A New, Broader, More Fundamental Case for Social Media Agent "Middleware" (Revised)

(Discussion draft post and deck, restructured, expanded 1/7/24, as noted below.)

Despite the efforts of business, government, and academia, there seems to be no adequate solution to the dilemma of managing any-to-any online media at global-scale. Too much central control by platforms or governments is a "loaded weapon on the table" ripe for authoritarian abuse, but media anarchy pollutes the public (and private) sphere -- and there are no bright lines. This is creating a deepening crisis not only in the world's political health, but in all aspects of public health: social, mental, and physical. 

How can we maintain freedom of thought while limiting harm from antisocial speech? Democracy is in crisis over who controls what is expressed online -- and what is impressed upon each of us in online feeds and recommendations. What are the legitimate roles of online platforms, government, communities, and individuals in such controls, and how does that depend on community and context? There are numerous efforts and proposals, many with significant support, but each has serious limitations. 

It recently struck me that three key solution elements that I have been advocating for many years have an importantly synergistic effect. I have become all too familiar with the objections for each element that have limited uptake -- and now see that the way to counter those concerns is to clarify and build on how these pillars work in combination – to reinforce one another and serve as a foundation for the full suite of remedies.

I offer this as a significant broadening of common thinking about "middleware" services (intermediaries between users and platforms) -- in a way that makes it far more powerful and important to civil discourse, and counters various concerns that have hindered its acceptance as a way to preserve democracy in the online era.

Middleware can support three essential pillars of discourse that synergize with each other to restore the human context that platforms have collapsed:
       1. Individual agency (the current focus)
       2. A social mediation ecosystem (now seen apart, fragmentary, even conflicting)
       3. Reputation and trust (now considered only in basic form).

Here are some brief notes -- followed by an embedded deck that serves as a working outline with more depth on my suggestions. 

Three pillars

The three pillars that synergize to restore human context as a foundation for managing online discourse are:

1. Individual choice and agency, over how we each use online media – this creates speaker/listener context. This gained significant recognition after Francis Fukuyama and his group at Stanford proposed it be enabled via “middleware” that sits between users and the platforms, as a democratic way to limit how platform power threatens democracy. The idea is to return power to users to steer our online “bicycles for our minds” for ourselves.
   
   
2. A social mediation ecosystem, which cooperatively applies collective intelligence, wisdom, judgment, and values, to serve users, as networked into social groups – this mediates context collectively. We have failed to directly integrate the traditional roles of more or less organized social groups into social media. The idea is for social media to leverage our social associations to promote “bridging” of the divides that social media now seem to highlight and reinforce – by rebuilding our processes for creating "social trust." Many have proposed aspects of this, but I take this much farther than I have seen suggested anywhere else (as explained further in the update below, and more deeply in the deck).*
   
   
3. Reputation and trust, both in individuals and in what they say – to evaluate speaker/mediator context and trustworthiness both individually and collectively. This is less widely advocated, and most proposals for this are relatively basic, but some have seen that much more powerful reputation and trust systems are possible -- much like how Google has applied reputation and trust to web search. The idea is to apply the kind of rich combination of individual and social judgements of reputation that guided traditional (pre-online) discourse.*

I now see user agent "middleware" as underlying all of the three pillars, enabling them to work together to restore the context that is essential to effective discourse. Most consideration of middleware seems to focus almost entirely on just the first of these pillars (important as it is), thus understating its true potential and raising concerns that the other pillars can reduce.

My primary focus here is “social” media – in its broadest sense. That also applies to hybrids of human and artificial intelligence (AI), as touched on briefly.

Context collapse

A key reason why online discourse is so problematic is that global any-to-any networks generally collapse the subjective mutual understanding of context -- who is speaking to what intended audience in what way. This has been understood as “context collapse.” These three pillars work together, through middleware, to restore this lost matrix of context, thus making the particular and subjective nuance of online discourse more understandable to both humans and algorithms. I suggest that can counter the feared pitfalls of each alone.

The broader need for middleware

As a long-time advocate for user agent middleware, I have seen it gain support with a primary focus on restoring the pillar of user choice and agency, but generally in ways that are narrowly centered on that, and open to important concerns. I now see the need to emphasize the synergy of each pillar with the other two more clearly – and to make the case that user agent middleware can and must support all three pillars as they work in concert - individual agency, social mediation, and reputation. The hope is that will provide a far more powerful benefit, and counter the common objections arising from narrower framings. 

That might lead to much broader uptake of this important strategy for reestablishing human context that I believe can provide a strong foundation for cutting through current dilemmas, using these and other supplementary strategies to enable online discourse of all kinds to have a far more positive influence on society, and sustain democracy -- for both individual and collective welfare. 

The fundamental synergy is the dialectic of a flexibly optimized blend of human freedom gently balanced by a degree of social nudging toward responsibility. Underlying  that synergy is the collective wisdom that humans embed in reputation. Middleware is the technology that supports this traditional human context in the online world of computer-mediated discourse. Think of it as contextware.

Working notes on this thesis -- a slide deck

As I have begun socializing this strategy, in preparation for a more formal presentation --and to draw in others who might join in developing these ideas -- I am sharing this working version of a deck. It explains these elements in more detail, including how they work together, and how all three are facilitated by middleware as the underlying connector -- and so can together counter objections commonly raised in response to each when considered individually.

(The deck can be viewed on Google Slides without a Google account here.)

Feedback on this is invited (intertwingled [at] teleshuttle [dot] com).

Major update 1/7/24:

Drawing on comments from many experts (acknowledged below), this post has been revised and the slide deck has been expanded and reorganized. The deck adds new sidebars, and section breaks to facilitate easy scanning:

• Summary
• Overview
• Synergies
• Thought as a Social Process
• Middleware as Foundation
• Sidebar: The App Store Analogy (new) 
  Analogy suggesting how transformative a middleware ecosystem can be. 
• Conclusion
• Sidebar: Fear of Middleware (new)
  Addressing the fears, especially fragmentation and partisan sorting. 

Added Sidebar:
Migrating our traditional social mediation ecosystem into social media 

Part of the fear that middleware might increase social fragmentation derives from the narrow way that social media middleware is generally understood -- as apart from the broader social mediation ecosystem and its historically central role. Consider this broadened perspective:

What we've got here is failure to re-integrate. The idea of a social mediation ecosystem integrating with social media feeds is a re-visioning of how things used to work. Context collapse is not a problem internal to online social media, but a broader failure to migrate our existing social mediation ecosystem -- our processes for "social trust" -- into the digital domain.  We seem to have forgotten how society has been building on such ecosystems for millennia. 

• The groups that comprise the social mediation ecosystem have historically served as a public square, or public sphere, ranging from informal gathering places like coffee shops and taverns, social & civic associations, the press, academia, churches, unions, workplaces, and other communities of interest. 
• Such associations develop norms and contexts for discourse. Our participation in them shapes what we see and hear of the world. That nudges us to speak "freely," but with sensitivity to those norms and values, so others will choose to listen to us. 
• Online media technology can enable restoration of that mediating role through enterprise-class affordances that support community operation (including integration with CRM systems) and let users interact both within and across communities. 
• Middleware can facilitate and enrich user-community interactions, and enable us to steer our feeds to blend content favored by whatever mix of communities we choose to include at a given time -- depending on our tastes, objectives, tasks, and moods. 
• For example, current curators of news could become attention agent services. Users might select a set of them to compose, with different relative weights in ranking. Eg: NY Times, CNN, MSNBC, Fox, The Atlantic, People (and other categories, such as civic, political, church, and special interest associations). Content ranking inputs could be from the community’s expert curators/editors and/or crowdsourced from the population that follows those curators.
• Importantly, as it was historically, this ecosystem must be open and diverse, and users must be able to draw on combinations of many mediation sources to maintain an open and balanced understanding of the world. 
• Many fear that middleware might increase fragmentation and partisan sorting. That will be a concern while there are just one or a few mediators, but being able to selectively combine exposure to many loosely connected communities is how open societies have always limited that ever-present risk. 
• (While there has been little attention to reintegrating this ecosystem into online social media, it has figured in many of my prior writings, as listed here, and dates back to my 2001-3 design for a richly distributed social media system that anticipated current and forward-looking ideas like multi-level, multi-homing federation, user-chosen middleware, and reputation-based attention agents -- as highlighted here. My inspiration for this perspective was from my work with the emerging open market in financial market data analytics around 1990, and from my work on the evolution of "intergroupware" around 1997.)

______________

Acknowledgements (apologies for any omissions):

Thanks to those who have provided stimulating feedback on these ideas, including Chris Riley, Luke Thorburn, Aviv Ovadya, Daphne Keller, Francis Fukuyama, Zoelle Egner, Ethan Zuckerman, Chand Rajendra-Nicolucci, Zach Graves, Luke Hogg, Harold Feld, Pri Bengani, Gabe Nicholas, Renée DiResta, Justin Hendrix, Richard Whitt, Ellen Goodman.

(Updated 2/3/24)

*[Added 7/26/24] To better clarify the distinction between pillars 2 and 3, note that social mediation is a process that occurs over time, while reputation/trust is a quality (a reputation for trustworthiness at a given point in time that results from the operation of the social mediation process). 

From Fediverse to the Pluriverse of the Future - Navigating A Linked Web of Communities in Many Dimensions

[Discussion Draft]

Where should the still-formative vision of the Fediverse be headed?

The "fediverse" of Mastodon and similar systems has gained attention as a shift in social media from centrally-controlled platforms. 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. 

I hope a longer-term vision for the fediverse will be a focus of discussion at the FediForum "unconference" (September 20-21, 2023). I hope to join in (or lead) a session on that -- shaping ideas for the long-range direction of how social media can better augment human discourse of all kinds, across all platforms. 

While there is obvious need to develop near-term features to make each competing tech platform and universe of platforms more compelling, there is also a need to articulate long-term objectives that many platforms can build toward. We are re-engineering human discourse for the online era -- that will be a long process -- but without dialog seeking a long-term vision, it will longer and more problematic.

I have been writing about this future of "bicycles for our minds" for many years (some in collaboration with Chris Riley). My prior blog post on "hopes for the Bluesky project" offers my most current overview of how these ideas might apply to existing and emerging social media architectures. While that post was written with a focus on Bluesky -- as currently pushing farthest in providing for the multidimensionality that will underlie a pluriverse -- those directional ideas apply equally well to the fediverse of Mastodon and other ActivityPub-connected systems. 

I look forward to discussing these (and alternative, or opposed) ideas at FediForum, to develop collective insight into how the fediverse might evolve to meet future needs.

The TL;DR of the pluriverse, as I envision it:

• The move to federation, the fediverse, and on to 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 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 created 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. 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, possibly weighted using sliders.
  
  
• Bluesky seems farthest along in pointing to this multidimensionality, building out (but not yet far in implementing) tools for separating the "speech layer" from the "reach layer" as described in blog posts on Composable Moderation, Moderation in a Public Commons, and Algorithmic Choice. My recent post suggests directions for taking that farther.
  
  
• Mastodon seems to also be moving in that direction, with discussion of a cross-instance groups structure, and shared moderation services that address the challenges of administering small communities.
  
  
• The vision I suggest will take time to build, develop, and be fleshed out by users, but having these ideas in mind as we architect and build in the near-term, will be important to being able to get where we will want to go in the future.

Summarizing the vision in my prior post: 

The sections of that post are summarized here (but I hope readers will look to the fuller explanations there):

• Hypercommunities
  Each person can "be" in 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 recommendation boils 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.
  
  
• User-selectable, multilevel feed composition composed from multiple algorithms
  A truly composable, steerable feed would provide a higher level interface that lets each of us merge a user-selected mix of 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. 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.
  
  
• Rebuilding our social mediation ecosystem
  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.
  
  
• 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 of labeling and ranking -- and federation
  
  
  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
  I see 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.
  
  
• "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 (and with whatever levels of moderation they want). This is the infancy of a flexible new social ecosystem, and whatever initial vibe chaos might arise will give way to a new order of vibe control. A fully functional social media pluriverse will be a virtual "shmoo" (a classic cartoon creature that tasted like whatever you wanted) -- 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 fediverse grows in scale and diversity and in the richness of interconnections desired by users with many interests and moods for diverse vibes.

Much of this flexible multidimensionality will emerge slowly, as technical, human, and social infrastructures co-evolve toward it -- a whole-of-society process that will take decades. 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 that is robust and generative.

My related works are listed on the Selected Items tab of this blog, with the following as most relevant:

Recent visions of the fediverse/pluriverse

• My Decades of Blueskying, and Hopes for the Bluesky Project (6/25/23) - my thinking on where I hope they will take us. Even if Bluesky, the company, fails to achieve critical mass, it has potential to lay the foundations for next generation protocols and services that will. 

• Context Lost; Context Regained – Comments on iDPI's “A Manifesto for a Smaller, Denser Internet” (4/5/23) - My comments and my own expansions on a manifesto from Ethan Zuckerman's iDPI team. A "pluriverse" of many linked diverse platforms and a "loyal" agent that "aggregates, cross-posts, and curates."  

• Into the Plativerse… Through Fiddleware? (Tech Policy Press, 12/20/22) - Suggesting a future that enables nuanced control -- and may enable the emergence of federated middleware (fiddleware?) to best serve users. See especially the last two sections here. (*My more recent 4/5/23 post favors the term "pluriverse" as more evocative.) 

In more depth on the vision:

• The Internet Beyond Social Media Thought-Robber Barons (Tech Policy Press, 4/22/21) - See especially Part 1 and beyond.
  

• The Augmented Wisdom of Crowds:  Rate the Raters and Weight the Ratings (7/22/18) - An earlier detailing of my vision.
  

Broad statements of direction and motivation:

• 
  From Freedom of Speech and Reach to Freedom of Expression and Impression (Tech Policy Press, 2/14/23) - Distilling and updating essential reframings from the Delegation series (just below). 

• The Delegation series in Tech Policy Press with co-author Chris Riley:

1. Delegation, or, The Twenty Nine Words that the Internet Forgot (2/28/22)
   
2. Understanding Social Media: An Increasingly Reflexive Extension of Humanity (4/27/22)
   
3. Community and Content Moderation in the Digital Public Hypersquare (6/17/22)
   
4. Contending for Democracy on Social Media and Beyond (9/22/22)

Where I am coming from: 

• The Roots of My Ideas on Tech Policy

[First posted 8/10/23, revised 8/17/23. Thanks to Jaz-Michael King for very helpful comments.]

My Decades of Blueskying, and Hopes for the Bluesky Project [Updated 6/28/23]

 ***See updates to Key Ideas section that follow below (6/28/23)***

This vision also applies to the future of the Fediverse, Mastodon, and all of social media. [Update 8/9/23]

Having been thinking for decades about the potential of social media to offer steerable "bicycles for our minds" individually and collectively -- and becoming increasingly concerned by the directions of the past decade -- I now see some very encouraging patches of blue. 

I have been following and commenting generally on the Bluesky project that Jack Dorsey spun out from Twitter, and this April wrote about a similarly aligned project, the Initiative for Public Infrastructure led by Ethan Zuckerman. That post explained how the iDPI effort aligned with my ideas, and where I hoped it might go.

After using Bluesky for nearly two months, and reading some of the growing body of their thinking (in their blog posts and related details on Github) it seems timely for me to respond to their requests for feedback by outlining my thinking on where I hope they will take us. Even if Bluesky, the company, fails to achieve critical mass in its mission “to develop and drive large-scale adoption of technologies for open and decentralized public conversation,” it has potential to lay the foundations for next generation protocols and services that will. What I have seen so far -- building out Composable Moderation, Moderation in a Public Commons, and Algorithmic Choice -- seems well-aligned with my vision.

KEY IDEAS FOR BLUESKY
[...and the Pluriverse of social media in general]

This is a first, brief, and informal discussion draft, summarizing and pointing to ideas I hope the Bluesky team will be pursuing [with updates below]. I don't know how much of this is already in their long-term architectural plan. Of course it is not reasonable to expect the team to be far along in implementing much of what I suggest here -- that will be a massive and extended whole-of-society effort. My objective is simply to paint the vision, in hopes that they (and others) will share it, to ensure that their architecture is designed to extend in these directions as it develops. The hope is that like the web, this architecture will be generative and extensible enough to evolve over decades to provide a rich backbone for augmenting nearly all human discourse -- and the processes of its social mediation.

Following this brief summary are pointers to works of mine that expand on this vision in some detail.

Hypercommunities

The What is Bluesky? blog post says “In the federated network, people can move between cities depending on what kind of community they’d like to be in.” This analogy takes a step in the right direction, but strikes me as missing the essential multi-dimensionality of humanity's social web. 

The beauty of online discourse is that I can "be in" many virtual communities at once – I don’t need to “move between” them. Because these communities are virtual, I can participate in many at once, and interact with community members who also participate in many communities at once, as a giant web of overlapping Venn diagrams. I can have multiple "home" communities. At any given time, I should be able to have a view of my own composed virtual community, a view that includes whatever mix of communities I wish to participate in or just observe, ranked into my attention as I choose at the time. Feeds (and searches) should be composable and steerable to provide that view. 

This hyperlinking of public (and/or private) spaces is explained in Community and Content Moderation in the Digital Public Hypersquare (co-authored with Chris Riley). Much as web sites form a hyperlinked web that can be seamlessly connected with varying degrees of openness (manually, with links, or using web services), we can build webs of hyper-communities that are connected by our webs of connections to them and to their members. I refer to that as semipermeability, like a membrane that selectively passes some things and not others. As Ted Nelson said, “everything is deeply intertwingled.”

Ranking as the core task

Perhaps it is implicit, not yet documented, or I have missed it in the Bluesky materials, but it seems to me nearly all mediation boils down to ranking. Except in the most egregious cases, "moderation as removal" is anathema. "Filtering" is often narrowly understood as weeding out, not as ranking up or down. Egregious content might be downranked with prejudice, and quarantined, but the value of most content is in the eye of the beholder, and in the eye of those communities that beholders participate in based on shared norms and values. 

Done well, 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, weighted as appropriate and desired. Rankings can be based on many dimensions of attributes, with items coming to our attention based on which attention agents uprank or downrank them my how much, and what weight is given to each of those agents.

Composability should also be dynamically steerable. Think of “bicycles for our minds” and how we can steer them at will. And remember that these bicycles should steer us through the multidimensional and semipermeably overlapping web of hypercommunities.

Multilevel feed composition composed from multiple algorithms

I hope the Bluesky architects have this in mind, but have not seen it clearly stated. Currently My Feeds gives a list of pre-defined feed algorithms that we can view one at a time. A truly composable, steerable feed would have a higher level interface that lets us merge a mix of feeds, with defined relative weights. A steerable feed would allow those mixes and weights to be easily changed at will to suit our tasks and moods. Obviously, this full capability and the appropriate UIs for it will take time to develop, but I hope the architecture is being designed to provide extensibility and protocol support for this. Some UI options might be very simple, and some might be suited to those who desire fine granularity of control.

Multi-dimensional reputation based on explicit and/or implicit signals

I view reputation as essential to making ranking work well. Reputation cannot be adequately captured by simple lists. I have written frequently about “rate the raters and weight the ratings” as an extension of Google’s PageRank algorithm to develop what Scott Aaronson has called "eigentrust" (=“eigenreputation”). I have suggested this use implicit ratings -- like, shares, comments (and perhaps more value-indicative signals) – as well as explicit ratings (which might include labels). Feed algorithms can use these methods in an infinite variety of ways. As a simple example, a feed might be composed in part based on implicit ratings from users of some mix selected from followers of Fox, MSNBC, the NY Times, or People magazine – or alumni of Harvard, Ohio State, or Texas A&M, or members of some church or union. The beauty of this kind of computed PageRank-style "eigenreputation" is that it is far more nuanced, current, and broadly sourced than binary lists of who is vouched for or not by some list curator.

This reputation system should ultimately be multidimensional. Reputation ratings may be segmented with respect to specific subject domains and value orientations, and can be selectively sourced from specific communities of interest and value. That way content and people can be ranked in different ways for different purposes. While doing this at scale may seem very complex, my understanding is that Google does similar context-specific segmentation for PageRank. Resources to do that are not yet in hand, but as such services reach scale, funding models will follow.

Rebuilding our social mediation ecosystem

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, where many of the same communities and institutions can add value. These signals of human judgment can be crowdsourced from their membership, but they can also derive from editorial curation sanctioned by these communities/institutions. Many providers of Bluesky algorithms might be tightly integrated into the technical infrastructure of these communities/institutions.

Context Lost; Context Regained – Comments on iDPI's “A Manifesto for a Smaller, Denser Internet”

Here are my preliminary comments on a “manifesto” we should all read and get behind! -- arguably the most concise, yet comprehensive, sensible, and understandable vision statement for where our social information infrastructure needs to go. I suggest it as required reading for anyone who cares about the sad state of social media now, and where civilization needs our media tools go.

The Three-Legged Stool: A Manifesto for a Smaller, Denser Internet by Chand Rajendra-Nicolucci, Michael Sugarman, and Ethan Zuckerman (3/29/23), is a white paper on the seminal ideas Zuckerman and his team at the Initiative for Digital Public Infrastructure have been developing for some time, and a roadmap for their future work. It steps back from the current narrow and failing logics of social media to re-envision what new logics are needed.

I expected it to be a notable contribution, and was pleased that it exceeded my expectations for both substance and presentation. Here are initial impressions of key points that resonate with my own work, plus a brief list of some objectives that I would expand on as important to a longer-term vision.

A core failing of our current online media infrastructure is “context collapse” – the loss of the rich social context for information and discourse that human society has learned to sense and rely on for millennia. Here are ideas for restoring and augmenting that richness of context.

The iDPI Manifesto

The authors envision a public sphere supported by these three legs (quoting):

1. Consists of many different platforms with a wide variety of scales and purposes;
2. Users can navigate with a loyal client that aggregates, cross-posts, and curates;
3. Is all supported by cross-cutting services rooted in interoperable data.

The first leg is fundamental, and stands out from much current discussion as a “Both/And” solution that recognizes that neither centralized platforms nor the decentralization of the “fediverse” fully suit needs for global interconnections that support human diversity. The answer is an open architecture that can integrate Very Large Online Platforms (VLOPs) like Facebook, Instagram, and Twitter with Very Small Online Platforms (VSOPs) like many free-standing online communities and the loosely interconnected fediverse of Mastodon instances. They call this a “pluriverse” a “world where many worlds may fit” with diverse “goals, norms, and governance.”

The manifesto explains why we need a fertile, flexible ecosystem of community-scale VSOPs which serve different purposes than the existing VLOPs, and how they can provide affordances and foster norms that nurture “civic social media” where “moderators are active guides and participants”

I have written over the years about a similar vision, both on my own, and in a recent four part Delegation series with Chris Riley. (Late last year, before Musk shut down the nascent linkage of Twitter’s platform to the fediverse via rudimentary cross-posters, I suggested the term “plativerse,” but agree that “pluriverse” is more descriptive.)

The second leg is also essential to this vision, and lets users delegate authority for accessing their data and managing their feeds to a “loyal client.” This is a “new architecture for choosing, customizing, and testing ‘lenses’ for your feed” based on “an open standard for developing and integrating third-party algorithms.” This has also been a major theme of my recent work, and something I wrote about as early as 2003, then brought up to date in 2021, before expanding on that in the Delegation series. I have referred to these as “filtering services’ and “feedware” and more recently lean toward “attention agents*” as being functionally descriptive and technology-agnostic. Others have made related proposals, and Francis Fukuyama’s group at Stanford brought their version, called “middleware” to the attention of the tech policy community – even though it is somewhat narrower in scope, lacking the multi-platform aggregation/cross-posting features.

The third leg provides further support, most importantly for what the authors call “The Friendly Neighborhood Algorithm Store” that would realize the open market in feed algorithms that many advocates of delegated attention agents, loyal clients, or user-controlled middleware envision. It highlights the need for algorithms to be tunable, auditable, combinable, and understandable. Few have recognized the importance of combinability – I have referred to this as composing and orchestrating

The manifesto also explains why achieving this vision will be difficult, but is essential, and how strategies such as contextual privacy, adversarial interoperability --and regulation -- can be important.

Looking down the road

As complement to this iDPI manifesto, I note some further directions that can enrich this infrastructure. (I do not know to what extent Zuckerman’s team might already have these in mind.) Having them in mind now can help ensure that the foundations are built for needed extensibility. These focus primarily on two aspects of context:

• richly flexible interconnection of both virtual and real-world communities, and
• attention agent algorithms that efficiently exploit social mediation at individual and community levels to augment context in our feeds.

(Note that I use the term “social media” in the broad sense of media that are open to user/social participation -- which will gradually extend to include nearly all media.)

[Ted Nelson]

Intertwingled communities: Communities in real life generally benefit from being interconnected through fuzzy and semi-permeable boundaries, often overlapping like Venn diagrams. This suggests that our media tools should support an interlinked web of communities (hyperspheres) with semi-permeable boundaries -- much as the web already does so well. (These ideas date from my 2003 design,** and were expanded on in the third installment of the Delegation series in 2022.)

The manifesto seems to refer to communities in terms of platforms (VLOPs and VSOPs), but those semantics seem likely to blur. I see this as primarily a functional issue of supporting a wide range of human communities, and only secondarily an matter of platforms. The structure of communities need not align with the structure of platforms. I suggest thinking of VLOCs and VSOCs (Large and Small-Communities) -- with a full spectrum of mid-sized or compound communities in between. This enhanced interconnection functionality might arise from both directions:

• Independently served VSOCs (as iDPI suggests and is developing) might generally have more independence, and might increasingly outsource complex support to an ecosystem of distributed services – including marketplaces for clients, mediation/moderation, privacy/security, etc. This is the direction that is very feasible now.
• Conversely, VLOPs might offer increasing levels of agency and support services to communities (VSOCs) that they host -- to approach the functionality tailoring of VSOPs. That is now offered in only basic form but could be enriched at scale quickly.
• Either way, loyal clients/attention agents could add support for semi-permeable community boundaries, in combination with underlying platform support affordances. This could add support for the many ways traditional communities are not islands, but interconnect via their common memberships and various levels of non-member participation.

Integration with real-world community infrastructures: Perhaps the greatest tragedy of our wrong path on social media is the accelerated disintermediation and decay of the vibrant ecosystem of real-world communities that have been the lifeblood of human society. The manifesto refers to Robert Putnam (Bowling Alone) and suggests that “libraries, newspapers, public broadcasters, cultural institutions, local governments and others have a role to play in re-envisioning the digital public sphere.”

Regardless of whether they are funded publicly or privately (such as private academies, churches, clubs, unions, etc.) all these real-world communities might regain their important role in the social mediation and context of discourse. Social media platform support for them should be better enabled and integrated with their existing analog and digital infrastructures. In addition to building anew from green fields, there is an urgent untapped opportunity to rejuvenate the community institutions we already have.

Context-enhanced algorithms:  Algorithms can be enhanced to select and discover content items based on the reputation, with respect to relevant interests and values, of those who post and interact with those items.

• Importantly, managing semi-permeable boundaries can make algorithms far richer, drawing on a wider universe of human judgement, focused through the lens of a network of communities chosen for interest and value.
• Applying a recursive reputation model much like Google’s PageRank (The Augmented Wisdom of Crowds: Rate the Raters and Weight the Ratings) would make such algorithms smarter, by drawing on a powerful way to refine the best of human judgement, and would be harder to spoof or game.
• Such algorithms could partition the user feedback they draw on to dynamically assemble and reassemble relevant communities of interest and value on demand.
• Challenges to doing this include a need for a stable identity that builds reputation history, and the need for contextual privacy strategies to provide controlled access to the interaction history and ID metadata needed to track and evaluate reputation.
• Reputation also provides a way to finesse the dilemma of anonymity, aliases, and verified identities (that Musk’s Twitter is now shining a harsh light on). Reputations can be developed even for a persistent alias, without the need to divulge a real identity that might need safety.

Interplay of community-context-driven algorithms:  Algorithms can be more powerful when controlled by delegated client agents that are aware of community boundaries and relationships. This builds on the idea of community level feeds that are composed into composite feeds.

• Community-level feeds can use community reputation for items and for people, to decide community-based rank in the feed for each community member.
• Openness to outside communities can be based on not only on explicit referrals from users and external recommenders, but also on implicit rankings for other similar members – both within a given community, and in related communities, based on cross-cutting connections.
• Thus (as iDPI may have in mind), feeds can draw on signals from within a community as well as from other communities that users choose to orchestrate together. Consider a Long Covid feed that draws on implicit metadata from communities of doctors, researchers, patients, and medical journalists. This could uprank items from global sources that are liked, shared, or commented on by members of those communities, weighted by community-specific reputation.

Business models: I would also look to innovation that would apply adaptively win-win business models that can seamlessly blend subscription, donations, ads, and public subsidies to make these new services sustainable without being extractive of user value and agency. Core principles are outlined in HBR and journals, as cited in my FairPayZone blog, and promise to enable revenue models that are affordable and offer fair value to a full range of users and ecosystem participants. The pluriverse may unleash a new era in business model innovation and experimentation that might bring these strategies into the mainstream.

I applaud the work iDPI is doing, and their manifesto provides a very helpful overview of why it is important. While there is much that went wrong with the VLOPs, we can remold the world of online media to the heart’s desire without shattering the platforms to bits.

(*Drawing on the "attention-allocation" terminology suggested by Ovadya and Thorburn, and emphasizing the issue of agency.) 

(**Update 7/14/24 - even earlier, to the days of extranets in 1997.)

Back From Absurdity: The Essential Reframing Needed to Manage Online Speech - Now on Tech Policy Press

My latest article in Tech Policy Press distills the core ideas needed to reform how we manage online speech: From Freedom of Speech and Reach to Freedom of Expression and Impression.

Key ideas: Problems in managing social media news feeds, including disinformation, extremism, and hate speech, are reducing to the absurd because of a misguided focus on censorious removal. A new framing is needed to reverse these problems, from the “Twitter Files” and Musk’s confusion about “freedom of speech, but not freedom of reach,” to the absurd pairing of cases on online speech headed to the Supreme Court. 

The solution is to shift focus to manage the other end of the proverbial “megaphone” – not the speaker’s end, but the listener’s. Dominant social media platforms are co-opting the “freedom of impression” that we listeners did not realize we had. Proposed remedies based on “delegation” and “middleware” promise to address this, but this new framing in terms of “freedom of impression” is necessary to clarify for all concerned why that is needed from human rights, legal, governance, economic, cultural, and technological perspectives -- and how it can work.

This new framing offers a way to apply both/and thinking to cut through many current dilemmas, and to set a direction for the future of freedom and democracy. It also illuminates a new path toward competition in this market.

Operationally, it suggests how to manage networked speech by refactoring the balance of three control points - to balance full freedom of expression with full freedom of impression:

• Censorship as posts and responses enter the network, entirely banning users or removing individual posts before they reach anyone at all -- a threat to freedom of expression.
• Selection of what is fed or recommended out to each user, individually -- a threat or exercise of freedom of impression, depending on who controls it.
• Friction and other measures to enhance the deliberative quality of human social mediation activity -- with little threat to freedom of thought.

There is no quick fix to the problems of social media, but we can quickly change course to begin to undo the disaster of the past decade or two -- and to avoid ill-conceived remedies that will fail or make things worse. As shown here, doing that means: 

• relying less on censorship (bans and removals that have questionable legitimacy, even when guided by the most well-intended proportionality), and
• giving users agency over the selection of what they see (to legitimately balance each speaker's freedom of expression with each listener's freedom of impression), and
• re-creating a truly open and generative social mediation ecosystem that is like what we have been evolving over the past centuries of analog society, but now augmented by digital media tech (instead of de-augmented and disintermediated by it).

Read the full article on Tech Policy Press...

...And dig deeper into the rationale and concepts behind this distillation and update of the Delegation series (co-authored with Chris Riley)

---

Podcast: For those who prefer listening to reading, most of these ideas are covered, along with some additional commentary (but a few months less currency), in the Lincoln Network's podcast, The Dynamist, hosted by Evan Schwarztrauber.

Reisman on The Dynamist Podcast - Reforming Social Media with Delegation and Freedom of Impression

Lincoln Network's The Dynamist podcast (Episode 5, 2/7/23) features an interview of me by Evan Schwarztrauber that covers many of the key ideas in my Delegation series in Tech Policy Press with Chris Riley, plus more that will be in an article to be published soon. [Update 2/14: Now online at Tech Policy Press: From Freedom of Speech and Reach to Freedom of Expression and Impression.]

It steps back to consider the inherent absurdity of current approaches to content moderation and the dilemmas of freedom of expression vs. censorship -- as exemplified in the cases now headed to the Supreme Court: some that would require platforms to carry "lawful but awful" speech, and others that would make them liable for carrying it. 

I suggest the way to cut through those dilemmas is by giving users agency to restore their freedom of impression, which the platforms have co-opted -- and to delegate that agency to services that can shape their feeds and recommendations in accord with criteria and values that they choose. (This was recorded on 10/13/22.)

Thought as a Cyclic Social Process: Thought => Expression => Social Mediation => Impression => ...

[Update 2/14/23: The article this previews is now published on Tech Policy Press.]

This is a preview of ideas from an article in the works, introducing some new diagrams seeking to distill and simplify key ideas addressed in the Delegation Series (with Chris Riley).

Freedom of thought, expression, and impression are not just isolated, individual matters, but an ongoing, cyclic, social process. Thought leads to expression, which then flows through a network of others – a social mediation ecosystem. That feeds impression, in cycles that reflexively lead to further thought.

Cutting through the dilemmas of managing networked speech will depend on balancing full freedom of expression with full freedom of impression, by augmenting the social mediation ecosystem with the right balance of three control points:

• Censorship as posts and responses enter the network, entirely banning users or removing individual posts before they reach anyone at all -- a threat to freedom of expression.
• Selection of what is fed or recommended out to each user, individually -- a threat or exercise of freedom of impression, depending on who controls it.
• Friction and other measures to enhance the deliberative quality of human social mediation activity -- with little threat to freedom of thought.

There is no quick fix to the problems of social media, but we can quickly change course to begin to undo the disaster of the past decade or two -- and to avoid ill-conceived remedies that will fail or make things worse. As shown here, doing that means: 

• relying less on censorship (bans and removals that have questionable legitimacy), and
• giving users agency over the selection of what they see (to legitimately balance each speaker's freedom of expression with each listener's freedom of impression), and
• re-creating a truly open and generative social mediation ecosystem that is like what we have been evolving over the past few centuries of analog society, but now augmented by digital media tech (instead of de-augmented and disintermediated by it).