, misinformation exacerbates feelings of alienation among marginalised communities, intensifying societal divisions. The convergence of misinformation with existing inequalities creates environments where individuals feel like strangers in their own land, eroding social cohesion and trust within communities. By lowering the cost of producing false narratives, generative AI risks entrenching a ‘post-truth’ culture, where ‘alternative facts’ dominate public debate (Gregory, 2023). This necessitates regulatory oversight to ensure AI is used responsibly, guided by principles of transparency, accountability, and ethical design.

Effectiveness of the UK’s Regulatory Framework

The Online Safety Act 2023 represents an important step toward addressing digital harms, imposing new duties on platforms to mitigate risks and remove illegal content. However, its effectiveness will depend on its implementation, the platforms’ commitment to compliance, and the adaptability of regulatory measures. Ofcom’s analysis of the Southport riots highlights several weaknesses in platform responses, including inadequate enforcement, revealing significant gaps in current practices.

One notable issue is the lack of consistent crisis protocols among platforms. While some companies acted swiftly to moderate harmful content, others allowed false narratives to persist for days, exacerbating the situation. This inconsistency undermines the Act’s potential to guarantee safety and stability during crises. Regulatory frameworks must balance competing values: the protection of free expression, the prevention of harm, and the promotion of public trust. One could argue that regulations should prioritise the most vulnerable to make sure that platforms do not exacerbate inequalities or marginalise minority voices (Rawls, 1971). This aligns with Ofcom’s recommendation for tailored measures to protect children and other at-risk groups from algorithmic harms.

Policy Recommendations for Improvement

To address the challenges posed by social media algorithms and generative AI, several measures are essential.

First, platforms must adopt greater algorithmic transparency, allowing regulators and researchers to audit the ethical implications of their systems (Bontridder & Poullet, 2021). Regulators should have access to conduct independent audits, ensuring that algorithms do not amplify harmful content. This transparency would also enable a better public understanding of how content is prioritised and moderated. This highlights the necessity of aligning technological practices and design with human values and societal norms (Gabriel, 2020; Nissenbaum, 2010).

Second, there is an urgent need for real-time moderation protocols. Platforms should establish dedicated teams and partnerships with civil society organisations and law enforcement enabling rapid responses to harmful content during emergencies. These teams must have the resources and training to identify and address harmful content effectively (Gongane, Munot & Anuse, 2022).

Third, media and AI literacy initiatives must be scaled up to foster critical engagement with online content. Such initiatives should focus on teaching users how to identify misinformation, recognise manipulative narratives, and report harmful content (Caled & Silva, 2022). To implement, media literacy and AI ethics education must be introduced early and integrated into educational curricula. This approach equips individuals, particularly young people, with the critical thinking skills needed to navigate the complexities of digital media and AI technologies (Dabbagh et al., 2024). This shift not only empowers individuals but also creates a foundation for a more informed and resilient society capable of countering the systemic spread of harmful content.

Finally, regulatory frameworks must address the unique risks posed by generative AI. This includes establishing guidelines for the ethical deployment of AI tools and holding platforms accountable for the misuse of their technologies such as for AI-generated misinformation that circulates on their services (Novelli, Taddeo & Floridi, 2024). The aim should be to make sure that technological innovation does not come at the expense of societal well-being.

Accountability and Governance

Accountability for the spread of misinformation and harmful content must extend beyond platforms to include the broader ecosystem of stakeholders and societal structures. Ofcom and the National Security Online Information Team play vital roles in overseeing compliance, but their efforts must be supported by independent oversight bodies and collaborative networks of stakeholders, involving civil society, academia, and independent oversight bodies. This advocates the model of ‘shared responsibility’, where all actors contributing to structural harms are accountable for mitigating their effects (Young, 2006).

The Southport riots highlight the need for a holistic approach to governance, integrating regulatory measures with public education and ethical design practices. By promoting a culture of accountability, the UK can lead the way in addressing the ethical and societal challenges of the digital age, posed by social media and generative AI.

Conclusion

The inquiry into social media, misinformation, and harmful algorithms is both timely and essential. The Southport riots exemplify the profound risks posed by digital misinformation and the amplifying effects of algorithmic systems. These technologies, while transformative, carry significant ethical responsibilities that cannot be ignored. By addressing the gaps in current regulatory frameworks, promoting transparency and accountability, and encouraging public resilience through media and AI literacy, the UK can safeguard democratic values and social cohesion.

I hope this submission provides valuable insights to inform the Committee’s work and contribute to the development of effective evidence-based policies that address the root causes of misinformation and harmful content, creating a safer and more equitable digital future for all.

17 December 202

References

Arendt, H. (1963). Eichmann in Jerusalem: A Report on the Banality of Evil. Viking Press.

Bontridder N, Poullet Y. (2021). The role of artificial intelligence in disinformation. Data & Policy 3:e32.

Bozdag, E. (2013). Bias in algorithmic filtering and personalization. Ethics and Information Technology, 15(3), 209-227.

Caled, D., Silva, M.J. (2022). Digital media and misinformation: An outlook on multidisciplinary strategies against manipulation. J Comput Soc Sc 5, 123–159.

Cinelli, M. De Francisci Morales, D. Galeazzi, A. Quattrociocchi, W. Starnini, M. (2021). The echo chamber effect on social media, Proc. Natl. Acad. Sci. U.S.A. 118 (9) e2023301118.

Gabriel, I. (2020). Artificial Intelligence, Values, and Alignment. Minds & Machines 30, 411–437.

Gongane, V.U., Munot, M.V. & Anuse, A.D. (2022). Detection and moderation of detrimental content on social media platforms: current status and future directions. Soc. Netw. Anal. Min. 12, 129.

Habermas, J. (1991). The Structural Transformation of the Public Sphere: An Inquiry into a Category of Bourgeois Society. MIT Press.

Mosleh, M., Martel, C., & Rand, D. G. (2024). Psychological underpinnings of partisan bias in tie formation on social media. Journal of Experimental Psychology: General. Advance online publication.

Nissenbaum, H. (2010). Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford University Press.

Novelli, C., Taddeo, M. & Floridi, L. (2024). Accountability in artificial intelligence: what it is and how it works. AI & Society 39, 1871–1882.

Rawls, J. (1971). A Theory of Justice. Harvard University Press.

Sam G. (2023). Fortify the Truth: How to Defend Human Rights in an Age of Deepfakes and Generative AI, Journal of Human Rights Practice, Vol. 15, Issue 3, pp. 702–714.

Sunstein, C. R. (2007). Infotopia: How Many Minds Produce Knowledge. Oxford University Press.

Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science 359 (6380), 1146-1151.

Williams, B. (2002). Truth and Truthfulness: An Essay in Genealogy. Princeton University Press.

Young, I. M. (2006). Responsibility for Justice. Oxford University Press.

Zellers, R., Holtzman, A., Bisk, Y., Farhadi, A., & Choi, Y. (2019). Defending against neural fake news. Proceedings of the 33rd International Conference on Neural Information Processing Systems, Article 812, 9054–9065.