(This is an early draft reposted from my wiki, where the latest version can be found.)

In my AtmosphereConf workshop in Vancouver last March I explored how research teams could create more collective intelligence by pooling each individual team members’ Atmosphere activity. The core idea is a win-win: institutions build structures which benefit both them and the researchers using them, with the knowledge generated also made more useful to wider audiences.

This post explores how those wider audiences can best benefit.

most of the knowledge created by research projects remains locked in the researchers’ skulls

This question is informed by the work I’ve done for Knowledge4Policy, the science for policy platform of the European Commission’s Joint Research Centre, so my focus is on bringing scientific knowledge to policymakers. The same ideas, however, also apply to communicating science to the wider public and transferring scientific results to industry, because all three audiences have the same problem: scientists publish their results as research paper PDFs, a form designed solely for other scientists, while most of the knowledge they create is tacit, and remains locked inside their skulls.

Project Toolkit

The approach is bottom-up, flexible and resilient, so let’s start at the scale of an individual University research team, Horizon Europe research project or similar. 

They decide to adopt, configure and customise a “Project Atmosphere Toolkit”:

This toolkit is designed to help the project and its researchers communicate their results, discover useful knowledge and use it, so it features:

• Conversations*
  • the project researchers discuss their work on Bluesky, with all posts curated into a “project team feed“— a simple Bluesky custom feed created by combining a list of the researchers and a hashtag: “here’s what the research team is saying about their work in this field”
  • “high-value community feed”  -  a second custom feed, driven by the same hashtag coupled to a larger list of the project’s researchers PLUS the scientists they trust: “here’s what this research team and scientists they respect from around the world in this field are talking about”.
• Content discovery and curation: some or all of these scientists use other Atmosphere apps — eg:
  • Sill to surface interesting content from their social graphs
  • Semble or Margin to curate knowledge into collections, and annotate or comment on them.

(*For more on this custom feed architecture, see How newsrooms, scientific institutions & governments can best use Bluesky.)

All these conversations, bookmarks, curated collections and more are stored in each researcher’s Personal Data Server (PDS). That content is public and its locations are known, as the team members and their trusted peers appear in the Lists used to build the custom feeds.

That makes it easy for the project website to aggregate and publish that content, transforming it from a static “here’s our team and its outputs” page into a window onto the project’s living, tacit knowledge: the conversations these researchers are having, the content they’re finding and organising across the web, the emerging consensus and open questions in their field.

And it’s just as easy for everyone else. Enter knowledge brokers and science communicators.

Project Cluster Toolkit

So what happens when multiple projects in the same field do this?

Once there are several projects in the same field doing this, something interesting becomes possible:

The figure shows a “Project Cluster toolkit” creating an aggregation layer above several individual, related projects:

• Aggregating the projects’ feeds and surfacing the best posts — “here are the top posts from all the researchers in these projects and the scientists they trust”
• Building a ranked knowledge base: every URL any of these researchers shared, bookmarked, curated or commented on, ranked by how many researchers across how many projects engaged with it, and filterable by researcher, project, or (AI-tagged) topic
• Providing an expert-finding service, because the cluster knows who is in each project and who they trust.

This is the collective tacit intelligence of a cluster of projects — not a list of PDFs maintained by a few harried project coordinators, but the ongoing intellectual activity of an entire, multi-team research community, automatically created from the tools they use in their day to day work.

This layer provides the source material for adding value to this work by science communicators, knowledge brokers, and industry. Feeding the outputs into an LLM with retrieval-augmented generation to synthesise findings and answer their questions is an obvious next step, with the usual caveats.

Building knowledge communities on the Atmosphere

The K4P audience research made it clear that policymakers will never search a database to download and read a dozen research papers.

Instead, they “want someone to talk to”: either a scientist or a knowledge broker, whose job it is to synthesise research and answer their questions.

So when a policymaker asks “what does our science say?”, the project cluster gives them and/or their knowledge broker somewhere to go. Not a database of PDFs, but a living, searchable knowledge layer icommunity: the conversations, curated content and emerging findings ... and the experts behind them.

Policymakers and knowledge brokers won’t be the only beneficiaries. Science journalists can dive into what these research communities are discussing, while industry can identify emerging research results faster.

Triggering a virtuous cycle

This doesn’t need to be implemented by one centralised authority, forcing everyone into a centralised platform.

Once research teams start using the Atmosphere in this way, anyone can cluster any of these projects together to aggregate and add value to their work. The more that happens, the more valuable using the Atmosphere project toolkit becomes to the research teams themselves (who doesn’t like having greater impact for free?), triggering a virtuous cycle.

But that cycle doesn’t start until individual research teams and projects demonstrate the toolkit’s benefits to themselves. Once they’re using these tools for their own purposes, the resulting knowledge communities will become useful to everyone else, and the benefits will begin compounding for everyone as adoption grows.

The result will be a more efficient and effective science knowledge infrastructure which is less vulnerable to capture by scientific publishers, and more resilient as a result.