Software ate the world. Can technology eat democracy?

Democracy is young, and not as common in reality as it is in the eyes of people that are accustomed to live in one. Democracy is neither inevitable nor guaranteed, and despite our perception of the democratic regime’s foundations as self-evident and naturally occurring, it is the various shades of tyranny and autocracy that have historically dominated the human power structures.

Because technological trends of an era evolve concurrently with the political power structures in vogue, examining past technological change helps explain the rise and fall of political regimes, and analyzing the current trends can indicate the governing structures that are most likely to dominate the political future.

In this two-part essay, I will demonstrate how democratic advantages were built concurrently with the dominant technologies of the 19th and 20th centuries, and how today those advantages are being eroded by two technological trends, specifically:

• behavior-modification abilities developed by profit-seeking entities that undermine the value of the expression of an individual’s free will
• the need for centralized information-processing structures in the AI era that paves the way for the establishment of autocratic regimes

Part 1 — Who’s in charge?

Democracy is hard: hard to measure, achieve, preserve, and often hard to understand — especially when either your salary or personal influence depends on your inability to understand it.

Yet functioning democracies do exist, and most of them are thriving. Manifold theories flourished after the democratic wave of the 19th century depicting their benefits: the Democratic Peace Theory, for example, explains how democratic countries don’t go to war since their leaders are accountable for the loss of life and property to a public that can vote them out. Democracies also reduce exploitation and prevent abuse of power, as they separate the branches of government with no single person having full control over either of them. Also, as long as the public believes that the leaders are elected fairly, the likelihood of a revolution is fairly low, as it is much harder to question the lawfulness or legitimacy of leaders when more than half of the country’s citizens said that they are okay with being governed by them… The list of advantages goes on.

But can we still reap those democratic benefits, that we often take for granted, in today’s tech-enabled world?

1.1. The individual in power

With seemingly self-evident ideas like freedom and universal voting rights, the democratic regime has empowered the individuals living under its ruling and cemented itself in their minds.

When defining a democracy, free and fair elections are top of mind — no functioning democratic regime has yet been established without them. The capacity of voters to participate freely (it’s your decision who to vote for) and meaningfully (your vote actually counts) in the shaping of the political and social systems is protected by safeguards, a politically-independent due process, and a system designed to guarantee the expression of citizen’s unaltered free will. Personal choices of citizens thus have a direct effect on who holds power: with our votes, we discern the winners and losers of the political arena.

Another defining feature of a functioning democracy is a free market economy, one in which no ruling oligarchy captures the proceeds of economic growth. When electing a government, the citizens are likely to pick one that promises a level-playing field for everyone and will vote out the crooks and embezzlers. Also, it has been demonstrated that economic freedom, guaranteed by democratic societies, leads to more freedom of choice (in employment, education, place of residence, etc) that in turn translates into more economic prosperity.

In “Why Nations Fail”, Daron Acemoglu illustrates that because of a system designed to share economic benefits, democracies get more creative destruction, innovation, and prosperity when they protect innovation by enforcing property rights, and let the invisible hand decide what the customers value. Economic freedom also leads to more interconnectivity: the benefits of economic integration reduce the policy choices open to governments, making war — which disrupts that integration — so unattractive, as to be practically unthinkable. But the important psychological byproduct of economic freedom is that most of us now believe that consumer sovereignty plays a major role in shaping the economic landscape. Sure, policies exist to guarantee fair competition for the contents of our wallets but it is our economic decisions that matter most and every economic transaction that we engage in makes us feel like we are picking the winners and the losers of an industry.

In places that we refer to as “the free world”, we believe that it is our free will that, by shaping market and power structures, correctly allocates both political power and economic resources to the worthiest.

Having said that, the liberal-democratic system is today being accused, often rightly and not solely so, of no longer providing or evenly distributing wealth, and segregating society into “working” and “ruling” classes. Alienation of the middle class and its withdrawal from debates over the future and the direction of technological development is laid out by Yuval Noah Harari, who attributes Donald Trump’s election and the result of the Brexit vote to “the masses [who] fear irrelevance and are frantic to use their remaining political power before it is too late”. While I agree that in many democracies, identity was elevated above policy by unscrupulous politicians, I believe that democracy’s foundation — free will — is threatened by a more existential and tech-enabled threat that I will explain in the following parts of this essay.

1.2. “The Human Use of Human Beings”

Originally, the concept of behaviorism was popularized by the Harvard professor B.F. Skinner: through an external input, or stimulus, he considered that it was possible to alter the subject’s behavior and beliefs, similarly to a machine reacting to a command from the user. Skinner’s findings have later on been extensively used in laying the foundations of behavioral economic science, which has since the early 2000s been a fertile ground for Nobel laureates. As demonstrated in my previous article, roughly at the same time, a combination of left-leaning philosophy claiming that “everything should be free on the internet”, the idolatry of innovators, and an urge to monetize their creations has pushed technology make money through personalized ads. Advertising, which used to be about companies trying to put their content in front of perfect strangers in hopes of boosting sales, now operates more like a behavior-modification empire researching subjects in real-time and altering their behaviors with global reach. The resulting ad-based big tech industry gathered enough data and developed the skills to nudge us, using a wide range of behavioral-modification stimuli exactly at the moment when we are most receptive to them. 14 minutes of Jaron Lanier’s TED talk is all the time one needs to understand the nefarious consequences that result from our acceptance to give away our data and how that enabled the tech giants to carry out behavioral modifications, at scale. Even our governments now openly discuss the use of social media for behavioral modifications to promote public health policies.

One will also note how the responsibility for this system is shifted onto the users: if you are smart enough to be using our free product, you know that we collect and keep all your data; and how entrenched the model has become: the absence of an overt price tag makes it hard to complain about the negative externalities of the product, but I digress again.

Tech platforms also increasingly and unilaterally decide what we get to see on our devices — on which we spend an ever-growing amount of our waking hours — their armies of attention designers are trained to make sure that the content reaches the right audience to produce the desired effect. With our paleolithic brains overloaded with an ever-increasing amount of free information, the valuable commodity that companies compete for has become our attention.

Do the concepts of free market, free will, the invisible hand, or the power of the people by the people — foundations that made liberal democracies so attractive and stable in the first place — still apply in a world where our fears and desires are analyzed to be shaped by profit-seeking businesses and unprincipled political campaigns? Should we even keep listening to the expression of the will of citizens, when behavior modification systems are widespread and proven to be effective? How about an AI picking the next president, based on a myriad of metrics, whose interrelations are too complex to understand for a human?

One of the prerequisites for a functioning democracy — free expression of its citizens’ will — can very well soon be declared outdated and meaningless. As most of us mindlessly scroll our personalized feeds, the purpose of the governed majority in tomorrow’s democracies may well be reduced to the bitter and disturbing point of data-generating subjects, enslaved to the algorithm-owning lords. One hopeful way out of that dystopia may come from pivoting to paid digital services and thus removing the current incentive that ad-based platforms have to surveil us for monetization.

Part 2 — The true Terminator advantages

What sets democracies apart from authoritarian regimes are not only morally just ideals and aspirations, but also a better system of distribution of information processing and decision-making. The rise of AI systems may annihilate the need for diffuse information processing and may well be another step taken in the death march on which our democracies embarked.

2.1 Diffused they stand

Instead of concentrating information and constantly decisions to a higher authority linked to the central power (usually in the capital), democracies do astonishingly well at subsidiarity: having the decision being taken at the lowest-possible hierarchical level, a concept borrowed from the church’s organization. Another advantage of democracies is outlined in Condorcet’s Jury Theorem: the increase of decision-makers, as long as they have a better-than-average understanding of the issue, is positively correlated with good decisions being taken. Diffusing the power of decision-making thus often leads to an increased decision quality. Note that the reciprocal is also true: an autocrat will rule with absolute power (sharing the slightest bit of it fragilizes his winning coalition and paves the way for his demise) and is unlikely to continuously be right.

The distributed structures of information processing and decision-making have thus made democracies more attractive and reconciled the heterogeneous groups that they contain with a sense of equality and empowerment. The effects of this are often well demonstrated on the battlefield: numerous military experts attribute the victory of Israel in the Six-Day War to the ability of its air force pilots to exhibit more autonomy in the face of unforeseen situations that are commonplace in battle. Combatants familiar with a distributed decision-making structure and individual empowerment tend to do better than the ones constantly referring up.

The historical root of this lies partly in the work of scientists like Charles Sanders Peirce and Karl Popper, who have refined the scientific method and highlighted the need for meticulously formalizing the observations made by science. The spike in information output during the first wave of industrialization resulted in vast amounts of codified data flowing into the books and minds of scientists, ready to be analyzed. The increasing quantity and rising complexity of available information prevented a single entity from being able to process and act on it: the world remained too complex for even the CIA to understand all repercussions and pull all the right strings. The technology of the late 19th and 20th centuries has thus reinforced and catalyzed a naturally-occurring advantage of democratic regimes: a distributed processing structure of information.

Distributed processing, together with diffuse decision-making structures, were set in place because of the inability of any single entity to compute the immense available data and decide how to act on it, and the absence of standardized data formats has only reinforced this trend. Today, a new kind of information-processing technology is emerging, one that may flip that advantage on its head, and democracy’s retreat is already visible today: Democracy Index, created by the Economist Intelligence Unit, has recorded an erosion in 2019 that is being accelerated by the pandemic. The causal relationship is hard to establish, but I believe AI technology to have a responsibility in this shift to autocracy.

2.2 A new handmaiden to authoritarianism

I have speculated that the structure of information processing tools and the rise of democratic regimes may be correlated. The emergence of AI may reverse this trend and instead favor centralized structures of information processing. In doing so, it will favor concentrated power structures and could lead to more authoritarianism in the world.

The amount, accuracy, and completeness of data are the main metrics of success for a machine learning AI algorithm training itself: the learning process involves repeated data analysis instead of a rigid set of rules. The GIGO rule (Garbage In = Garbage Out) has never been more valid than in the context of artificial intelligence algorithms, training themselves on data and instead of executing pre-written scripts. For example, AlphaZero did not learn any of the classical chess openings but reached Grandmaster level in four hours by playing against itself and discerning patterns in moves that lead to checkmate. We already have multiple examples of AI algorithms using attributes in the data that humans did not previously rely on, discern patterns and provide expected outcomes: new ways to diagnose cancer were discovered by an AI, based on attributes of cells that humans have neglected.

When collecting the training data for its AI, a state that ignores privacy and human rights will inevitably get an edge on one that cares about those: amount, accuracy, and completeness of data all suffer when it is anonymized. As the saying goes, “Data can be either useful or perfectly anonymous, but never both”. Akin to democracies getting an advantage in the late 19th and 20th centuries, when the multiplication of information brought forward diffused systems, autocracies and tyranny will be on the rise tomorrow, when AI will require invasive, personal, and comprehensive data to be trained on.

The future does not look any better when speculating about how those systems will be used. Hollywood movie scripts have imagined that AI will one day turn on us, but my fear is of the exact opposite: an AI that never rebels is one that solidifies the ones in power. An autocrat will use any means available to retain power, and in a chilling example could rely on a behavior-modification AI to nudge his people towards obedience. With such a system, the ones in power that own or access the algorithm will only perpetuate their ruling. Brazil’s current president is already looking into how he could use new personal-data-fueled technologies to tighten his grip on power.

History has shown that real threats for democratic regimes are almost always internal. With enough time, a shameless leader can subvert and sabotage democracy completely: we are seeing this in Venezuela where the opposition is all but eliminated, in Turkey where religious fundamentalism toppled good policy, or in India where the Bharatiya Janata Party (of the current PM) has invaded most institutions including the electoral commission… The list goes on.

Many of the current technological trends, like powerful AI tools and behavior-modification capabilities, may well slide us towards a more undemocratic world, but it would be foolish to declare that the way of processing data is the sole or even the major predictor of how we organize power in our societies. Surely, the current state of those two technologies gives little cause for optimism for the future of democracies, considering also the polarization of views through social media, algorithmic biases, AI-powered deep fakes, and current incentive structures of tech companies. But if we are to secure a bright future for democracies, and humankind overall, the solution lies in having a holistic, structured approach for oversight and planning of the technological changes that are ahead. It is time to see governmental agencies step up to the task and work jointly with associations, journalists, and activist groups that are today filling the gap left by our administrations.