If you want to reduce income inequality, kill a robot. If you want to increase income inequality, kill a union.

Could that be the choice we face in today’s economy? Labor Day is often the moment when we look back at the history of the trade union movement, and reflect on all the contributions it has made to our collective well-being: little things like pensions, workplace safety rules, and the minimum wage. But it is also a moment to look forward, and to consider the forms of worker advocacy and social support we will need in the years ahead.

At this moment, that means thinking very carefully about the relationship between income inequality, computerization, and unionization. The three are closely related, as Tali Kristal argues in “The Capitalist Machine: Computerization, Workers’ Power, and the Decline in Labor’s Share within U.S. Industries.” Kristal takes a hard look at the argument that rising income inequality is a direct result of computerization. Economists often assume that technology has driven down workers’ share of income due to “the increase in machine productivity relative to workers’ productivity.” But Kristal analyzed 60 years of data across 43 different US industries, which tells a more complicated story. It’s not computerization that drives down workers’ share of national income, relative to the share that is earned by capital (i.e. company owners and investors). Rather, it’s the way that computerization affects unionization rates: by weakening unions, technology has changed the balance of power between labor and capital, and allowed the owner/investor class to claim a larger share of income.

Kristal points to two mechanisms through which digital technology has weakened the labor movement, and thus, workers’ relative bargaining power, allowing a greater share of profits to go to the owner/investor class:

First, the computer revolution’s strengthening of management control may have empowered employers and management, allowing them to use more legal and illegal anti-union tactics designed to intensely monitor and punish union activity. Second, computer technology is linked to skill polarization in the workforce, which may have undermined workers’ solidarity, thereby reducing the likelihood of working-class cohesion and collective action.

This dynamic is “class-biased technological change”: innovation that intrinsically and systemically privileges some socio-economic groups over others. Kristal’s analysis is not only useful in understanding the role of technology in the levels of income inequality we see today, but also in anticipating the potential impact of the latest wave of tech innovation—like artificial intelligence and the explosion of the collaborative economy.

In “The History of Technological Anxiety and the Future of Economic Growth: Is This Time Different?“, Mokyr et al. note the concern that “if artificial intelligence and robotics continue on their present trend, future machines will be able to carry out these human capabilities, at least in certain contexts and to a certain extent. Thus, it seems frighteningly plausible that this time will be different, and large sections of the labor market will be dislocated.” And as Sara Horowitz writes in “Freelancers in the U.S. Workforce,”

Online work platforms, such as Uber, Airbnb, Etsy, and Elance, that connect workers directly to consumers and clients are completely reimagining the work relationship. Under the National Labor Relations Act, however, these workers and the millions of other freelancers are largely prohibited from forming unions.

While Mokyr et al. use historical precedent to argue that our anxiety about the impact of these kinds of changes may be misplaced, even they acknowledge that “[w]hile the predictions of widespread technological unemployment were, by and large, wrong, we should not trivialize the costs borne by the many who were actually displaced.”

If unions have been the historical buffer that protects workers (and the worker class) from the worst excesses of tech-driven disruption, how are we to address the challenges of technologies that weaken unionization itself? Given the relationship between unionization and income inequality, it’s a pressing question: quite apart from the crucial question of social justice, this election season has sharpened our awareness of the risks that widening income inequality poses to our society and our political system. We could be looking at a future in which technology depresses unionization rates, drives more and more workers into the freelance economy, widens the income gap, and drives social inequality and fragmentation.

Or we could be looking at a future in which we learn from both the history of tech-driven economic disruption and of the labor movement itself.  Mokyr et al. ponder a possibility that has been raised by many observers of the rising freelance economy:

In particular, we believe that there is a distinct possibility that wages for some classes of workers may need to be supplemented through some income redistribution. In addition, it may be necessary to expand the set of publicly provided goods to include certain “primary goods” (Rawls 1971) such as food, housing, education, and health care that are necessary for a modern life to go well. For many others, cheaply produced goods and increasingly automated and freely available services should allow access to increasing levels of material well-being and health.

It’s a rosy picture that will appeal to those of us who’ve been inspired by the currency-free societies of techno-utopias like the Star Trek universe. But the precedents for that kind of social safety net lie in the benefits that were wrested by the labor movement: health insurance, unemployment insurance, and pensions all came about through union bargaining and labor-organized political struggle. Kristal gives us good reason to worry that the very technologies that drive our need for a next-generation safety net will, ironically, cripple the labor movement that could help us achieve it.

But she also gives us an implicit road map for addressing the tech-enabled obstacles to worker organizing and class-based collective action. If computerization weakens organizing power by enabling management surveillance and control, we can resist and regulate the technologies of intrusion, embracing privacy-enhancing technologies like encryption and anonymous web browsing (if only to serve as camouflage for the friends and colleagues who rely on those technologies to organize). If computerization weakens class solidarity by dividing highly skilled workers from less-skilled workers, we can fight that erosion with associations and unions that draw attention to the common structural position and interests of Uber drivers, Airbnb hosts, Etsy artisans, and freelance coders.

If that sounds like a tall order, remember that the job of building class solidarity and broad-based labor unions must have looked just as daunting to the workers of the Industrial Revolution a century or two ago. But they recognized the urgent need for collective action in the face of technological changes that disrupted traditional forms of social and economic organization—just as we need to recognize the social, political, and economic imperatives of the technological changes we face today.