To Fix Tech, Democracy Needs to Grow Up

Calls to “democratize technology” ring hollow when both systems seem to be failing. The key is realizing that democracy is not yet in its final form.
Photo collage of an I Voted sticker a person with a projection of codedata over them and bees working on a honeycomb
Photo-illustration: Jacqui VanLiew; Getty Images

There isn’t much we can agree on these days. But two sweeping statements that might garner broad support are “We need to fix technology” and “We need to fix democracy.”

There is growing recognition that rapid technology development is producing society-scale risks: state and private surveillance, widespread labor automation, ascending monopoly and oligopoly power, stagnant productivity growth, algorithmic discrimination, and the catastrophic risks posed by advances in fields like AI and biotechnology. Less often discussed, but in my view no less important, is the loss of potential advances that lack short-term or market-legible benefits. These include vaccine development for emerging diseases and open source platforms for basic digital affordances like identity and communication.

At the same time, as democracies falter in the face of complex global challenges, citizens (and increasingly, elected leaders) around the world are losing trust in democratic processes and are being swayed by autocratic alternatives. Nation-state democracies are, to varying degrees, beset by gridlock and hyper-partisanship, little accountability to the popular will, inefficiency, flagging state capacity, inability to keep up with emerging technologies, and corporate capture. While smaller-scale democratic experiments are growing, locally and globally, they remain far too fractured to handle consequential governance decisions at scale.

This puts us in a bind. Clearly, we could be doing a better job directing the development of technology towards collective human flourishing—this may be one of the greatest challenges of our time. If actually existing democracy is so riddled with flaws, it doesn’t seem up to the task. This is what rings hollow in many calls to “democratize technology”: Given the litany of complaints, why subject one seemingly broken system to governance by another?

At the same time, as we deal with everything from surveillance to space travel, we desperately need ways to collectively negotiate complex value trade-offs with global consequences, and ways to share in their benefits. This definitely seems like a job for democracy, albeit a much better iteration. So how can we radically update democracy so that we can successfully navigate toward long-term, shared positive outcomes?

The Case for Collective Intelligence

To answer these questions, we must realize that our current forms of democracy are only early and highly imperfect manifestations of collective intelligence—coordination systems that incorporate and process decentralized, agentic, and meaningful decisionmaking across individuals and communities to produce best-case decisions for the collective.

Collective intelligence, or CI, is not the purview of humans alone. Networks of trees, enabled by mycelia, can exhibit intelligent characteristics, sharing nutrients and sending out distress signals about drought or insect attacks. Bees and ants manifest swarm intelligence through complex processes of selection, deliberation, and consensus, using the vocabulary of physical movement and pheromones. Humans are not even the only animals that vote. African wild dogs, when deciding whether to move locations, will engage in a bout of sneezing to determine whether quorum has been reached, with the tipping point determined by context—for example, lower-ranked individuals require a minimum of 10 sneezes to achieve what a higher-ranked individual could get with only three. Buffaloes, baboons, and meerkats also make decisions via quorum, with flexible “rules” based on behavior and negotiation. 

But humans, unlike meerkats or ants, don’t have to rely on the pathways to CI that our biology has hard-coded into us, or wait until the slow, invisible hand of evolution tweaks our processes. We can do better on purpose, recognizing that progress and participation don’t have to trade off. (This is the thesis on which my organization, the Collective Intelligence Project, is predicated.)

Our stepwise innovations in CI systems—such as representative, nation-state democracy, capitalist and noncapitalist markets, and bureaucratic technocracy—have already shaped the modern world. And yet, we can do much better. These existing manifestations of collective intelligence are only crude versions of the structures we could build to make better collective decisions over collective resources.

In some sense, the democratic structures we have now, such as majority voting, lifetime appointments, and archaic levels of representation, operate by even cruder rules than many algorithms. So do our markets, which ignore a wide array of subtle value negotiations in favor of rudimentary optimizations for cost, profit, or share price. These badly need to be updated. In fact, long-term fears about runaway artificial intelligence models over-optimizing misaligned outcomes and destroying humanity parallel, in many ways, the dangers of committing to these crude optimizations in our existing decision-making.

Adopting a frame of collective intelligence lets us see existing democracy as a starting point, rather than as a finished project. We mustn’t relinquish democratic ideals—without which we are left with pure technocracy, and an erosion of shared political judgment into mere technical expertise. But to tackle the challenges of technological advances, we will need to do far more than simply shore up our crumbling democratic institutions. Instead, we must leverage emerging technology towards building better, more intelligent collective institutions for the deeper exercise of collective agency.

Mechanisms, Technologies, and Systems

To understand how to construct better building blocks for collective intelligence, we can break the CI “stack” down into mechanisms, technologies, and systems.

Mechanisms elicit, process, and aggregate collective intelligence from diverse sources. They let us combine people’s preferences and local information into decisions. Direct majority voting is an example of a basic CI mechanism. Others include various modes of price-setting and exchange, consensus-based deliberation, top-down agenda setting, delegation to representatives, jury selection, and so on.

Building toward better collective intelligence could begin with fairly minor updates, like replacing direct majority voting with ranked-choice voting or quadratic voting. These can allow for more granular expression of preferences (by taking into account second and third choices, or enabling different allocations of “voice credits” to different issues). This could expand into a range of mechanisms, combining elements from things like prediction markets (which incentivize predicting the probability of future events), auctions and exchange mechanisms (which can triangulate notions of shared value), retroactive funding (which enables post hoc allocation of resources), liquid democracy (which allows a polity to dynamically allocate representation), and sortition (in which a representative sample of stakeholders make deliberative decisions).

These possibilities are brought to life via technologies, which help CI mechanisms scale and interoperate across contexts. Take Wikipedia, an extraordinarily valuable manifestation of CI (accounting for almost half of the value of all Google searches) that is itself run via nested CI mechanisms for entry editing, expansion, and dispute resolution. Wikipedia didn’t invent the idea of a collaborative encyclopedia—in fact, the Oxford English Dictionary arguably adopted a similar, albeit analog, process of volunteer contributions in the late 19th century—but technology like basic internet protocols enabling hyperlinks and wiki structures made it possible at scale in a way previously unimaginable.

Recent advances in coordination technologies could have similar stepwise effects—the key to governing rapidly advancing tech may come in the form of utilizing this technology for governance. For example, decentralized organizations using blockchain-based technologies are experimenting with token-based delegation, quadratic voting, impact certificates, non-transferable voting tokens, and the like.

Deliberative democratic platforms, like pol.is, have already found great success using basic machine-learning algorithms to determine points of consensus among large groups with differing opinions. Future augmentative applications of artificial intelligence could change the landscape of effective CI by representing preferences in deliberative contexts, using language models to identify argumentation gaps or points of consensus, or computing consensus across clusters of opinions.

This brings us to CI systems, which are the crux of where radical improvements can manifest. Without well-designed systems, there may never be consequences for bad decisions, or mechanisms may just surface insights that are never used (as happens with many “send information to your representative”-style technology projects). Expanding CI systems would allow us to move far beyond existing, nation-state notions of democratic participation into a rich web of participation and progress across sectors, borders, and scales.

These systems can take the form of:

  • Collective cognition – arriving at useful answers or truths about a situation through collective input
  • Collective coordination – synchronizing individual activities to achieve a common goal
  • Collective cooperation – facilitating cooperation among actors with different and possibly misaligned self-interests

Small-scale experiments have already shown the possibility of each of these ingredients. This could look like a CI-enabled workplace that incorporates inputs from internal prediction markets (as has been previously attempted at Google), as well as team deliberation processes (enabled by a deliberation platform such as Loomio) and collective ownership structures (as developed by Exit to Community).

This can be taken even farther. Imagine a world of abundant public goods, funded via consortium-based taxes coupled with active participation in investment based on shared need. Right now, many projects with incredible social-value returns, from transportation infrastructure to small businesses that would benefit many in a community, can languish in an innovation valley of death because they’re not well set up for pure public funding, and yet are misaligned with the incentives of private capital. CI mechanisms for pooled funding could address this gap, redirecting state and philanthropic resources into financially supporting these projects in proportion to their deemed benefit to a given community, and dynamically reallocating when necessary. This could better direct scientific and research funding, creating massive positive externalities not well incentivized in the current system, or direct public funding for industrial policy (instead of direct subsidies, which can miss necessary local information or be prone to capture and cronyism).

Or imagine a world of CI-enabled firms prioritizing collective provision at scale through effective economic democracy. Advanced CI mechanisms could expand collective input and ownership beyond voice in a single workplace, or a vote every four years. Imagine networks of production run with input from local and global stakeholders, using transaction fees as taxlike sources of funding for long-term investment—radically updating the regulated monopoly. Or AI assistants helping communities navigate value trade-offs, scaling up commons-based governance practices while computing a range of quantitative and qualitative measures to optimize—instead of just maximizing share price. Or platforms that enable individuals and communities to track the effects of new technologies, internalizing externalities from environmental degradation to pandemic risk.

Of course, getting there won’t be easy—changing power structures never is. Scholars from John Dewey to Helene Landemore have emphasized the material changes and conditions of “education and freedom”—from access to basic necessities to economic security—needed for democracies to truly enable collective intelligence. Incentive-alignment work, political shifts, base-building, and public advocacy are essential for conveying the urgency of shifting transformative technology toward collective intelligence and input.

What This Could Look Like

For a deeper dive into building a CI system, we can turn to the example of data governance, an area that is doubly important given the centrality of information flows to enabling CI across contexts.

The existing data economy (mirroring the digital economy as a whole) is a primary engine of shared growth and progress—and a leaky, power-concentrating, fractured mess. Data brokers sell and resell personal data with little oversight. Huge networks like Facebook and Google capture the information of billions of people and use it in the service of a few shareholders’ narrow interests. It is only during brief moments of generosity during a crisis, like when  Google provided mobility data to cities during the Covid pandemic, that the public can even see how vast these data stores are, and how helpful they might be in building shared safety and prosperity.

This has led to a resurgence of interest in antitrust and firm breakup. But these remedies are highly imperfect. They may entrench existing business models and misunderstand the reality of network effects and capital costs underlying emerging tech. And given that they must be carried out by nation-states, they often do not take into account global stakeholders, who may have very different views on what a good outcome is than, say, an American senator. Top-down regulation by nation-states is certainly necessary to curb the harms of the current system, but it cannot build a better system, particularly when fractured across borders. Other alternatives aren’t particularly promising either. Proposed “data marketplaces” re-create the network effects that lead to current monopolies, while adjacent “own your own data”-style proposals are infeasible and mostly serve to further enclose the digital commons. Neither of these pure state- or market-based approaches provide real ways to either incorporate collective input or build toward collective outcomes.

A CI-based ecosystem of data governance would enable greater privacy, greater control, and greater access. We might start with mechanisms for decentralized governance, like the digital identity-based delegation systems being piloted in decentralized organizations. These can provide a platform to delegate responsibility to a trusted fiduciary or to choose to spend time voting on specific proposals that one deems of high importance (maybe you trust your local credit union to handle your financial data, except in cases of cross-border transfers, where you want to have more input). As a result, larger numbers of people could participate productively in data stewardship,

Choosing the right fiduciary might take some up-front engagement—more than we’re used to in a world of single-click “I Agree” buttons, which are now used to justify tracking and black-box data marketplaces. But after initial allocations, a shifting set of data intermediaries would be well placed to look out for your interests, with pathways for updates over time. Eventually, these decisions might be supported by augmentative AI assistants to navigate the complexity.

This allows for shared governance of data networks, rather than responsibility for one’s “own” data, which is practically impossible. After all, which individual “owns” an email, or a Wikipedia page, or even genomic data that inherently contains family information? As our inputs into digital ecosystems grow, so do the networks of data that we might have privacy, financial, or other shared interests over. CI-based data governance would more closely mirror citizen science initiatives, allowing for the pooling of high-quality information to solve problems that data holders find important. Of course, this requires a range of privacy protections. In particular, this would incorporate machine-learning techniques that have made it easier to share data insights, without sharing the underlying data sets, making information transfer possible with lower privacy costs.

Such CI systems for data governance are in early stages, with ongoing experiments across distributed data governance (e.g., PoolData), privacy-preserving data sharing (e.g., OpenMined), data coalitions and cooperatives (e.g., the Data Freedom Act), and institutional innovation (e.g., the Data Act in the EU). These efforts are supported by growing academic and research interest in data intermediaries. But far more work is needed.

This is particularly important in the age of transformative AI. Large-scale models like OpenAI’s GPT-3 are trained on hundreds of billions of words written by people, capturing centuries of human knowledge, thought, and insight, everything from books to blogs to wikis. These models are incredibly capable, yes, but it is because they are engines of collective intelligence, not just artificial intelligence. And yet, they garner billions of dollars of investment with no recompense for those creating the content. In the future, such models may aim to do everything from large-scale labor automation to implicitly controlling decision processes. Developing the mechanisms, tools, and systems to take a CI approach to data governance can be a stepping stone into taking similar approaches to transformative AI—recognizing economic input and financial interests, to be sure, but also utilizing scalable ways to build toward the preferences, values, and needs of the humans that make the system what it is. 

Pushing the Frontier

While the engines of progress for other technological fields are well oiled, many manifestations of CI are underfunded when compared to their promise. But when appropriately resourced and championed, the CI approach has been remarkably successful. Just look to Estonia, which has built a full-stack digital democracy. Or to Taiwan, which has rolled out cutting-edge experiments in deliberative governance, decentralized coordination, and technological innovation, in which more than half of the country’s 24 million residents have participated. These nations are expanding our notions of what's possible: They’ve built platforms for digital voting, legislative input, and collaboration, alongside reforms including innovative tax policy, investments in public infrastructure, and collaborative, open source technical solutions for issues from transportation regulation to climate change.

From my vantage point within the tech governance ecosystem of the US, the situation often feels as polarized as our broader political system. Techno-solutionists eschew democracy while techno-pessimists eschew technology, resulting in a tech ecosystem increasingly divorced from the collective interest and a politics of technology increasingly against even the possibility of shared progress. But in reality, we are as far from the best democratic systems we could have as we are from the frontiers of technology-enabled flourishing. And we can’t have one without the other—at least, not without embracing either a technocratic dystopia or a stagnant one.

This means we need to not only “fix democracy” and “fix technology,” but find ways to leverage each toward the pursuit of the other. Getting there will require policymakers to initiate and finance positive alternatives, not just enact regulation to curb the harms of the current system. It will require political systems willing and able to raise and deploy funding into collective intelligence experimentation, via subsidies, sandboxes for fast innovation, and investment into basic research funding and digital public infrastructure. It will require technologists and researchers to develop metrics beyond artificial benchmarks or maximizing engagement; in turn, it will require funders and journals to reward research breakthroughs that augment collective intelligence and collaboration. It will require civil society organizations to expand beyond (necessary) criticism of existing technology ecosystems into convening communities to imagine and contribute to actionable, better futures. And it will require collective intelligence experiments of all kinds—from the local to the global, from the digital to the physical, from theory to practice. This isn’t just a job for institutions; it’s a job for all of us who are invested in both participation and progress.

For all its flaws, the early internet, the foundation of many CI instances today, was built with public funding, research, civil society input, and private innovation. It has gone on to restructure our age. The almost insurmountable challenges of this century will require coordination on an even more massive scale. But the rewards are likely to be even greater. We should invest accordingly.