Big Tech’s monopoly over online discourse threatens democracy. “Middleware” promises a path forward by adding competitive, customizable layers of recommendation algorithms. But can middleware decentralize social media without falling prey to old monopolistic patterns? In new research, Madhavi Singh argues that without targeted regulatory guardrails—including mandatory API access, structural separation, and stakeholder empowerment—even middleware could succumb to concentration.
Big Tech’s monopoly control of the online flow of information and its ability to curate and regulate social and political discourse in the digital public sphere has worried many about its impact on democracy. The United States legal system has struggled to respond, constrained in their efforts by First Amendment protections and immunity granted to market players like Meta and X for information published by third-party actors on their platforms under Section 230. Against this backdrop, the proposal for middleware, third-party algorithms to recommend and curate content for users, appears to avoid these legal conundrums and focuses instead on reforming the underlying technological and business structure of social media platforms to break their monopoly on online information flows.
But if middleware is to diffuse Big Tech’s hold over information flows, regulators must be involved to prevent this section of the marketplace of ideas from collapsing back into old patterns of consolidation. A solution based entirely on free-market middleware competition is no solution at all. Specifically, the market for middleware must be set up to require that 1) social media platforms be compelled to provide mandatory and uniform access to APIs (the software by which a user and platform communicate); 2) platforms be prohibited from offering their own recommender algorithms (structural separation); and 3) attempts to de-concentrate social media must be accompanied by concurrent efforts to empower groups affected by concentration, such as publishers and advertisers.
Middleware and its different versions
The Stanford Working Group on Platform Scale, one of the biggest proponents of middleware, defines it as “software and services that would add an editorial layer between the dominant internet platforms and consumers.” Simply put, middleware creates a marketplace for recommender and curatorial algorithms on top of social media platforms. Social media users are currently tied to platforms’ proprietary recommender algorithms when they create an account. These proprietary algorithms account for a specific set of metrics (like location, age, gender, sexuality, activity history, etc.) of the user and optimize for a fixed goal like “engagement” or “time spent.”
Users cannot customize these metrics or choose alternative optimization goals. The issue with proprietary recommender algorithms isn’t that their optimization goals, such as maximizing “engagement” or “time spent,” are problematic in themselves. Rather, the deeper problem is that any platform relying exclusively on a single algorithm must inevitably choose from among a wide and highly contested range of optimization values—such as accuracy, privacy, inclusiveness, or equity—over others. Deploying a single proprietary algorithm with its narrow set of priorities to serve a user base as culturally, economically, and politically diverse as social media users inevitably results in homogenization. Such a “one-size-fits-all” approach deprives users of meaningful control over the structure, operation, and inherent value trade-offs shaping their social media experiences.
Proponents of middleware propose to remedy this by allowing third parties to provide a range of recommender algorithms that operate on top of social media platforms. In a competitive market for middleware, free speech absolutists could choose a recommender algorithm that doesn’t remove any content and simply ranks it based on relevance, time posted, or other generic metrics. These algorithms could signal their value by highlighting their independence from the financial and ideological interests that could influence platforms.
On the other end of the spectrum, parents of children using social media or others who wish to avoid exposure to certain types of content could choose middleware with more stringent content moderation policies that filter out all violent or sexual content. Yet others could choose middleware that displays only fact-checked content from verified journalistic sources.
Tech and political thinkers such as Jack Dorsey, Stephen Wolfram, Richard Reisman, Francis Fukuyama, Daphne Keller, and Ethan Zuckerman have all been advocates of middleware. Versions of this idea are also being tested in the market—indeed, the rise of decentralized networks was the most transformative trend in social media in 2024. Bluesky offers a range of customizable recommender algorithms. Federated social media platforms like Mastodon or Reddit pursue decentralization by allowing individual server admins to set their own moderation policies. Non-profit options like the Initiative for Digital Public Infrastructure’s Gobo project have also created an infrastructure which aggregates content from across different platforms and allows users to customize their own “lenses,” choose their own algorithms for sorting the aggregated posts, and even use third-party services to assist in filtering and sorting. Instead of building an entire new middleware layer, slices of this idea also exist in the form of third-party plugins like Block Party, an anti-harassment tool designed specifically for X, formerly Twitter, (which is now on an indefinite hiatus after X placed its API behind a paywall).
The promise
Middleware promises to act as a broad-spectrum structural solution. Firstly, it could solve the intractable problem of content moderation. Setting bright-line content moderation rules which satisfy the diverse expectations of a large consumer base is extremely complex. Middleware facilitates the exercise of user autonomy to choose their own content moderation standards.
Secondly, it mitigates risks to the marketplace of ideas from social media market concentration by adding a new competitive editorial layer between the social media platform and the end users. Overhauling the structure of social media through middleware would offer a renewed opportunity for all stakeholders (including users, publishers, and advertisers) to renegotiate their power relations and attempt to correct entrenched power asymmetries.
Thirdly, middleware could also tap into the powers of the market to produce additional benefits. At present, social media platforms have very little incentive to disclose details about how their recommender algorithms function. In contrast, middleware providers would have to disclose much more information about their algorithms to distinguish themselves from competitors, paving the way for greater transparency and explainability.
Lastly, middleware’s proponents celebrate it as a flexible architectural solution that decentralizes control and enhances user autonomy. It avoids the piecemeal behavioral prescriptions that require continuous government monitoring common in antitrust cases, as seen in Google Search, and in new statutes like the European Union Digital Markets Act and Digital Services Act. Past experience shows that firms with strong economic incentives can almost always find technical loopholes and workarounds to comply with the text of the prescription while defeating its larger goals, making structural solutions necessary.
The perils
However, the optimism of middleware’s proponents often ignores that there are regulatory and institutional prerequisites necessary for its success. They take for granted the strengths of free-market solutions. However, our experience, especially with digital platforms, exposes the limits of a non-interventionist approach which presumes that competition will take care of itself provided markets are not straddled with inefficient regulation.
Indeed, proponents of older technologies, like the web browsers, search engines, blockchain, and social media, made similar utopian claims of self-regulating decentralization and deconcentration at the time of their inception. If middleware is to successfully promote competition and decentralization where its predecessors failed, its proponents must not fall into the same pitfalls and take into account the role of law and institutions in shaping the trajectory of technology. Julie Cohen’s and Amy Kapczynski’s seminal accounts of informational capitalism highlights the role of law in shaping the political economy of the digital sphere and in determining the asymmetries of power embedded in technologies. In the absence of complementary legal and institutional changes, middleware will meet the same fate as its predecessor technologies and veer toward a monopolized market.
How middleware succeeds
If middleware is to promote competition in the digital marketplace of ideas, regulators must pursue and maintain a three-pronged strategy.
1. Mandatory and uniform access to APIs
Regulators must require that social media platforms make available their APIs, so that third parties can create recommender algorithms that can integrate with the platform. Many social media platforms already provide access to APIs, but they often control the terms of this access opportunistically to destroy competitive threats (recall the recent arbitrary revocation of free access to Reddit’s and X’s APIs). Instead of relying on the largesse of social media platforms, mandatory and uniform API access to middleware providers should be enforced using existing regulatory frameworks like antitrust or other statutory mandates.
2. Structural separation
Even after imposing mandatory and uniform API access, platforms will continue to have strong economic incentives to enable their own recommender algorithm to integrate better or more seamlessly with the platform. Alternatively, users themselves might end up choosing the recommender algorithm of established social media platforms due to brand familiarity instead of giving new third parties a chance. Or the platform’s algorithm may be set as the default, resulting in default bias. Scholars like Tim Wu and Lina Khan have argued that actors who control the underlying platform infrastructure should be separated from those who control the venues of access or other layers of the information ecosystem. When the same entity has a stake in different layers of the platform, it produces an inherent conflict of interest. For example, Amazon, in its capacity as both the marketplace and a seller on its own marketplace, has the perverse incentive to self-promote its products. To avoid such conflicts, the social media platform handling content inventory should not be allowed to offer middleware, i.e., recommender services.
3. Empowering the affected stakeholders
These first two requirements will help create a marketplace for middleware but not prevent that market from tending towards consolidation due to network effects, efficiencies of scale (especially those associated with data), or even choice fatigue amongst users. Thus, in addition to reducing monopoly power, simultaneous attempts should be made to empower the groups affected by monopoly power in case some degree of consolidation eventually emerges. For instance, middleware providers might be incentivized to engage in even more extractive data practices to make their respective recommender algorithms more engaging or even addictive. We cannot rely alone on competitive free markets to automatically produce more pro-privacy alternatives. Instead, we need to secure concurrently the rights of consumers through data protection laws.
Similarly, while the introduction of a middleware layer might provide another opportunity to publishers and advertisers to renegotiate their power relations with the new platform intermediaries, such an opportunity would be wasted if we do not allow these groups to act collectively and exert some bargaining power. This might entail labor law reform or other provisions to facilitate more equitable bargaining between powerful platform intermediaries and affected groups.
Conclusion
Assuming that the introduction of middleware will alone solve all our problems associated with social media concentration is reductionist and suffers from historical amnesia. Discussions about creating a market for middleware must focus more on identifying and setting up regulatory guardrails to secure its success.
Author’s Disclosures: Madhavi Singh is a contributing editor at ProMarket. The author report no conflicts of interest. You can read our disclosure policy here.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
