The Advent of AI

Is it the Opportunity Democracies have been Craving or will it Destroy the Entire System?

The development of artificial intelligence (AI) has ushered in large-scale transformation across various sectors worldwide including finance, education, and journalism.[1] Nonetheless, its rapid development has raised concerns that AI poses a real existential threat to humanity. This threat is neither imminent, nor entirely plausible.[2] However, artificial intelligence does present a distinct and current threat to democracy. AI possesses the potential to undermine the entire democratic system, especially through the uncontrolled dissemination of misinformation and disinformation. Yet, AI has now become deeply ingrained in society, and its influence on democracy and politics will only grow. Instead of fearing its growth, democratic governments must harness the potential of AI to enhance system. Simultaneously, governments must approach developments with caution and implement sufficient safeguards to mitigate potential harm. This essay identifies the dangers and potential opportunities for democracies while also listing available solutions to mitigate these threats.

The Threats to the Public Sentiment Interpretation

Generative AI threatens the three central pillars of democracy: representation, accountability and – arguably most important – trust.[3] For a democratic government to function, constituents must be able to trust their government. They should have the ability to obtain information about government activities from a reliable source and trust that the government is giving due regard to their concerns and preferences. According to Robert Dahl, in 1972[4] democracy also requires “the continued responsiveness of the government to the preferences of its citizens”. However, with the advent of AI, it is becoming far more challenging for authorities to grasp constituent sentiment. This new technology now has the capability to generate false constituent sentiment at scale by creating messages taking any position with remarkable uniqueness.[5] A field experiment conducted in the US in 2020 demonstrated that when state legislators received a number of letters written by both AI and humans, they were unable to discern the difference between the two.[6] This also indicates that there is an opportunity for malicious actors to successfully conduct astroturf campaigns, where a skewed or biased view of public opinion is presented as a genuine, grassroots movement,[7] which could then be used to influence legislators. For government officials, this completely undermines their efforts to understand public sentiment, which in turn threatens the accuracy of democratic representation. This could have huge repercussions on the legislative process, rendering almost obsolete the public-comment process, as it could lead to malicious actors distorting legislators’ perceptions of citizens’ feelings as well as which issues are most important to them.[8]

The Threat of Disinformation and Misinformation

Concurrently, disinformation and misinformation can erode the public’s trust in their government. The European Commission’s strategy ‘Shaping Europe’s Digital Future’ acknowledges that “disinformation erodes trust in institutions along with digital and traditional media and harms our democracies by hampering the ability of citizens to take informed decisions”.[9] AI equips malicious actors with numerous techniques which enable them to promote the dissemination of disinformation. One such method is through micro-targeting. Experts warn that “AI will make it far easier for malicious actors and legitimate advertisers alike to track user behaviour online, identify potential new users to target, and collect information about users’ attitudes, beliefs, and preferences”.[10] This could also be used to reinforce certain biases. What is even more concerning is the rise of “deep fakes”. These are video and audio spoofing tools that allow malevolent agitators to manipulate and mimic voice and imagery of any person. A popular example of their use was the mimicking of Carrie Fisher in one of the movies in the popular franchise ‘Star Wars’.[11] Deep fakes have become increasingly accurate in the last number of years, evolving from mere jokes into scams. There has already been evidence of their successful application to deceive individuals. For instance, a CEO of UK based energy plant fell prey to a scammer who used deepfake technology.[12] The scammer imitated the voice of the CEO’s partner company leader to request a wire transfer of €200,000. The CEO sent the money and was unable to receive any compensation upon discovering the deception.[13] These deep fakes are now even making their way into politics. In Belgium, former Prime minister, Sophie Wilmès, was targeted by an activist group – Extinction Rebellion –[14] in mid 2020. The group posted a video imitating her giving a speech on the links of COVID-19 and climate change. Despite the group’s efforts to make it clear that the speech was fake, some people still believed that it was real. The implications of these incidents are grave, particularly for the future of trust in politics. For voters, this also threatens their ability to hold elected officials accountable for their actions, contributing to an erosion of democratic accountability. Any politician could deny their own wrongdoing declaring it a deep fake.

Opportunities for Democracy

Yet, despite the threat it poses to democracy, artificial intelligence could become a huge asset for democracies. There is huge potential to harness AI’s capabilities to enhance fairness within democratic societies. Firstly, AI could play a role in informing citizens about various policy issues, position stances and other aspects of democratic life through a public large language model (LLM).[15] More ambitiously, some experts proposed that AI has the potential to enable widespread radical democracy.[16] They suggest the creation of massive political chat rooms that could function as town hall meetings, allowing people to deliberate on a wide range of issues that might not have had the opportunity to be discussed otherwise. AI could then automatically summarize participant opinions, moderate debates and facilitate consensus building.[17] This would provide politicians with a better understanding of their constituents’ views, especially when combined with human expertise.[18] The potential of AI to evaluate public sentiment could also benefit the government, particularly policymakers. They could harness the abilities of AI to enrich and expediate policy making, from identifying issues to evaluating previously implemented policies.[19] One study indicates that the use of artificial intelligence systems in this way does not undermine citizens’ trust, provided there is clear disclosure of the use of AI as well as a guarantee of human oversight.[20]

Mitigating the Threats

To harness the full potential of AI, certain safeguards must be put in place to protect the privacy and freedom of individuals, as well as the integrity of democracy. One such method would be to train the same programs that generate AI content to detect AI-generated content. The neural networks used in AI for text generation “know” the types of language, words and sentence structure used in creating artificial content.[21] Therefore, they can be used to detect the same hallmarks and patterns in a piece of work and determine whether it was created by artificial intelligence or by humans.[22] On the other hand, the proliferation of disinformation and misinformation often relies on people’s inability to distinguish lies from the truth.[23] Thus, the key to combatting disinformation and misinformation is to have an informed public. In Germany, there is a list of possible deep fake indicators available online on their Federal Office for Information security.[24] Other countries should follow suit and ensure their own citizens are aware and informed. Additionally, there are laws and regulations being implemented to address the dangers posed by AI. In the EU, the European Commission's draft regulation on AI systems comprises numerous regulations, including a requirement to label all content generated using deep fake technology.[25] It also imposes specific regulations on high-risk AI software, and this list includes software used to influence voters in political campaigns. The regulation of AI is still in its “early days” in the US, but the Biden administration has been “fast-tracking” talks with various AI companies, civil society groups and academics.[26]

To conclude, the rapid proliferation of AI presents a clear and distinct threat to democracies around the world. It has permeated many sectors of life encroaching on the realm of politics, contributing to the spread of misinformation and disinformation, which in turn fuels mistrust, doubt and fear. Yet, researchers have discovered a number of ways in which AI can improve the landscape of politics and policy making. For a path forward, governments must enforce the necessary regulations and ensure citizens are properly informed about the potential threats of AI and how to protect themselves. With the right balance of regulation and awareness, democracy has the potential not only to be preserved, but enhanced, ultimately leading to an improved society as a whole.

Lily O’Keeffe

[1] Kreps, S. and Kriner, D. (2023) ‘How AI threatens democracy’, Journal of Democracy, 34(4), pp. 122–131. doi:10.1353/jod.2023.a907693. [2] ibid [3] ibid [4] ibid [5] ibid [6] ibid [7] Adam, M. and Hocquard, C. (2023) Artificial Intelligence, democracy and elections, European Parliament Briefing. Available at: https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/751478/EPRS_BRI(2023)751478_EN.pdf (Accessed: 05 November 2023). [8] Kreps, S. and Kriner, D. (2023) ‘How AI threatens democracy’, Journal of Democracy, 34(4), pp. 122–131. doi:10.1353/jod.2023.a907693. [9] COLOMINA, C., SÁNCHEZ MARGALEF, H. and YOUNGS, R. (no date) The impact of disinformation on Democratic ... - european parliament. Available at: https://www.europarl.europa.eu/RegData/etudes/STUD/2021/653635/EXPO_STU(2021)653635_EN.pdf (Accessed: 05 November 2023). [10] Arsenault, A. (2020) Microtargeting, automation, and forgery: Disinformation in the age of ..., uO Research. Available at: https://ruor.uottawa.ca/bitstream/10393/40495/1/ARSENAULT%2C%20Amelia_20201%20-%20WEB.pdf (Accessed: 05 November 2023). [11] Ray, A. (2021) ‘Disinformation, Deepfakes and democracies: The need for legislative reform’, University of New South Wales Law Journal, 44(3). doi:10.53637/dels2700. [12]Quirk, C. (2023) The High Stakes of Deepfakes: The Growing Necessity of Federal Legislation to Regulate This Rapidly Evolving Technology, Princeton University. Available at: https://legaljournal.princeton.edu/the-high-stakes-of-deepfakes-the-growing-necessity-of-federal-legislation-to-regulate-this-rapidly-evolving-technology/ (Accessed: 05 November 2023). [13] ibid [14] Ray, A. (2021) ‘Disinformation, Deepfakes and democracies: The need for legislative reform’, University of New South Wales Law Journal, 44(3). doi:10.53637/dels2700. [15] Schneier, B., Farrell, H. and Sanders, N.E. (2023) The tech investment we should make now to avoid A.I. disaster, Slate Magazine. Available at: https://slate.com/technology/2023/04/ai-public-option.html#:%7E:text=It%20could%20plausibly%20educate%20citizens%2C%20help%20them%20deliberate%20together%2C%20summarize%20what%20they%20think%2C%20and%20find%20possible%20common%20ground. (Accessed: 05 November 2023). [16] Adam, M. and Hocquard, C. (2023) Artificial Intelligence, democracy and elections, European Parliament Briefing. Available at: https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/751478/EPRS_BRI(2023)751478_EN.pdf (Accessed: 05 November 2023). [17] ibid [18] ibid [19] Wiczew, D. (2023) Conversing with ai: How llms are transforming governments and organizations, Medium. Available at: https://medium.datadriveninvestor.com/conversing-with-ai-how-llms-are-transforming-governments-and-organizations-aaa5d019297d (Accessed: 05 November 2023). [20] Bender, S. (2022) ‘Algorithmic elections’, Michigan Law Review, (121.3), p. 489. doi:10.36644/mlr.121.3.algorithmic. [21] Kreps, S. and Kriner, D. (2023) ‘How AI threatens democracy’, Journal of Democracy, 34(4), pp. 122–131. doi:10.1353/jod.2023.a907693. [22] ibid [23] Krimmer, R., Rabitsch, A. and Kužel, R. (2022) Elections in Digital Times, UNESCO. Available at: https://unesdoc.unesco.org/in/documentViewer.xhtml?v=2.1.196&id=p%3A%3Ausmarcdef_0000382102&file=%2Fin%2Frest%2FannotationSVC%2FDownloadWatermarkedAttachment%2Fattach_import_72f09c8b-5b8e-436a-8951-dac7ce4ebfc9%3F_%3D382102eng.pdf&locale=en&multi=true&ark=%2Fark%3A%2F48223%2Fpf0000382102%2FPDF%2F382102eng.pdf#Elections-in-Digital-Times_WEB_20230522.indd%3A.39628%3A9490 (Accessed: 05 November 2023). [24] Deep fakes – threats and countermeasures (2022) Federal Office for Information Security. Available at: https://www.bsi.bund.de/EN/Themen/Unternehmen-und-Organisationen/Informationen-und-Empfehlungen/Kuenstliche-Intelligenz/Deepfakes/deepfakes_node.html#doc1011620bodyText6 (Accessed: 05 November 2023). [25] Adam, M. and Hocquard, C. (2023) Artificial Intelligence, democracy and elections, European Parliament Briefing. Available at: https://www.europarl.europa.eu/RegData/etudes/BRIE/2023/751478/EPRS_BRI(2023)751478_EN.pdf (Accessed: 05 November 2023). [26] Kang, C. (2023) A.I. regulation is in its ‘early days’ - the new york times, New York Times. Available at: https://www.nytimes.com/2023/07/21/technology/ai-united-states-regulation.html (Accessed: 05 November 2023).