When the UK voted to leave the EU and when Donald Trump became the 45th president of the United States last year, most polls were expecting a different result. One of the reasons mentioned by various journalists and by involved campaign teams for the unexpected results is the use of big data and artificial intelligence to influence people on social media. While these techniques were also used during the Obama campaigns in 2008 and 2012 (and in various other elections throughout the world), it is argued that these elections were the first in which these techniques have been widely used and arguably the first in which they have had a major impact. Although the on- and offline targeting of potential voters is completely legal, a report by Carole Cadwalladr in the Guardian (2017) has sparked a debate about whether the used techniques harm our democracy. In this debate, it remains unclear how AI is exactly being used but it does not improve public opinion on AI (which is already notorious for replacing millions of jobs in the future). Whether these techniques are ethically correct is a question regarding norms and values which society should answer. However, it’s beneficial to realise how data is used to change ‘audience behaviour’.
Facebook is the central platform on which voters have been influenced during the mentioned campaigns. Be harvesting publicly available Facebook data (likes, friendships, etc.) a personality profile can be made for each user. This is done by embedding psychological insights, using the OCEAN personality traits, into an algorithm which can then by optimized by AI. Subsequently, users can be categorized based on various variables including, personal characteristics (naivety, anxiousness, etc.) or geographical location. Based on the needs of the campaign team, very specific groups can then be influenced via the sponsorship of content on Facebook.
The above-described method of data-analysis has been developed by Michal Kosinski, a psychologist and data scientist at Cambridge University. The SCL Group, a “global election management agency” hired top researchers from Kosinski’s team to create Cambridge Analytica in 2013. Although Cambridge Analytica denies any involvement in the UK elections, CEO Alexander Nix claims that they have had a significant impact on the Trump election: “We collect up to 5,000 data points on over 220 million Americans, and use more than 100 data variables to model target audience groups and predict the behaviour of like-minded people.”.
This brings us closer to an explanation on why these recent elections have sparked the debate about the use of AI in elections. Where the collection of the necessary substantial sets of personal data used to be the disadvantage of using psychological analysis (based on the OCEAN model), this problem has now been overcome using cloud computing in combination with AI. Cloud computing has never been so widely available and affordable which enabled the possibility of using these techniques on an enormous scale, as confirmed by Nix’s statement. Does this mean that we should worry about our democracy, given the possibilities of Artificial Intelligence?
Brynjolfsson and McAfee (2017) recently published an article discussing the possibilities and the risks and limits of the increasing usage of AI. Regarding the limits, they refer to Pablo Picasso’s observation about computers: “But they are useless. They can only give you answers.”. This might not entirely hold anymore but the essence remains the same. While it might not be always possible to fully understand the answers which the ‘computer’ gives because of deep neural learning, it is eventually up to businesses, entrepreneurs, scientists, etc. to ask the question in the first place. In the case of Cambridge Analytica, influencing the election behaviour of specific groups was the main goal for which AI was used. But is this fundamentally any different than going around neighbourhoods campaigning for votes? AI just offers a more efficient way of targeting and reaching out to people in this digital era and Cambridge Analytica has been the first to apply it on this scale. Trump and the Leave.EU campaign might have won by being an early adopter of AI but the increasing accessibility of AI will not disrupt our democracy as it is becoming available for all. Sounds very democratic, right?
References:
– Brynjolfsson, E. and McAfee, A. (2017). The Business of Artificial Intelligence. Harvard Business Review. Retrieved from: https://hbr.org/cover-story/2017/07/the-business-of-artificial-intelligence
– Cadwalladr, C. (2017, May 7). The great British Brexit robbery: how our democracy was hijacked. The Guardian. Retrieved from: https://www.theguardian.com/technology/2017/may/07/the-great-british-brexit-robbery-hijacked-democracy
Suggested further reading:
– Helbing, D., Frey, B. S., Gigerenzer, G., Hafen, E., Hagner, M., Hofstetter, Y. & Zwitter, A. (2017, February 25). Will Democracy Survive Big Data and Artificial Intelligence. Scientific American.
Interesting analysis Rutger. Another way to look at AI in elections, is that If AI learns more about what each individual wants in a leader, you could say that AI is making the elections more democratic, as all the competing politicians need to cater to the needs of increasing niche-groups if they want to get votes.