Prathm Juneja
DPhil Student
Prathm Juneja is a student on the DPhil in Social Data Science, and was previously a student on the MSc in Social Data Science.
Introduction
Political parties want to win elections. To win elections, it is essential to understand voters. To understand voters, as demonstrated by the successful presidential campaigns of Donald Trump and Barack Obama, political parties now rely on data and statistics.
The integration of data into politics has already transformed our politics by creating an insatiable demand for increasingly large amounts of high-quality data. Recent developments in Artificial Intelligence (AI) will further drive and reinforce this transformation into a data-driven political system.
At first, this might sound scary. Certainly, the world’s attention has been captured recently by articles that paint a doomsday picture for politics, driven by AI-enabled misinformation campaigns through the proliferation of “deepfakes” on social media. It’s important not to completely discount these risks, but there is evidence that many of these fears are “overblown”.
Though it is essential to take election security seriously, the current focus on hypothetical risks created by AI distracts from what is perhaps an even more important story: how AI is changing and transforming political parties themselves.
To understand this transformation, it is important to look behind AI at the data, infrastructure, and technical expertise driving its development and implementation. These three factors will be the primary drivers in the creation of a new political system that is more personal, but less human, and less public than ever before.
Data
Politically aligned data brokers already hold thousands of data points for any given individual voter in America; this incredible amount of information is used heavily by political parties. Parties rely on data to identify key issues, craft and validate their advertisement campaigns, improve voter turnout in key districts, and even build psychological profiles of voters.
It was the creation of psychological profiles that brought the political use of data truly into the public debate, a side effect of the Cambridge Analytica scandal. The public outrage that followed these revelations brought increased scrutiny of how campaigns obtain and use data, leading many private sector stakeholders to put limits on political data-gathering and targeted political ads.
Although certain forms of political advertisement may have been limited as a result, such restrictions have done little to effect the use of data in elections. As advanced statistical and AI-based models demand data in larger amounts, campaigns and data firms will be forced to gather even more information about voters. This will lead to a new era of personalized politics.
Infrastructure
To manage the collection and usage of data for elections, political parties formed new organizations to centralize capacity. They have created the necessary infrastructure to run data-heavy political campaigns.
This infrastructure works as follows. Campaigns will collect some of their own data (e.g., who donates, whose doors they knock on), but will store it in a politically aligned customer relationship software, and then connect it to, yet another, politically aligned data analytics platform.
These data are then further enhanced by combining them with additional external sources of data about voters, such as what one likes or dislikes or who your candidate preference may be. The result? An in-depth detailed voter file, where large amounts of personal data are linked to a specific individual’s voter information, including voting turnout history, that will be leveraged to influence who you vote for.
This is already happening today, but this infrastructure is not cheap, and a large volume of funding is required to ensure the sustainability of this political technical infrastructure. AI will strengthen these dependencies on external, and often privatised, infrastructure.
Technical Expertise
As the ways in which data are used by political campaigns to win elections grow, so too does the demand for technical expertise. Political parties will still require skilled campaigners, marketing teams, and volunteers to run a successful campaign, but, once again, AI is changing things.
To leverage cutting edge AI abilities, political parties are relying on externally provided talent, politically-aligned start-ups, and larger internal data teams more than ever before. There are already clear examples of this: in 2016 most of the major technology companies had employees embedded in both political campaigns, and there is evidence that in some cases these employees acted more like political operatives than advertising partners.
What is emerging is a potential future for politics that is not only about which party has the best candidate, but which campaign has been able to hire the best data scientists.
Conclusion
It is likely that the use of AI in American elections will, just like social media and big data in previous elections, mark a transition towards a new kind of politics.
Campaigns will be hyper-personal, driven by advances in AI that allow campaign staff to automatically tailor messages to specific groups. Elections will be more personal than ever before, but they will feel less human. If a campaign no longer needs one message to share with everyone, but rather a series of tailored messages to smaller groups, how important will the candidate’s own words be?
How campaigns are run, and who gets a say, will also change. As the importance of AI grows for political parties, they will require more computational power that is, today, only in the hands of a few companies, further entangling politics with business. Campaigns will also spend more on data than before, placing more power in the firms that curate such data.
Not all of these changes will happen in the 2024 election, and not all possible questions will be answered in the next few years. Yet, it is inevitable that AI will make its way into our elections – the potential advantages are too strong for campaigns to ignore.