Assessing the Threat posed by Artificial Intelligence and Computational Propaganda

Marijn Pronk is a Master Student at the University of Glasgow, focusing on identity politics, propaganda, and technology. Currently Marijn is finishing her dissertation on the use of populist propagandic tactics of the Far-Right online. She can be found on Twitter @marijnpronk9. Divergent Options’ content does not contain information of an official nature nor does the content represent the official position of any government, any organization, or any group.


Title:  Assessing the Threat posed by Artificial Intelligence and Computational Propaganda

Date Originally Written:  April 1, 2020.

Date Originally Published:  May 18, 2020.

Author and / or Article Point of View:  The Author is a Master Student in Security, Intelligence, and Strategic Studies at the University of Glasgow. The Author believes that a nuanced perspective towards the influence of Artificial Intelligence (AI) on technical communication services is paramount to understanding its threat.

Summary: 
 AI has greatly impacted communication technology worldwide. Computational propaganda is an example of the unregulated use of AI weaponized for malign political purposes. Changing online realities through botnets which creates a distortion of online environments could affect voter’s health, and democracies’ ability to function. However, this type of AI is currently limited to Big Tech companies and governmental powers.

Text:  
A cornerstone of the democratic political structure is media; an unbiased, uncensored, and unaltered flow of information is paramount to continue the health of the democratic process. In a fluctuating political environment, digital spaces and technologies offer great platforms for political action and civic engagement[1]. Currently, more people use Facebook as their main source of news than via any news organization[2]. Therefore, manipulating the flow of information in the digital sphere could not only pose as a great threat to the democratic values that the internet was founded upon, but also the health of democracies worldwide. Imagine a world where those pillars of democracy can be artificially altered, where people can manipulate the digital information sphere; from the content to the exposure range of information. In this scenario, one would be unable to distinguish real from fake, making critical perspectives obsolete. One practical embodiment of this phenomenon is computational propaganda, which describes the process of digital misinformation and manipulation of public opinion via the internet[3]. Generally, these practices range from the fabrication of messages, the artificial amplification of certain information, to the highly influential use of botnets (a network of software applications programmed to do certain tasks). With the emergence of AI, computational propaganda could be enhanced, and the outcomes can become qualitatively better and more difficult to spot.

Computational propaganda is defined as ‘’the assemblage of social media platforms, autonomous agents, algorithms, and big data tasked with manipulating public opinion[3].‘’ AI has the power to enhance computational propaganda in various ways, such as increased amplification and reach of political disinformation through bots. However, qualitatively AI can also increase the sophistication and the automation quality of bots. AI already plays an intrinsic role in the gathering process, being used in datamining of individuals’ online activity and monitoring and processing of large volumes of online data. Datamining combines tools from AI and statistics to recognize useful patterns and handle large datasets[4]. These technologies and databases are often grounded in in the digital advertising industry. With the help of AI, data collection can be done more targeted and thus more efficiently.

Concerning the malicious use of these techniques in the realm of computational propaganda, these improvements of AI can enhance ‘’[..] the processes that enable the creation of more persuasive manipulations of visual imagery, and enabling disinformation campaigns that can be targeted and personalized much more efficiently[4].’’ Botnets are still relatively reliant on human input for the political messages, but AI can also improve the capabilities of the bots interacting with humans online, making them seem more credible. Though the self-learning capabilities of some chat bots are relatively rudimentary, improved automation through computational propaganda tools aided by AI could be a powerful tool to influence public opinion. The self-learning aspect of AI-powered bots and the increasing volume of data that can be used for training, gives rise for concern. ‘’[..] advances in deep and machine learning, natural language understanding, big data processing, reinforcement learning, and computer vision algorithms are paving the way for the rise in AI-powered bots, that are faster, getting better at understanding human interaction and can even mimic human behaviour[5].’’ With this improved automation and data gathering power, computational propaganda tools aided by AI could act more precise by affecting the data gathering process quantitatively and qualitatively. Consequently, this hyper-specialized data and the increasing credibility of bots online due to increasing contextual understanding can greatly enhance the capabilities and effects of computational propaganda.

However, relativizing AI capabilities should be considered in three areas: data, the power of the AI, and the quality of the output. Starting with AI and data, technical knowledge is necessary in order to work with those massive databases used for audience targeting[6]. This quality of AI is within the capabilities of a nation-state or big corporations, but still stays out of reach for the masses[7]. Secondly, the level of entrenchment and strength of AI will determine its final capabilities. One must differ between ‘narrow’ and ‘strong’ AI to consider the possible threat to society. Narrow AI is simply rule based, meaning that you have the data running through multiple levels coded with algorithmic rules, for the AI to come to a decision. Strong AI means that the rules-model can learn from the data, and can adapt this set of pre-programmed of rules itself, without interference of humans (this is called ‘Artificial General Intelligence’). Currently, such strong AI is still a concept of the future. Human labour still creates the content for the bots to distribute, simply because the AI power is not strong enough to think outside the pre-programmed box of rules, and therefore cannot (yet) create their own content solely based on the data fed to the model[7]. So, computational propaganda is dependent on narrow AI, which requires a relatively large amount of high-quality data to yield accurate results. Deviating from this programmed path or task severely affects its effectiveness[8]. Thirdly, the output or the produced propaganda by the computational propaganda tools vary greatly in quality. The real danger lies in the quantity of information that botnets can spread. Regarding the chatbots, which are supposed to be high quality and indistinguishable from humans, these models often fail tests when tried outside their training data environments.

To address this emerging threat, policy changes across the media ecosystem are happening to mitigate the effects of disinformation[9]. Secondly, recently researchers have investigated the possibility of AI assisting in combating falsehoods and bots online[10]. One proposal is to build automated and semi-automated systems on the web, purposed for fact-checking and content analysis. Eventually, these bottom-top solutions will considerably help counter the effects of computational propaganda. Thirdly, the influence that Big Tech companies have on these issues cannot be negated, and their accountability towards creation and possible power of mitigation of these problems will be considered. Top-to-bottom co-operation between states and the public will be paramount. ‘’The technologies of precision propaganda do not distinguish between commerce and politics. But democracies do[11].’


Endnotes:

[1] Vaccari, C. (2017). Online Mobilization in Comparative Perspective: Digital Appeals and Political Engagement in Germany, Italy, and the United Kingdom. Political Communication, 34(1), pp. 69-88. doi:10.1080/10584609.2016.1201558

[2] Majo-Vazquez, S., & González-Bailón, S. (2018). Digital News and the Consumption of Political Information. In G. M. Forthcoming, & W. H. Dutton, Society and the Internet. How Networks of Information and Communication are Changing Our Lives (pp. 1-12). Oxford: Oxford University Press. doi:10.2139/ssrn.3351334

[3] Woolley, S. C., & Howard, P. N. (2018). Introduction: Computational Propaganda Worldwide. In S. C. Woolley, & P. N. Howard, Computational Propaganda: Political Parties, Politicians, and Political Manipulation on Social Media (pp. 1-18). Oxford: Oxford University Press. doi:10.1093/oso/9780190931407.003.0001

[4] Wardle, C. (2018, July 6). Information Disorder: The Essential Glossary. Retrieved December 4, 2019, from First Draft News: https://firstdraftnews.org/latest/infodisorder-definitional-toolbox

[5] Dutt, D. (2018, April 2). Reducing the impact of AI-powered bot attacks. CSO. Retrieved December 5, 2019, from https://www.csoonline.com/article/3267828/reducing-the-impact-of-ai-powered-bot-attacks.html

[6] Bolsover, G., & Howard, P. (2017). Computational Propaganda and Political Big Data: Moving Toward a More Critical Research Agenda. Big Data, 5(4), pp. 273–276. doi:10.1089/big.2017.29024.cpr

[7] Chessen, M. (2017). The MADCOM Future: how artificial intelligence will enhance computational propaganda, reprogram human culture, and threaten democracy… and what can be done about it. Washington DC: The Atlantic Council of the United States. Retrieved December 4, 2019

[8] Davidson, L. (2019, August 12). Narrow vs. General AI: What’s Next for Artificial Intelligence? Retrieved December 11, 2019, from Springboard: https://www.springboard.com/blog/narrow-vs-general-ai

[9] Hassan, N., Li, C., Yang, J., & Yu, C. (2019, July). Introduction to the Special Issue on Combating Digital Misinformation and Disinformation. ACM Journal of Data and Information Quality, 11(3), 1-3. Retrieved December 11, 2019

[10] Woolley, S., & Guilbeault, D. (2017). Computational Propaganda in the United States of America: Manufactoring Consensus Online. Oxford, UK: Project on Computational Propaganda. Retrieved December 5, 2019

[11] Ghosh, D., & Scott, B. (2018, January). #DigitalDeceit: The Technologies Behind Precision Propaganda on the Internet. Retrieved December 11, 2019, from New America: https://www.newamerica.org/public-interest-technology/policy-papers/digitaldeceit

Artificial Intelligence & Human-Machine Teaming Assessment Papers Cyberspace Emerging Technology Influence Operations Marijn Pronk