AI, Inequality, and Democratic Backsliding

This post is part of a series of contributor perspectives and analyses called "The Coming Age of Tech Trillionaires and the Challenge to Democracy." Learn more about the call for contributions here, and read other pieces in the series as they are published here.

As a system of elected representative government, democracy is predicated on the principles of popular sovereignty, contingent consent of the governed, universal suffrage that grants every adult identical voting power, the protection of their ability to vote for any political preferences without state intimidation, and the rule of law to guide collective decision making. Due to normative commitments about the attractiveness of collective self-rule and self-determination, this system guarantees basic fundamental rights like enfranchisement, equality, freedom of consciousness, expression, assembly, political participation, and more.

At the best of times, democracies face principle and practical challenges. Common challenges include disengagement from politics, unemployment rates, media bias, the need to modernize voting systems under budget constraints, and the influence of special interests. These issues, however, are overshadowed by the changes brought about by AI. Perhaps only extended constitutional crises introduce as much risk of democratic backsliding.

AI expands computing beyond routine, pre-programmed tasks through learning capabilities and flexibility. This breakthrough enables the automation of non-routine work. However, when driven solely by brute capitalist control over these productive forces and private investors’ expectations of returns, these advances raise serious concerns about inequality, wealth distribution, surveillance, and un- and under-employment driven by automation. Indeed, sociologically, the success of democracy depends on limiting and reversing social and economic inequality, while economic crises can trigger political instability and reduce a country’s ability to cope with adverse shocks. And as I argue, redistribution is essential to ensure that AI bolsters, rather than weakens, democracy.

Refocusing on the fundamentals

Many focus on AI’s threat to democracy through information manipulation—the fear that misinformation will lead voters astray or malicious actors will undermine election security, erode trust, and destabilize politics. However, this view may oversimplify the relationship between AI and democratic systems. A more fundamental concern involves property relations, the subsequent control rights, and how over-accumulation of finite assets leads to concentration in the hands of a few. For this reason, it is necessary to plot the social impact of AI by examining how it changes or preserves the existing balance of power between labor and capital.

Due to current inequalities of power, the development and deployment of AI are shaped by global capitalist social relations and power structures. The evolution of AI, driven by private investment and built on the commodification of personal data for advertising and behavioral manipulation, has three main implications for social change. First, revolutionary technological systems often trigger economic transformations that redistribute wealth and opportunity, creating inequalities, stratifications, cleavages, and class resentments that alter established politics. Next, like other digital tools, AI is transforming who controls information flows and how such information can be leveraged, also altering politics and the balance of wealth. Lastly, revolutionary technologies introduce countless social transformations, shifting norms and ambitions with open-ended consequences. One should not assume that democracy automatically thrives or survives these changes.

To simplify, in the recent past new technologies not only automated existing tasks but also created new ones, leading to job growth, shared prosperity, and broad economic benefits. The post-war experience of rapid growth and shared prosperity was partly because of a production paradigm where automation allowed workers to undertake new tasks. Unions also had a significant role. However, in the neo-liberal era, automation has contributed to a widening gap between capital owners and labor, with a rising share of national income going to capital and a falling share to labor.

The market structure’s evolution for AI providers will impact how these tools are integrated into business applications and who profits. There are two potential outcomes. One possibility is the emergence of a competitive environment fostered by open-source models and the entry of numerous small generative AI providers. Although not a solution to all the problems, this scenario would prevent the monopolization of these tools and the massive datasets collected by the tech industry.

Alternatively, an oligopolistic structure could arise, where a few companies with a lead in data collection and large, all-purpose generative AI model development become dominant, using this commanding position to co-opt and buy potential competitors. In this scenario, these companies would gain a disproportionate share of the profits from new AI tools, intensifying inequality and potentially increasing their influence on technology’s direction and the centralization of information control.

In short, private investor-driven AI development is anathema to broad-based economic prosperity, social cohesion, and the material conditions that enable democracy.

The consequential problems of unfair distribution

The extreme wealth of capitalists does not exist in isolation—it directly impacts everyone else. Active social participation and democratic decision-making are more likely when people believe they have a chance at economic success and are receiving their fair share of economic benefits. Critics often claim that this represents ‘the politics of envy’ and insist that people should focus on their own success rather than concerning themselves with others’ wealth.

Capitalists out-compete ordinary people for ownership of and influence in politics, politicians, and media. Ultimately, ordinary, working people find themselves in an increasingly unwinnable competition as the growing wealth gap makes it harder for them to compete in the market. As the middle class experiences a loss of assets, and capitalists amass wealth, the wealthy increasingly control essential resources.

This is the true effect of social and wealth inequality, which the Musk situation makes painfully clear. When the rich can systematically outbid others and when they command more assets and resources, their actions, regardless of intention, fundamentally undermine democratic principles. This explains why we should care about inequality and the distribution of wealth and power. The only effective protection against this trend is collective action—people working together as a unified political force to counter these undesirable processes.

Another AI world is possible

The ruling class need not capture computation. Instead, digital society requires significant restructuring to facilitate greater democratization. AI development should prioritize human flourishing over corporate profit. This requires challenging the capitalist imperative to optimize for profit at the expense of people.

Government policy may be necessary to redirect AI in a way that is more just. This involves many different policy actions in many different areas, as well as an administrative state. Here are two priorities. First, attention should be given to where the tax code currently incentivizes investment in algorithms. A fairer tax system, where the marginal taxes for hiring and training labor are equivalent to those for investing in equipment and software, might encourage fair distribution as there are fewer incentives to save on labor costs.

Second, establishing well-functioning data markets is necessary. Current laws enable the use of freely available data with few constraints, which can lead to the manipulative use of information and discourage investment in high-quality data. Legislation establishing property rights for workers and citizens over their data could support a functioning data market, discouraging business models based solely on maximizing user attention and engagement. Efforts like these can accumulate to have dramatic spillover effects for a fairer distribution of wealth and decentralization of power.

To build a more equitable society, especially given the substantial rise in inequality over the past forty years, it is crucial to find ways to utilize new AI technologies to foster shared prosperity. If new technologies sideline workers, discourage political participation, and promote manipulation on online platforms, they may raise new risks to democracy. Risks against shared prosperity and democracy are tightly entangled. As I wrote in Algorithms and the End of Politics, a wide array of technological trajectories are possible, and if one endorses frameworks of justice developed by John Rawls, T. M. Scanlon, and others, perhaps there are some we are morally obligated to pursue.