Beyond the State of the Art in Social Media (and AI) -- Cruising Your "Vibes" with a Feed Mixer -- on "Bicycles for Our Minds"

Context: I have been thinking and writing about this future of "bicycles for our minds" for decades in ways that look well beyond the generally understood state of the art. Many of these ideas have flourished, and many are gaining recognition and being implemented, but many more remain largely unrecognized. This post highlights those that seem most important to the path forward. It assumes familiarity with the current state of the art -- and its discontents. 

Where should social media (and AI) be headed?

There is growing dissatisfaction and deepening concern about social media and its effects on people and society-- and its threats to democracy. But few really understand where we are -- and how we could be going in a far better direction. These same issues are also emerging for AI, as it fuses with social media by 1) incorporating socially user-created content and 2) being used in social media feeds and recommenders.  

Many long for a liberation of social media from the "enshittification" of centrally-controlled platforms -- and now with even greater urgency -- a counter to the incumbent platforms' capitulation to authoritarian influence that now threatens democracy, open discourse and the very foundations of human sense-making. 

The "ATmosphere" of Bluesky and the "Fediverse" of Mastodon, along with Project Liberty's DSNP and other similar efforts have gained attention as more decentralized and giving users more choice over how these powerful tools serve them. This shift to a "federated universe" of interoperating systems can better serve the context- and norm-specific needs of discourse among individuals and and the diverse communities they participate in. Even Meta has given a partial nod to this trend -- by federating Threads with the Activity Pub protocol of Mastodon and other services -- thus edging toward what is better described as a "pluriverse".

Context: I draw heavily on Bluesky and AT protocol and the framing of the Free Our Feeds initiative to solidify its openness -- as currently farthest along in providing for the multidimensionality that will underlie a full-function pluriverse. But those directional ideas apply equally well to the fediverse of Mastodon and other ActivityPub-connected systems -- and to Project Liberty's Distributed Social Network Protocol -- and to other current and future protocols and services with similar objectives that might integrate and harmonize (or supplant) these early shoots of growth in a better direction. 

Global networks, insularity, and the problem of "vibe"

All social media services currently have issues of what communities they help us assemble and participate in, and of what norms apply to them. The dominant global platforms suffer from "context collapse," bringing  diverse communities into collisions without sufficient context to avoid misunderstandings and polarization. The more decentralized pluriverse seeks to avoid that by empowering more community context. Bluesky with its AT protocol has pulled into the lead over the Mastodon fediverse by combining user choice with ease of use, openness, flexibility, and extensibility, reaching over 35 million users. However, its appeal has been limited by its reputation as being dominated by liberals (driven from X/Twitter) and for how some perceive its "vibe." People are wondering where to turn for an online experience the offers the people and vibe they want.

The answer is in a deeper vision of what these "bicycles for our minds" can do, and how we must allow time for us to shape these still formative tools -- and to learn to manage how they shape us. What we see now is just the infancy of a radically new medium that is subsuming all media. This post (adapted from an earlier post) offers a vision of how this infant that just barely crawls, will grow into the nimble bicycle that Steve Jobs had in mind (based on how human mobility was far less efficient than many animals, but how a human on a bicycle could travel far more efficiently than the most efficient animal, a condor.) Our trick is that humans are tool makers, but our downfall might be when the tools makers build tools to serve themselves, and not those who use them.

While there is obvious need to develop near-term features to make each competing tech platform and universe of platforms into more efficient tools, there is also a need to articulate long-term objectives that many tools can build toward to be not only efficient, but effective in serving us, as their users. We are re-engineering human discourse for the online era -- that will be a long process -- but without a long-term vision of how our tools for discourse should work, and how we want to use them, it will be longer and more problematic.

Key ideas 

Technically: A "feed mixer" is a key missing layer. There is much current discussion of online service feeds -- the good, the bad, and the ugly, and the hurdle of ease of use as limiting greater user control -- but little recognition of the need for a user-controlled feed mixer. That would serve to simplify the combination of 1) handlebars for steering our bicycle, and 2) pedals, brakes, and gear shifts for controlling its speed and responsiveness. With flexible control of our feeds that orchestrates multiple algorithms and works across whatever networks we participate in, it will be easy to tune into whatever vibe we want. This post explains and puts that in context -- along with other layers that have been generally ignored.

Sociotechnically: Individual agency over feeds and other details is only #1 of three essential pillars. Neglected are #2, the "Social Mediation Ecosystem" that our ideas get mediated by, and #3 the Reputation Systems that determine whose mediating efforts we trust. All three synergize to help humans, as individuals in an open society, to refine and apply our unique collective intelligence and human values to make sense of the world and flourish. As Marshall McLuhan and his colleagues said, “We shape our tools and thereafter our tools shape us.” Modern liberal society has been powerfully shaped by print and broadcast media. Now we must again relearn -- to reshape how society adapts -- and how we shape online media to manage its far greater power, reach, and speed.

Further TL;DR of the pluriverse, as I envision it 

• The move to decentralization, federation, and on toward the rich diversity of the pluriverse, reflects the realization that human society is far too complex, diverse, and nuanced to be served by any one centrally managed global "public square."
  
  
• 
  However, current steps toward decentralization will need to better support the hyperlinked multidimensionality of how individuals and communities interconnect. These communities reflect a diversity of interests, values, and norms. But IRL (In Real Life) individuals participate in many communities. They are rarely bound by any one community, and wish to have global views into many, as both speakers and listeners, depending on their interests, goals, and moods as they vary from time to time. Ted Nelson invented hypertext because "everything is deeply intertwingled."
  
  
• Users will inevitably need multi-homing tools that give variable "lenses" for looking into and participating in many communities. Cross-community feeds and recommenders will be essential for individuals to navigate the abundance of riches in the pluriverse to meet their needs and find their vibe. This may work at at least two levels: 1) low-level recommenders for up- or down-ranking ranking feed items based on specific objectives, and 2) higher-level UX tools for composing and steering mixes of lower level rankings into a consolidated feed.
  
  
• 
  Think of that higher level UX tool as a feed mixer. Just as a music mixing console takes in many individual sound tracks and blends them into a dynamically orchestrated, multi-dimensional composition, an information feed mixer should do the same for individual sub-feeds. Music mixers take tracks from voices, instruments, and other listening points, then adjust overall volume, apply tonal adjustment effects or filters, and balance the levels of each track in the mix. This may be controlled by a specialized mixing operator (an agent) -- often using selectors, knobs, and slider controls.
  
  
• Before objecting that such a tool would be too hard and time-consuming for lazy users to master, consider how feed services can be branded, and how that can make it easy for users to grasp a brand identity -- who is included with what vibe -- and mix feeds based on that intuition of a vibe. That is how we select CNN or Fox or MSNBC or PBS without studying a specification of their editorial curation policies. 
  
  
• Bluesky seems farthest along in pointing to this multidimensionality in our feeds, providing for (but still in early stages of implementing) tools for separating the "speech layer" from the "reach layer" as described in their early blog posts on Composable Moderation, Moderation in a Public Commons, and Algorithmic Choice. My more detailed post from 6/23 suggests directions for taking that farther.
  
  
• Mastodon seems to also be moving in that general direction, with discussion of a cross-instance groups structure, and shared moderation services that address the challenges of administering small communities, but seems to prefer to remain relatively insular. I suggest they can have both, making their communities semi-permeable. 
  
  
• A similar effort by Project Liberty also has some traction (and significant funding from Frank McCourt) and a vision that seems similar to that of AT Protocol, instead based on Distributed Social Network Protocol (DSNP).
  
  
• The objective should be for all of these -- as well as services using alternative decentralized protocols and current closed platforms -- to harmonize to allow users to seamlessly participate in an integrated "pluriverse" with a multi-homing feed mixer. My recent discussions with activists from all three of these current efforts shows a shared recognition of the need to converge from disparate silos to a true pluriverse with high interoperability.
  
  
• The vision I suggest will take time to build, develop, and be fleshed out by users, but to get where we will want to go in the future will require having these ideas in mind as we architect and build toward that vision. But even given the limits of our imagination, the beauty of open interoperability is that it supercharges the ability to markets to innovate. Just consider how the open interoperation of smartphone apps enabled the growth of a vibrant ecosystem far beyond what Apple or Google could ever provide by themselves. ...Or how the openness of the web took us far beyond what AOL or CompuServe could offer.

Summarizing key elements and features of the vision

Sections of my older and longer post are summarized and updated here. (Serious readers may wish to look to some fuller explanations there):

• Hypercommunities
  Each person can be a member of many communities (/groups) at once, as many layers of overlapping Venn diagrams in many dimensions -- shifting our view and level of participation as desired.
  
  
• Ranking as the core task
  Nearly all "moderation" and "curation" recommendations boil down to ranking. Downranking can provide safety from bad content, and upranking can bubble up quality and value. Composability of ranking tools can work at both individual and community levels to blend a mix of rankings that draw on the wisdom of each community. ("Moderation as removal" should instead be effected by "downranking with extreme prejudice" that ensures items will not appear in feeds, but may remain accessible by direct request, subject to appropriate restrictions on access to illegal content)
  
  
• Feed Mixing Agent Services as a core tool -- User-selectable, multilevel feed composition composed from multiple algorithms
  A truly composable, steerable feed would provide a higher level feed mixer interface that lets each of us easily manage and merge a user-selected mix of lower-level feeds, with user-defined relative weights. A steerable feed would allow those mixes and weights to be easily changed at will to suit our varying tasks and moods, including options for stored pre-sets. This would restore user agency to choose and orchestrate from an open market in independent attention agent services -- providing choices of UXs, algorithms, and human mediation providers. 
  
  
• Multi-dimensional reputation based on explicit and/or implicit signals
  Wiser use of algorithms is needed -- not to replace human wisdom, but to distill it based on human judgments and reputations as judged by other humans, all under user control. I view reputation as essential to making ranking work well, and have written frequently about “rate the raters and weight the ratings” as an extension of Google’s PageRank algorithm to develop a socially derived and reputation-weighted reputation. Reputations have multiple dimensions, including subject domain, value systems, and community context, and change over time, being slow to develop, but easy to lose. An effective reputation system motivates individuals to seek and maintain a good reputation.
  
  
• Support for rebuilding our Social Mediation Eecosystem
  Communities and mediating services can be decoupled. The speech layer may be more tightly tied to specific communities than the reach layer. Real-life communities and institutions may be re-enabled to mediate our online discourse, both for their direct membership and those who wish to follow them. The ecosystem that shaped and stabilized discourse in the real world should be reconstituted in the virtual world. This mediation ecosystem shapes how messages flow and evolve, interacting and synergizing with both user agency and reputation, as discussed more fully in Tech Policy Press, Three Pillars of Human Discourse (and How Social Media Middleware Can Support All Three).
  
  
• Classification/labelling and ranking
  Rankings can be based on many dimensions of attributes -- so rankings could take a hybrid form that includes classification or label attributes. Adding a quantifier for the strength of a classification/label (how strongly positive or negative it might be) would ultimately be essential to achieving nuance, and could also include quantification of the rater's confidence level in that value rating.
  
  
• Broader issues and federation/subsidiarity in labeling and ranking
  
  
  Our notions of truth and value -- and authority about that -- are contingent, changeable, and heavily influenced by our broader social mediation ecosystem. That has been central to the generative success of human society. Thus our social media should reflect that social contingency, and provide for a high degree of subsidiarity in how decisions are made. That is the essence of what I call freedom of impression, and how it serves to balance freedom of expression. 
  
  
• Further thoughts on the federated architecture
  There is need for algorithmic choice at multiple levels. At a lower level is an open market in basic algorithms with very specific objective functions in terms of subjects, values, and vibes/moods. At a higher level is an open market in UX-level services that enable composition and orchestration of those lower level algorithmic rankings to present an consolidated view that blends multiple objective functions, and to allow steering that view dynamically as the user's moods and needs change.  
  
  
• Enabling subsidiarity of "moderation" of the "lawful but awful"
  Federation is based on the principle of subsidiarity: that idea that most moderation/mediation decisions should be local, to best reflect relevant local/community interests, values, and norms. This would apply a nuanced blend of top-down controls to limit dissemination of the truly unlawful (with trust and safety teams, tools, and services), along with mostly bottom-up tools and services to manage more contingent (context-, value-, and norm-dependent) levels of awfulness -- and goodness! -- in multiple dimensions. This should apply at the level of 1) membership communities (servers/instances plus other communities/groups) and 2) cross-community attention/mediation agent services that users choose to opt into. (Given the role of Mastodon instance operators as "benevolent dictators," the current ActivityPub "fediverse" is really more of a "confediverse." The ATmosphere (of AT protocol) seems more supportive of the nuanced multidimensional division of control in federation.)
  
  
• 
  "Vibe"--  seeking "the shmoo of social media"
  There is much talk of the "vibe" of different platforms, but "we ain't seen nothin' yet." With selectable, composable feeds, users will be able to create views that tune into whatever vibe they want (with whatever levels of moderation they want). This is the infancy of a flexible new social ecosystem online, and whatever initial vibe chaos we might see now will give way to a new order of shaping a vibe style and viewpoint, and tuning into it. A fully functional social media pluriverse will be a "shmoo" (a classic cartoon creature that tastes like whatever you want) -- with diverse communities, but flexible lenses into as many as desired. This provides a level of flexibility and user control of their experience that will grow in importance as the pluriverse grows in scale and diversity and in the richness of interconnections desired by users with many interests and moods for diverse vibes.

These levels of choice may seem complex and overwhelming, but just as we easily choose what mix of CNN or Fox or MSNBC we want at any given time, branded middleware services could make that easy -- as I outline with an example for the NY Times. And consider that easily "channel surfed" linear TV channels have been largely replaced by an enshittified kludge of streaming walled gardens. The streaming platforms have resisted enabling a simple high level feed mixer user interface that crosses programs and streaming services. We might see that open to a better high-level user experience specific to video, but as social media eat the world, we might better hope to see such innovation applied to all of our information feeds, of all modalities.

Much of this flexible multidimensionality will emerge slowly, as technical, human, and social infrastructures co-evolve toward it -- a whole-of-society, socio-technical process that will take decades, and may be very disruptive for a time (much as the era of warfare related to society's sociotechnical absorption of Gutenberg's printing press). But if we do not plan for what we can foresee, and build for extensibility to what we do not yet foresee, it will be even harder to find a path toward a new stability that is robust and generative.

My related works are listed on the Selected Items tab of this blog.

Original Branding

 This vision of social media future is meant to complement and clarify the vision behind many of my other works (such as this, see list of selected pieces at the end). It assumes you have come here after seeing at least one of those (but includes enough background to also be read first).

Business opportunity – start now, and grow from there:

●     Managers of the NY Times, small local news services, or any other organization that has built a strong community can use the following model to build a basic online middleware service business, starting now.

●     For example, Bluesky could be a base platform for building initial proof-of-concept services along these lines that could develop and grow into a major business.

[If you are impatient, jump to the section on "Branding"]

It is clear that social media technology is not serving social values well. But it is not so clear how to do better. I have been suggesting that the answer begins in learning from how we, as a society, curated information flows offline. (These issues are also increasingly relevant to emerging AI.)

This piece envisions how an offline curation “brand” with an established following – like the New York Times, or many others, including non-commercial communities of all kinds – could extend their curatorial influence, and the role of their larger community, more deeply into the digital future of thought. (Of course, much the same kind of service can be built as a greenfield startup, as well, but having an established community reduces the cold-start problem.)

Building on middleware – the Three Pillars

I and many others have advocated for “middleware” services, a layer of enabling technology that sits between users and platforms to give control back to users over what goes into each of our individual feeds. But that is just the start of how that increased user agency can support healthy discourse and limit fragmentation and polarization in our globally online world.

 The pillars I have been writing about are:

1. 
   Individual agency, the starting point of democratic free choice over what we say to whom, what individuals we listen to, and what groups we participate in.
2. Social mediation, the social processes, enabled by an ecosystem of communities and institutions of all kinds that influence and propagate our thoughts, expression, and impression. (For simple background, see What Is a Social Mediation Ecosystem?)
3. Reputation, the quality metrics, intuitively developed and shared to decide which individuals and communities are trustworthy, and thus deserve our attention (or our skepticism).

Middleware can sit on top of our basic social networking platforms to support the synergistic operation of all three pillars, and thus help make our discourse productive.

In the offline world of open societies, there is no single source of “middleware” services that guide us, but an open, organic, and constantly adjusted mix of many sources of collective support. People grow up learning intuitively to develop and apply these pillars in ever-changing combinations.

Software is far more rigid than humans. Online middleware is a technique for enabling the same kind of diversity and “interoperation” – of attention agent services for us to choose from, and to help groups fully participate in them – so we can dynamically compose the view of the world we want at any point in time.

Bluesky currently offers perhaps the best hint at how middleware services will be composed, steered, and focused – as our desires, tasks, and moods change. Just keep in mind that current middleware offerings are still just infants learning to crawl.

As we may think …together

Vannevar Bush provided a prescient vision of the web in 1945 (yes, 1945!) – in his Atlantic article “As We May Think.” Its technology was quaint, but the vision of how humans can use machines to help us think was very on-point, and inspired the creation of the web. Now it is time for a next level vision – of how we may think together – even if the details of that vision are still crude.

Current notions of middleware have been focused primarily on user agency, and just beginning (as in Bluesky) to consider how we need not just a choice of a single middleware agent service, but to flexibly compose and steer among many attention agent services. Steve Jobs spoke of computers as “bicycles for our minds.” As we conduct our discourse, middleware-based attention agent services can give us handlebars to steer them and gear shifts to deal with varying terrain and motivations. They can give us “lenses,” for focusing what we see from our bicycles.

To build out this capability, we will need at least two levels of user-facing middleware services:

●     Many low level service agents that curate for specific objectives of subject domain, styles, moods, sources, values, and other criteria.

●     One or more high level service agents that make it easy to orchestrate those low level agents, as we steer them, shift gears, and change our focus, creating a consolidated ranking that gives us what we want, and screens out what we do not want, at any given time.

Just how those will work will change greatly over time as we learn to drive these bicycles, and providers learn to supply useful services – “we shape our tools and our tools shape us.” Emerging AI in these agents will increase the ease of use, and the usable power of the bicycles – but even in the age of AI, the primary intelligence and judgment must come from the humans that use these systems and create the terrain of existing and new information and ideas (not just mechanically reassembled tokens of existing data) that we steer through.

====================================================
Here is the business opportunity:
====================================================

Branding – a “handle” for intuitively easy selection  -- and signaling value

Yes, choosing middleware services seems complicated, and skeptics rightly observe that most users lack the skill or patience to think very hard about how to steer these new bicycles for our minds. But there are ways to make this easy enough. One of the most promising and suggestive is branding – a powerful and user-friendly tool for reliably selecting a service to give desired results. Take the important case of news services:

●     If we try to select news stories at the low level of all the different dimensions of choice – subject matter, style, values, and the like – of course the task would be very complex and burdensome.

●     But many millions easily choose what mix of CNN, MSNBC, Fox News, PBS, or less widely used brands they want to watch at any time. The existing brand equity and curation capabilities of such media enterprises are now being squandered by digital platforms that offer such established service brands only rudimentary integration into their social media curation processes. With proper support, both established and new branded middleware services can establish distinctive sensibilities that can make choice easy.

Importantly, branding also serves marketing and revenue functions in powerful ways that can be exploited by middleware services. Once established and nurtured, a brand attracts users on the basis that it offers known levels of quality, and as catering to selective interests and tastes. "It's Not TV, It's HBO" encapsulated the power of HBO's brand in the heyday of premium TV.

The New York Times as a branded curation community: 

Consider the New York Times as just one example of branded curation middleware that could serve as a steerable lens into global online discourse. It could just as well be News Corp, CNN, Sports Illustrated, or Vogue – or your local newspaper (if you still have one!) – or your town or faith community, a school, a civil society organization, a political party, a library, a bowling league – or whatever group or institution that wants to support its uniquely focused (but overlapping and not isolated) segment of the total social mediation ecosystem.

Consider how all three pillars can work and synergize in such a service:

User agency comes in by our participation as readers, and as speakers in any relevant mode – posts, comments, likes, shares, letters to the editor, submissions for Times publications. This can be addressed at at least two levels:

●     Low level attention service agents that find and rank candidate items for our feeds and recommenders. This is much as we now choose from an extensive list of available email newsletters from the Times.

●     Higher level middleware composing agents would help compose these low-level choices – and facilitate interoperation with similar services from other communities – to build a composite feed of items from the Times and all our other chosen sources. They could offer sliders to decide what mx to steer into a feed at any given time, and saved presets to shift gears for various moods, such as news awareness/analysis, sports/entertainment, challenging ideas, light mind expansion, and diversion/relaxation.

(Different revenue models may apply to different services, levels, and modes of participation, just as some NY Times features now may cost extra.)

Social mediation processes come in to our user interface at two levels of curation:

●     User-driven curation: Much like current platforms, the Times low-level services can rank items based on signals from the community of Times users – their likes, shares, comments, and other signals of interest and value. This might distinguish subscribers versus non-subscribing readers. Subscribers might be more representative of the community, but non-subscribers might bring important counterpoints. Other categories could include special users, such as public figures in various political, business, or professional categories. As such services mature, these signals can be expanded in variety to be far more richly nuanced, such as to give clearer feedback and be categorized by subject domains of primary involvement. 

●     Expert-driven curation: The Times editorial team can be drawn on (and potentially augmented with supportive levels of AI) to provide high quality expert curation services in much the same way, in whatever mix desired. This could include both their own contributions, and their reactions to readers’ contributions.

Reputation systems that keep score of quality and trust feedback on both users and content items – that arise from those mediation processes – can also be valuably focused on the Times community:

●     At a gross level, we might make gross assumptions that differentiate the editorial and journalism staff, subscribers, and non-subscribing readers (as part of the basic mediation process), but a reputation system could distinguish among very different levels of reputation for quality of participation in many dimensions, such as expertise, judgment, clarity, wisdom, civility, and many more – in each of many subject domains.

●     Reputation systems might also be tuned to Times reporters and editors, and their inputs to reputations of content items and users. But the true power of this kind of service is its crowdsourcing from not just the Times staff, but from its unique extended community. One could choose to ignore the staff, and just turn their lens on the community, or vice versa.

Enterprise-class community support integration – and simple beginnings

To fully enable this would require new operational support services that integrate the operation of open online social media platform services (like Bluesky now, or maybe someday Threads) with the operations of the Times. As the technology for multi-group participation is built out beyond current rudimentary levels, it can integrate with the operation of each group, including the enterprise-class systems that drive the operations of the Times. This might include the kind of functionality and integration offered by CRM (customer relationship management) systems for managing all of the Times’ interactions with its customers, as well as the CMS (content management system) used to manage its journalism content, and the SMS (subscription management systems) that manage revenue operations.

Doing all of this fully will take time and effort – but some of it could be done relatively easily, such as in an attention agent that ranks items based on the Times community members signals as distinct from those of the general network population. The Times could begin a trial of this in the near term by exploiting the basic middleware capabilities already available by creating a Bluesky server instance (using the open Bluesky server code and interoperation protocols) and their own custom algorithms. 

A large, profitable (or otherwise well-funded) business like the Times could develop and operate middleware software itself (if the social media platform allows that, as Bluesky does), but smaller organizations might need a shared “middleware as a service” (MaaS) software and operations provider to do much of that work.

A user steered, intuitively blended, mix of diverse sub-community feeds

Even at a basic level, imagine how doing this for many such branded ecosystem groups could enable users to easily compose feeds that bring them a diverse mix of quality inputs, and to steer and adjust the lenses in those feeds and searches to focus our view as we desire, when we desire.

Similar middleware services could be based all kinds of groups – for example:

●     Local news and community information services – much like the Times example, for where you live now, used to live, or want to live or visit.

●     Leadership and/or supporters of political parties or civil society organizations – issues, platforms/policies, campaigns, turnout, surveys, fact-checking, and volunteering.

●     Professional and/or amateur players and/or coaches for sports – catering to teams, fans, sports lore, and fantasy leagues.

●     Faculty, students, and/or alumni from universities – selecting for students, faculty, alumni, applicants, parents.

●     Librarians and/or card holders for library systems – selecting for discovery, reading circles, research, criticism, and authors.

●     Leaders and/or adherents to faith communities – for community news, personal spiritual issues, and social issues.

Consider how the Times example translates to and complements any of these other kinds of groups (most easily if enabling software is made available from a SaaS provider). Users could easily orchestrate their control over diverse sources of curation and moderation – selecting from brands with identities they recognize – without requiring the prohibitive cognitive load of controlling all the details that critics now argue would doom middleware because few would bother to make selections. New brands can also emerge and gain critical mass, using this same technology base.

By drawing on signals from expert and/or ordinary members of groups that have known orientations and norms, users might easily select mixes that serve their needs and values – and shift them as often as desired.

Context augmentation

Peter Steiner in The New Yorker

"On the Internet, no one knows you are a dog" -- or a lunatic, or a bot. Famously observed by Peter Steiner's 1993 cartoon, this became known as "context collapse," broadly understood as a core reason why internet discourse is so problematic. Much of the meaning derives from context external to the message itself -- who is speaking to whom, from and to what community, with what norms and assumptions. That has largely been lost in current social media (and in emerging AIs). 

Consider how the kind of social mediation ecosystem processes envisioned here differ from what current major platforms offer in the way of community support -- and thus fail to provide essential context: 

• They let you create a personal set (a unidirectional pseudo-community) of friends or those you follow, but increasingly focus on engagement-based ranking into feeds -- because they want to maximize advertising revenue, not the quality of your experience. 
• They rank based on likes, shares, and comments from a largely undifferentiated global audience, with little opportunity for you to influence who is included. 
• They may favor feedback from rudimentary "groups" that you join, but provide very limited support to organizers and members to make those groups rich and cohesive. 
• They may cluster you into what they infer to be your communities of interest, but with out any agency from you over which groups those are, except for the rudimentary "groups" you join.
• And, even if they did want to serve your objectives, not theirs, they would be hard-pressed to come anywhere near the richness and diversity of truly independent, opt-in, community-driven middleware services that are tailored to diverse needs, contexts, and sustaining revenue models.

Doing moderation the old-fashioned way – enabled by middleware

Instead of being seen as a magical leap in technology, or an off-putting cognitive burden on users, middleware can be understood as a way to recreate in digital form the formal and informal social structures people have enjoyed for centuries – individually composed interaction with the wisdom of organically evolved social mediation ecosystems and intuitive informal reputation systems.

What at first seems complicated, from the perspective of current social media, is at core, little more complicated than the structure of traditional human discourse – building on key functions and elements of the social mediation and reputation ecosystems – all legitimized by choices of individual agency. Yes, that is complicated, but humans have learned over millennia to intuitively navigate this traditional web of communities and reputations. Yes, make it as simple as possible, but no simpler!

Creating an online twin of such a web of community ecosystems will not happen overnight, but many industries have already built out online infrastructures of similar complexity – in finance, manufacturing, logistics, travel, and e-commerce. Middleware is just a tool for enabling software systems to work together in ways similar to what humans (and groups of humans) do intuitively. The time to start rebuilding those ecosystems is now.

____________________

Related works:

●     Overview on Tech Policy Press -- Three Pillars of Human Discourse (and How Social Media Middleware Can Support All Three).

●     My November 2023 post introducing the pillars framing – A New, Broader, More Fundamental Case for Social Media Agent "Middleware" – introduced the Three Pillars framing, and embeds a deck that adds details and implication not yet fully addressed elsewhere.

●     Core ideas addressed more formally in my April 2024 CIGI policy brief, New Logics for Governing Human Discourse in the Online Era.

●     Very simply -- What Is a Social Mediation Ecosystem? (and Why We Need to Rebuild It). 

●     Other related works are listed on my blog.

Jump link Tag test

This vision of social media future is meant to complement and clarify the vision behind many of my other works (such as this, see list of selected pieces at the end). It assumes you have come here after seeing at least one of those (but includes enough background to also be read first).

Business opportunity – start now, and grow from there:

●     Managers of the NY Times, small local news services, or any other organization that has built a strong community can use the following model to build a basic online middleware service business, starting now.

●     For example, Bluesky could be a base platform for building initial proof-of-concept services along these lines that could develop and grow into a major business.

[If you are impatient, jump to the section on "Branding"]

It is clear that social media technology is not serving social values well. But it is not so clear how to do better. I have been suggesting that the answer begins in learning from how we, as a society, curated information flows offline. (These issues are also increasingly relevant to emerging AI.)

This piece envisions how an offline curation “brand” with an established following – like the New York Times, or many others, including non-commercial communities of all kinds – could extend their curatorial influence, and the role of their larger community, more deeply into the digital future of thought. (Of course, much the same kind of service can be built as a greenfield startup, as well, but having an established community reduces the cold-start problem.)

Building on middleware – the Three Pillars

I and many others have advocated for “middleware” services, a layer of enabling technology that sits between users and platforms to give control back to users over what goes into each of our individual feeds. But that is just the start of how that increased user agency can support healthy discourse and limit fragmentation and polarization in our globally online world.

 The pillars I have been writing about are:

1. 
   Individual agency, the starting point of democratic free choice over what we say to whom, what individuals we listen to, and what groups we participate in.
2. Social mediation, the social processes, enabled by an ecosystem of communities and institutions of all kinds that influence and propagate our thoughts, expression, and impression. (For simple background, see What Is a Social Mediation Ecosystem?)
3. Reputation, the quality metrics, intuitively developed and shared to decide which individuals and communities are trustworthy, and thus deserve our attention (or our skepticism).

Middleware can sit on top of our basic social networking platforms to support the synergistic operation of all three pillars, and thus help make our discourse productive.

In the offline world of open societies, there is no single source of “middleware” services that guide us, but an open, organic, and constantly adjusted mix of many sources of collective support. People grow up learning intuitively to develop and apply these pillars in ever-changing combinations.

Software is far more rigid than humans. Online middleware is a technique for enabling the same kind of diversity and “interoperation” – of attention agent services for us to choose from, and to help groups fully participate in them – so we can dynamically compose the view of the world we want at any point in time.

Bluesky currently offers perhaps the best hint at how middleware services will be composed, steered, and focused – as our desires, tasks, and moods change. Just keep in mind that current middleware offerings are still just infants learning to crawl.

As we may think …together

Vannevar Bush provided a prescient vision of the web in 1945 (yes, 1945!) – in his Atlantic article “As We May Think.” Its technology was quaint, but the vision of how humans can use machines to help us think was very on-point, and inspired the creation of the web. Now it is time for a next level vision – of how we may think together – even if the details of that vision are still crude.

Current notions of middleware have been focused primarily on user agency, and just beginning (as in Bluesky) to consider how we need not just a choice of a single middleware agent service, but to flexibly compose and steer among many attention agent services. Steve Jobs spoke of computers as “bicycles for our minds.” As we conduct our discourse, middleware-based attention agent services can give us handlebars to steer them and gear shifts to deal with varying terrain and motivations. They can give us “lenses,” for focusing what we see from our bicycles.

To build out this capability, we will need at least two levels of user-facing middleware services:

●     Many low level service agents that curate for specific objectives of subject domain, styles, moods, sources, values, and other criteria.

●     One or more high level service agents that make it easy to orchestrate those low level agents, as we steer them, shift gears, and change our focus, creating a consolidated ranking that gives us what we want, and screens out what we do not want, at any given time.

Just how those will work will change greatly over time as we learn to drive these bicycles, and providers learn to supply useful services – “we shape our tools and our tools shape us.” Emerging AI in these agents will increase the ease of use, and the usable power of the bicycles – but even in the age of AI, the primary intelligence and judgment must come from the humans that use these systems and create the terrain of existing and new information and ideas (not just mechanically reassembled tokens of existing data) that we steer through.

====================================================
Here is the business opportunity:
====================================================

Branding – a “handle” for intuitively easy selection  -- and signaling value

Yes, choosing middleware services seems complicated, and skeptics rightly observe that most users lack the skill or patience to think very hard about how to steer these new bicycles for our minds. But there are ways to make this easy enough. One of the most promising and suggestive is branding – a powerful and user-friendly tool for reliably selecting a service to give desired results. Take the important case of news services:

●     If we try to select news stories at the low level of all the different dimensions of choice – subject matter, style, values, and the like – of course the task would be very complex and burdensome.

●     But many millions easily choose what mix of CNN, MSNBC, Fox News, PBS, or less widely used brands they want to watch at any time. The existing brand equity and curation capabilities of such media enterprises are now being squandered by digital platforms that offer such established service brands only rudimentary integration into their social media curation processes. With proper support, both established and new branded middleware services can establish distinctive sensibilities that can make choice easy.

Importantly, branding also serves marketing and revenue functions in powerful ways that can be exploited by middleware services. Once established and nurtured, a brand attracts users on the basis that it offers known levels of quality, and as catering to selective interests and tastes. "It's Not TV, It's HBO" encapsulated the power of HBO's brand in the heyday of premium TV.

The New York Times as a branded curation community: 

Consider the New York Times as just one example of branded curation middleware that could serve as a steerable lens into global online discourse. It could just as well be News Corp, CNN, Sports Illustrated, or Vogue – or your local newspaper (if you still have one!) – or your town or faith community, a school, a civil society organization, a political party, a library, a bowling league – or whatever group or institution that wants to support its uniquely focused (but overlapping and not isolated) segment of the total social mediation ecosystem.

Consider how all three pillars can work and synergize in such a service:

User agency comes in by our participation as readers, and as speakers in any relevant mode – posts, comments, likes, shares, letters to the editor, submissions for Times publications. This can be addressed at at least two levels:

●     Low level attention service agents that find and rank candidate items for our feeds and recommenders. This is much as we now choose from an extensive list of available email newsletters from the Times.

●     Higher level middleware composing agents would help compose these low-level choices – and facilitate interoperation with similar services from other communities – to build a composite feed of items from the Times and all our other chosen sources. They could offer sliders to decide what mx to steer into a feed at any given time, and saved presets to shift gears for various moods, such as news awareness/analysis, sports/entertainment, challenging ideas, light mind expansion, and diversion/relaxation.

(Different revenue models may apply to different services, levels, and modes of participation, just as some NY Times features now may cost extra.)

Social mediation processes come in to our user interface at two levels of curation:

●     User-driven curation: Much like current platforms, the Times low-level services can rank items based on signals from the community of Times users – their likes, shares, comments, and other signals of interest and value. This might distinguish subscribers versus non-subscribing readers. Subscribers might be more representative of the community, but non-subscribers might bring important counterpoints. Other categories could include special users, such as public figures in various political, business, or professional categories. As such services mature, these signals can be expanded in variety to be far more richly nuanced, such as to give clearer feedback and be categorized by subject domains of primary involvement. 

●     Expert-driven curation: The Times editorial team can be drawn on (and potentially augmented with supportive levels of AI) to provide high quality expert curation services in much the same way, in whatever mix desired. This could include both their own contributions, and their reactions to readers’ contributions.

Reputation systems that keep score of quality and trust feedback on both users and content items – that arise from those mediation processes – can also be valuably focused on the Times community:

●     At a gross level, we might make gross assumptions that differentiate the editorial and journalism staff, subscribers, and non-subscribing readers (as part of the basic mediation process), but a reputation system could distinguish among very different levels of reputation for quality of participation in many dimensions, such as expertise, judgment, clarity, wisdom, civility, and many more – in each of many subject domains.

●     Reputation systems might also be tuned to Times reporters and editors, and their inputs to reputations of content items and users. But the true power of this kind of service is its crowdsourcing from not just the Times staff, but from its unique extended community. One could choose to ignore the staff, and just turn their lens on the community, or vice versa.

Enterprise-class community support integration – and simple beginnings

To fully enable this would require new operational support services that integrate the operation of open online social media platform services (like Bluesky now, or maybe someday Threads) with the operations of the Times. As the technology for multi-group participation is built out beyond current rudimentary levels, it can integrate with the operation of each group, including the enterprise-class systems that drive the operations of the Times. This might include the kind of functionality and integration offered by CRM (customer relationship management) systems for managing all of the Times’ interactions with its customers, as well as the CMS (content management system) used to manage its journalism content, and the SMS (subscription management systems) that manage revenue operations.

Doing all of this fully will take time and effort – but some of it could be done relatively easily, such as in an attention agent that ranks items based on the Times community members signals as distinct from those of the general network population. The Times could begin a trial of this in the near term by exploiting the basic middleware capabilities already available by creating a Bluesky server instance (using the open Bluesky server code and interoperation protocols) and their own custom algorithms. 

A large, profitable (or otherwise well-funded) business like the Times could develop and operate middleware software itself (if the social media platform allows that, as Bluesky does), but smaller organizations might need a shared “middleware as a service” (MaaS) software and operations provider to do much of that work.

A user steered, intuitively blended, mix of diverse sub-community feeds

Even at a basic level, imagine how doing this for many such branded ecosystem groups could enable users to easily compose feeds that bring them a diverse mix of quality inputs, and to steer and adjust the lenses in those feeds and searches to focus our view as we desire, when we desire.

Similar middleware services could be based all kinds of groups – for example:

●     Local news and community information services – much like the Times example, for where you live now, used to live, or want to live or visit.

●     Leadership and/or supporters of political parties or civil society organizations – issues, platforms/policies, campaigns, turnout, surveys, fact-checking, and volunteering.

●     Professional and/or amateur players and/or coaches for sports – catering to teams, fans, sports lore, and fantasy leagues.

●     Faculty, students, and/or alumni from universities – selecting for students, faculty, alumni, applicants, parents.

●     Librarians and/or card holders for library systems – selecting for discovery, reading circles, research, criticism, and authors.

●     Leaders and/or adherents to faith communities – for community news, personal spiritual issues, and social issues.

Consider how the Times example translates to and complements any of these other kinds of groups (most easily if enabling software is made available from a SaaS provider). Users could easily orchestrate their control over diverse sources of curation and moderation – selecting from brands with identities they recognize – without requiring the prohibitive cognitive load of controlling all the details that critics now argue would doom middleware because few would bother to make selections. New brands can also emerge and gain critical mass, using this same technology base.

By drawing on signals from expert and/or ordinary members of groups that have known orientations and norms, users might easily select mixes that serve their needs and values – and shift them as often as desired.

Context augmentation

Peter Steiner in The New Yorker

"On the Internet, no one knows you are a dog" -- or a lunatic, or a bot. Famously observed by Peter Steiner's 1993 cartoon, this became known as "context collapse," broadly understood as a core reason why internet discourse is so problematic. Much of the meaning derives from context external to the message itself -- who is speaking to whom, from and to what community, with what norms and assumptions. That has largely been lost in current social media (and in emerging AIs). 

Consider how the kind of social mediation ecosystem processes envisioned here differ from what current major platforms offer in the way of community support -- and thus fail to provide essential context: 

• They let you create a personal set (a unidirectional pseudo-community) of friends or those you follow, but increasingly focus on engagement-based ranking into feeds -- because they want to maximize advertising revenue, not the quality of your experience. 
• They rank based on likes, shares, and comments from a largely undifferentiated global audience, with little opportunity for you to influence who is included. 
• They may favor feedback from rudimentary "groups" that you join, but provide very limited support to organizers and members to make those groups rich and cohesive. 
• They may cluster you into what they infer to be your communities of interest, but with out any agency from you over which groups those are, except for the rudimentary "groups" you join.
• And, even if they did want to serve your objectives, not theirs, they would be hard-pressed to come anywhere near the richness and diversity of truly independent, opt-in, community-driven middleware services that are tailored to diverse needs, contexts, and sustaining revenue models.

Doing moderation the old-fashioned way – enabled by middleware

Instead of being seen as a magical leap in technology, or an off-putting cognitive burden on users, middleware can be understood as a way to recreate in digital form the formal and informal social structures people have enjoyed for centuries – individually composed interaction with the wisdom of organically evolved social mediation ecosystems and intuitive informal reputation systems.

What at first seems complicated, from the perspective of current social media, is at core, little more complicated than the structure of traditional human discourse – building on key functions and elements of the social mediation and reputation ecosystems – all legitimized by choices of individual agency. Yes, that is complicated, but humans have learned over millennia to intuitively navigate this traditional web of communities and reputations. Yes, make it as simple as possible, but no simpler!

Creating an online twin of such a web of community ecosystems will not happen overnight, but many industries have already built out online infrastructures of similar complexity – in finance, manufacturing, logistics, travel, and e-commerce. Middleware is just a tool for enabling software systems to work together in ways similar to what humans (and groups of humans) do intuitively. The time to start rebuilding those ecosystems is now.

____________________

Related works:

●     Overview on Tech Policy Press -- Three Pillars of Human Discourse (and How Social Media Middleware Can Support All Three).

●     My November 2023 post introducing the pillars framing – A New, Broader, More Fundamental Case for Social Media Agent "Middleware" – introduced the Three Pillars framing, and embeds a deck that adds details and implication not yet fully addressed elsewhere.

●     Core ideas addressed more formally in my April 2024 CIGI policy brief, New Logics for Governing Human Discourse in the Online Era.

●     Very simply -- What Is a Social Mediation Ecosystem? (and Why We Need to Rebuild It). 

●     Other related works are listed on my blog.