Democracy and the birth of ‘Clean AI’

Despite Cambridge Analytica waking the world up to the manipulative power of AI, algorithmic encroachment on the democratic process shows no signs of abating. Dr. Christian de Vartavan says coding transparency is needed to ensure AI enhances democracy, rather than derails it

AI is increasingly seen by some of us as some sort of super inorganic butler discharging us of duties, responsibilities, reflection and even choice. When this attitude is applied to democracy, including voting, or government and justice, it is of great concern.

I have written elsewhere about AI and justice, asking whether we wish to be judged and sanctioned by robots. It’s a matter of personal dignity and self-respect, and raises the following question: Would we surrender our freedom of choice, or freedom as such, if a superior and wiser alien race landed on earth? Never. So why would we surrender to a non-biological entity, namely AI, when a judge of this country with 40 years in the higher courts tells me that a lifetime of experience is often not sufficient to judge some cases.

Finding the balance

Like how the concepts of justice and government are tied together, so are the concepts of government and democracy. In each of these areas, AI is now injected. Why? Because the prospect of delegation is inherent to our species. Some may even argue it is a sign of our intelligence. Why do something if someone, or in this case ‘something’, can do it for us?

The answer is because if we delegate everything, there will be a heavy price to pay. The solution clearly resides in finding the right balance between where AI should and shouldn’t be deployed. In this article, we’ll tackle the problem of democracy, exploring how AI is currently being used in the democratic process and whether it is getting involved in areas where it shouldn’t.

AI & democracy

The period leading to any election is packed with mammoth political campaigns. Every voter has their own ideas and expectations for their candidate. The holy grail for each political party is to succeed in influencing the voter’s choice, even better, swaying it.

Usually, this is done by politicians or their teams during one-to-one meetings with voters. The limitations of this traditional approach are obvious as each candidate cannot meet each voter face to face. Enter mighty AI. Politicians want to use AI to remedy this restriction. Not to meet each voter (this is equally impossible), but to influence them.

In his 2008 presidential campaign, Barack Obama did not have today’s AI capabilities. But he did have big data analysis. Suddenly, big data became not just big money, but big politics. Obama rose to White House with a targeted $1 billion email campaign, allowing him to decline public money to fund his campaign, a first in American political history.

2016’s watershed moment

The 2016 US campaign took big data and AI to a completely new level. And not in a pretty way. We all remember the Cambridge Analytica scandal. The UK-based data firm rolled out an extensive advertising campaign targeting persuadable voters based on their profile and psychology. Cambridge Analytica understood the power of fear. Paranoid voters received fear-mongering ads. While Conservative voters were targeted with ads based on tradition and community.

Using voluntarily disclosed social media data, AI presented a different version of Trump to each section of the population according to their psychology and expected emotional triggers. As these methods were covert and the AI-powered messages insincere, the project was a great unethical manipulation. Indeed, it’s more accurate to describe it as an exercise in insidious manipulation.

At one point Cambridge Analytica operated in 68 countries, in each manipulating votes on an industrial scale. Left unchecked, democracy and democratic voting had become big business. Before Cambridge Analytica’s activities were uncovered, this level of political deception would have been flagged as a conspiracy theory. Unfortunately, it was not, and Trump was elected. The tripartite of data sets, AI algorithms and buckets of money enabled Trump to publish millions of whitewashed ads to the right voters.

This year’s presidential race

The involvement of AI in the 2020 US presidential race is fascinating and well worth dissecting. Gary Coby, Trump’s 2020 digital campaign director, compared the current strategy to high-frequency financial trading. According to reports, Facebook has built an algorithmic ad-buying system turbo-driven toward results. It’s no surprise that Trump’s campaign team is pushing the engine to its limits, spending tens of millions of dollars on the political track.

The new system is the opposite of self-driving cars or glassy AI robots but a deeply dull and invisible devil-in-the-box that is dramatically reshaping the democratic process. To the point that political campaigns have ceased to communicate with voters but with the AI instead. As a result, Facebook’s AI-powered targeted advertising has not become a crucial but perhaps the crucial component to winning the 2020 US campaign.

The scary part is that not only does this AI outperform any human manager but that it works in such an ineffable way that not even Facebook truly grasps it. This is why multiple campaign ads are run simultaneously, as in truth, no one has a clue which ad will fetch ad the highest engagement until after the system has worked its magic. Last October the Washington Post revealed that Trump-supporting PACs (Political Action Committees) had put misleading AI-driven ads on voters’ Facebook pages. For Brad Parscale, Trump’s campaign manager, Facebook’s ad transparency tools were “kind of like a gift,” enhancing the reach of Trump’s views.

What is certain is that the soulless campaign-god is now out of the machine and shaping the future of the US (and hence the world) without human control. This science fiction scenario is as real as that which we are currently living through with Covid-19. Who would have imagined such power would be unleashed in our midst? Science fiction authors.

The technology itself is not unethical. Those allowing it to run in the dangerous way it is currently being applied should be blamed. AI can in fact do a lot of good for politics. To start with, targeting voters and informing them about important political issues. AI can also provide intelligence to help voters decide who to vote for, but it should only do so if it is done in a neutral way. Perhaps AI should be banned from any democratic election until neutrality can be guaranteed, but it is no more possible to ban AI from elections than from our lives where it is now permanently and increasingly a part of the fabric of society. But we can ethically regulate and repurpose it to support democracy.

Any AI is only as good or bad as the ethical and moral values embedded in the code of its algorithms. Democratic values must, therefore, form the backbone of AI programming and this backbone must be built by developers and checked by democratic institutions that respect these values. In sum, we must set boundaries for what AI can or cannot do where elections or voters are concerned.

Transparency in the coding is hence crucial in this respect as we need to know who is coding what. In other words, the AI must be clean. ‘Clean AI’, a simple new concept I am pleased to here introduce (please make a quick Google search), is in my opinion the future alongside clean politics, although is admittedly somewhat much harder to get. One ethically driven, with well-defined moral values or principles for its functioning code. Code which should be regulated and audited by governmental bodies. So that, instead of vilifying it, AI enhances democracy.