What Unanimity Hides

Opening up a dictionary does not give you understanding of the language you are looking at. You can see the shapes, the outlines, perhaps read the translated word — but you do not understand how the language is formed. For many complex systems there is an inherent tension between form and function: knowing the parts does not tell you how they fit together, and knowing the rules does not tell you what actually happens.

An economy has property laws and trade agreements, but its productive structure — which industries cluster together, which capabilities travel with which — emerges from millions of independent choices about what to produce and what to trade. Countries that are good at making wine also tend to produce olive oil. To be able to produce cars, you need an infrastructure for making tires out of rubber. Nobody designed these relationships. They crystallized from repeated action, and they have to be recovered from the pattern.

International institutions have the same problem. They have founding treaties, formal rules, stated objectives — but the rules rarely tell you how members actually organize their attention in practice. What they repeatedly prioritize, which concerns travel together, who occupies which position in the issue space: that part of the structure is not written in any document. It has to be inferred.

Consensus institutions make this apparent. When every decision is unanimous, formal outcomes carry almost no information about internal differentiation. Everyone agreed — but agreed from where? With what emphasis? Having raised what concerns? The unanimity is real, but it flattens a richer landscape of distinct positions into a single signal: yes.

In a recent paper, my co-authors and I developed a way to recover that hidden structure. We drew on an idea from economic complexity — that you can infer the relatedness of activities by looking at who repeatedly co-specializes in them — and applied it to international governance. The test case is the Antarctic Treaty System, one of the longest-running consensus institutions on Earth, where sixty-plus years of submissions reveal a structured “space of concerns” that no vote or treaty text would show you.

This post walks through the core ideas of that paper: how we construct the space, what it reveals, and how actors move through it. The paper itself contains the full technical detail and robustness checks; here I want to focus on building intuition for the construction step by step, with interactive visualizations along the way.

The Data: What Countries Put on the Table

The Antarctic Treaty System holds annual consultative meetings where member states submit working papers, information papers, and other documents. Each submission is tagged by the ATS secretariat to one or more of roughly 45 recurring topic categories — things like “Environmental Protection”, “Tourism and Non-Governmental Activities”, “Marine Acoustics”, “Inspections”, or “Sub-glacial Lakes”.

This tagging gives us something valuable: a long, structured record of what each actor repeatedly chooses to raise. Not how they vote — there are no meaningful votes in a consensus system — but what they put on the table, year after year. That revealed pattern of attention is the raw material.

Measuring Specialization: Who Cares About What?

The first step is to figure out which actors are genuinely specialized in which topics, as opposed to simply being active across the board. A country like Australia submits a lot of papers on a lot of topics. That volume alone does not tell us much. What we want to know is: relative to the size of their overall engagement, do they disproportionately focus on certain concerns?

This is where we borrow from economic complexity. In trade economics, revealed comparative advantage asks: does a country export more of a product than you would expect given the country’s total exports and the product’s global share? The same logic applies here. We compute a revealed policy advantage for each actor-topic pair: does this actor submit more on this topic than you would expect given how active they are overall and how popular the topic is across the system?

whereis the number of submissions actormakes on topic. When this ratio exceeds 1, the actor is disproportionately engaged with that topic — they specialize in it relative to the baseline.

This is a simple but important normalization. It separates signal from volume. A small country with three submissions, two of which are about tourism, has a strong revealed specialization in tourism. A large country with two hundred submissions spread evenly across every topic does not specialize in anything in particular.

From Specialization to Relatedness: Building the Space

Now comes the key construction. We have a matrix of who specializes in what. The question is: which topics are related to each other?

The insight, borrowed again from economic complexity, is that you do not need external knowledge about what topics are semantically similar. You can infer relatedness from the pattern of co-specialization itself. If actors who specialize in “Drilling” also tend to specialize in “Marine Acoustics”, those two topics are related — not because the words sound alike, but because the same actors repeatedly treat them as part of the same portfolio of concerns.

We measure this by computing a proximity between every pair of topics. The proximity between topicsandis the minimum of the conditional probabilities: given that an actor specializes in, how often do they also specialize in, and vice versa? Topics that are frequently co-specialized in by the same actors end up close together. Topics that rarely share specialists end up far apart.

The result is a network — a weighted graph where nodes are topics and edges reflect co-specialization strength. This is the space of concerns. The interactive below walks through each step of the construction on a small synthetic example — step through with the arrows or click the dots.

Applied to the full ATS record, the construction produces the network below.

What the Space Looks Like

The space is not a random cloud. It has structure.

At the center sits a dense cluster of procedural and environmental topics — opening statements, exchange of information, environmental monitoring, environmental protection, liability, inspections. These are the institutional plumbing: the topics that almost every active member engages with to some degree.

From that core, thinner corridors extend outward in different directions. One arm reaches toward resource and science frontiers — drilling, sub-glacial lakes, marine acoustics, environmental domains analysis. Another extends toward tourism, site guidelines for visitors, and search and rescue. A third connects to operational coordination topics.

This geometry is not arbitrary. It reflects the institutional history of the ATS itself. The 1959 Antarctic Treaty demilitarized the continent and channeled strategic interests into procedure and science. The 1991 Madrid Protocol banned mineral extraction but allowed tightly regulated scientific drilling. In the space, this shows up as a resource-and-science corridor where capabilities with legitimate scientific value but potential strategic spillovers remain tethered to the procedural core — close enough to be monitored and negotiated.

Bridge topics organize the transitions. Inspections act as a compliance hinge between the governance core and frontier activities. Marine Protected Areas and Human Footprint connect the core to an access-oriented chain of tourism, pollution prevention, and site management. Climate change appears as a connector rather than a detached outlier.

Three Recurring Positions

If the space captures meaningful structure, the next question is whether actors cluster within it. They do — into three broad engagement modes.

Mode 1 — Coordination and Exchange. Newer entrants and smaller players: Switzerland, Malaysia, Estonia. They focus on scientific integration, education, and operational reporting — the entry-level concerns of a state finding its footing in the system.

Mode 2 — Compliance and Management. The practical middle ground: inspections, liability, protected areas, environmental management. This is where technical bodies and intermediate actors sit — UNEP, COMNAP, and Belarus.

Mode 3 — Strategy and Resources. The high-politics end: mineral resources, tourism pressure, multiyear strategic planning. Dominated by the system’s architects and longest-standing members — the United States, Russia, Australia, New Zealand.

These are not rigid blocs. Actors commonly bridge adjacent modes, and some major players like Australia maintain portfolios that span the entire spectrum while anchoring most of their weight in Mode 3. But the pattern is clear: mode position aligns more strongly with how long an actor has been active in the ATS than with external geopolitical categories like NATO, the EU, or BRICS. The structure is internal to the institution.

How Actors Move Through the Space

The space is not just a snapshot. Because the ATS record spans six decades, we can ask: how do actors change their positions over time?

The answer is that movement is local. When an actor picks up a new topic, it tends to be one that sits close to their existing portfolio in the concern space — a neighboring concern, not a distant one. Adoption odds drop sharply with distance. And once a topic is adopted, it tends to stick: ATS portfolios show strong persistence rather than year-to-year reinvention.

At the mode level, this translates into high stability. In rolling five-year windows, over 80% of actors stay in the same mode. Transitions, when they happen, are almost always to an adjacent mode. Direct jumps from Mode 1 to Mode 3 — skipping the middle entirely — are rare (about 1% of transitions).

This is the behavioral payoff of the space: it is not just a descriptive map, but a constraint on where attention tends to go next. Change in the ATS looks more like local recombination — extending an existing portfolio into neighboring concerns — than like random diffusion across the full issue topology.

Who Sets the Agenda?

If growth is local and path-dependent, then position matters. We asked: which actors tend to enter topics early, before the rest of the system catches up?

The answer is not a single dominant bloc. Agenda pioneers appear across all three modes. What distinguishes them is not which mode they sit in, but how they sit in it: early movers tend to have broader portfolios and stronger anchoring within their mode. They are not generalists drifting across the whole space, and they are not narrow specialists stuck in one corner. They are actors with a clear position and enough breadth to reach into adjacent concerns before others do.

This reframes influence in the ATS. It is less about formal status or raw volume and more about where you sit in the concern space and how you move through it.

The Space of Concerns as a Tool and Implications Beyond Antarctica

The space of concerns is, at its core, a way of making consensus institutions legible. When formal outcomes are unanimous and voting records are absent, you need a different kind of signal. Revealed patterns of attention — who repeatedly focuses on what — provide that signal. And the relational structure between those concerns — which topics travel together — provides the geometry.

For the Antarctic Treaty System, this reveals an institution that is neither uniformly governed nor cleanly partitioned. It is a shared arena with structured positions, local movement, and selective complementarity. The unanimity is real, but it sits on top of a richer landscape of differentiated and evolving concerns.

The method is not specific to Antarctica. Any institution that produces a long record of categorized activity — a legislature, a standards body, an international organization — could, in principle, be mapped the same way. The question is always the same: what does the pattern of repeated attention reveal about the structure that the formal record hides?