Stable vs Fluid Teams

There is a small island in the middle of the South Atlantic Ocean called Tristan da Cunha. It is often described as the most isolated inhabited island in the world: home to only 250 people, nearly 2,800 km from the nearest continent, reachable only by a supply ship that sails from Cape Town roughly ten times a year and takes about seven days each way. The population is closed enough that only four or five family names exist on the island, and diseases like asthma affect more than half of the population due to the limited genetic diversity from a founder effect. You cannot simply move there, buy property, or open a business; permanent residency is extremely restricted and practically impossible for outsiders to obtain. One of the inhabitant described life there as follows: "it’s about living with the land, making do with what you have, and looking out for each other, no matter how small the island or how big the challenge. There’s no Tesco, no Amazon Prime deliveries, no endless options when it comes to food or entertainment. But what we do have is a tight-knit community, a deep respect for the environment, and a tradition of self-reliance that runs through everything we do. Life may be slower, simpler, and at times a little more chaotic, but there’s a kind of beauty in it—a beauty that’s harder to find in the hustle and bustle of modern life. And while you won’t find any Instagram-worthy beaches or fast-food outlets, you’ll find something much more rare: a place where everyone knows your name, and you’re never really alone."

At first, it looks like a paradox: an isolated place with only a few people who are far less lonely than most of modern society. One reason modern life produces more loneliness despite far more people and options is choice. When you have many possible options (for goods or relationships), you commit less deeply to any one of them, because you know that there's an alternative. This is sometimes called choice overload, or the paradox of choice. However, in a stable, closed community like Tristan da Cunha, relying on other people isn't optional; it's a necessity, and "thick" relationships dominate. When your livelihood depends on collective effort (fixing the only boat, sharing a harvest, responding to an emergency with no outside rescue coming quickly), individual interest and group interest naturally align. In modern society, those thick relationships have transformed into "thin" ones that are single-purpose, optional, and transactional. The problem is that, according to evolutionary psychologists, we are not wired for these "thin" relationships; humans evolved in small groups of roughly 150 people, where survival depended directly on the group. By contrast, modern environments let a person survive in near-total social isolation, and the result is a mismatch: deep belonging in small, stable groups has been eroded by choice overload and mobility, while the psychological need for that belonging hasn't changed at all.

The Same Pattern Shows Up in Teams

This pattern, it turns out, shows up in teams almost unchanged. Stable teams, the workplace equivalent of Tristan da Cunha, tend to perform better on tasks that require members to trust each other enough to admit mistakes, ask for help, and speak up. Organizational psychologists call this psychological safety, and it relies on mutual trust that come almost for free once people have worked alongside each other for years.

Contrast this with fluid teams, much more typical of modern organizational life, where interdependence is often artificial: you can fail without much personal cost, or succeed without much collective effort, because the team forms fast and doesn't stay together long. These teams can still function well, but the trust they have is different. Researchers call it "swift trust", built on professional role and category rather than personal history. A film crew, a surgical team, or a project task force can coordinate effectively on day one because everyone trusts the role, not the person. But this kind of trust is rarely as deep as the one that a stable membership builds.

Why is trust important? Many studies show that in teams with weak trust, mistakes are not surfaced, because admitting an error, especially your own, feels risky in a group where you can't be sure how it will be received. The same errors that get discussed in a stable team often stay hidden in a fluid one. More importantly, people do not challenge a decision if they do not trust each other; it is not competence that drives arguments, but rather whether the relationship has enough accumulated trust to make that feel "survivable", or at least less psychologically costly. A team that never challenges anything, or does not know its errors, stops learning. Unanimity does not mean correctness here, but rather that everyone independently decided it wasn't worth the social cost to object.

The Downsides of a Closed Team

However, stable groups aren't automatically better; there are real caveats to the stable-team advantage. A large scale study shows that performance rises as the team settles in, peaks around two to three years, and then declines because the team stops reaching outside for information and develop a bias against outside ideas. In general, overly stable groups have less tolerance for dissent, more pressure to conform, slower diffusion of new ideas, and conflict that's harder to escape. Groupthink and a lack of diverse viewpoints are real risks once people have worked together for years.

Some typical warning signs of an overly closed team:

What's missing across all of this is the same thing: diversity of thought, and tolerance for productive friction.

The Downsides of a Fluid Team

The opposite extreme has its own failure list:

What's missing across all of this is close to the mirror image of the closed team's problem: interdependence, accumulated trust, and group (or institutional) memory.

Pulling Toward the Middle

Just as Tristan da Cunha isn't the "good" form of society and individualism isn't the "bad" one, a fully stable team isn't simply better than a fully fluid one. They are two different sets of trade-offs. Necessity-driven interdependence builds specific capacities (trust, mutual reliance) at the direct cost of others (diversity of thought, tolerance for outside ideas).

A good analogy is the immune system. It has to be reliable and fast against threats it has already seen by fine-tuning the B cells that recognise a given threat (exploitation), while retaining enough diversity by randomly producing brand-new, untested B cells on the chance that one of them will match a threat nobody has seen before (exploration). The problem that both come with costs and the body has finite resources, so which one should it favour, and to what extent? An immune system that is too exploitation-heavy develops autoimmunity; one that is too exploration-heavy ends up with immunodeficiency.

Behavioural viewpoint

This exploration versus exploitation dilemma shows up in complex adaptive systems theory, well beyond biology, and is solved similarly everywhere it appears: any system that has to survive under resource-constrained conditions and faces a real failure cost tends to exploit, because refining a known-good solution has a more certain, immediate payoff than gambling on something untested. It invests in exploration only when it can afford the risk. For example, the immune system favours exploitation over exploration as it ages, since its resources gradually decline. In individual psychology, both prospect theory and status quo bias explain the same bias toward exploitation over exploration under scarcity and high stakes; the immediate payoff of the known method is usually worth more, psychologically, than a probabilistically better but uncertain alternative, especially when a failed exploration could be fatal rather than just costly. Since this was fairly common in the past, the less systematic but faster part of our brain favours exploitation. Of course, there are individual differences, hence there are risk-tolerant and risk-averse people.

A similar decision heuristic also appears at the level of groups or organisations. Take a closed lab versus a fluid lab: every group has to balance refining and efficiently using what already works against searching for new alternatives whose payoff is uncertain. Exploitation has faster, more certain, easier-to-measure payoffs, so it often wins over exploration even if the team members are risk-tolerant. Indeed, a study analysed decades of chemistry research and found that scientists systematically favour "tried and true" combinations of known reagents over novel ones, producing safer, more citable but less innovative work. This shows that group behaviour depends more on the rewards (here, citations), which are in turn a tunable parameter, than on whether the members are individually risk-tolerant.

The behaviour of grouping isn't explained only by the exploration-exploitation trade-off directly, since we tend to group even without resource constraints. Some of these explanations are still arguably about risk management and heuristics at the end that we learnt to decrease the uncertainty of the payoff (homophily and trust and reciprocity, favouring the familiar and the proven over the uncertain). But others are separate: identity (in-group favouritism, favouring one's own group even without any information to justify it) and a basic psychological need for belonging both evolved as intrinsic drives rather than heuristics (so they are an end in themselves, like food-seeking, and not a shortcut for estimating something).

Structural viewpoint

There is a different, structural (rather than behavioural) viewpoint on this trade-off in social science that connects back to the thick versus thin relationships mentioned earlier: bonding versus bridging capital. Bonding happens when you connect with people similar to you; bridging happens when you connect across difference. A closed group has bonding capital in abundance and almost no bridging capital, since essentially nobody new arrives and nobody leaves to bring outside experience back. PhD students who become permanent members of the very lab where they earned their degree, without ever leaving, are a textbook case of this. Katz's aforementioned seminal work on long-tenured R&D project teams found a curvilinear performance curve; team performance rises in the first few years, and then declines. This shows that teams that stop maintaining external relationships perform worse over time, even while everything inside the team looks perfectly fine. A healthy team or organisation needs both bonding and bridging capital: bonding gives people a secure base and a sense of belonging, while bridging is what lets a team innovate and solve bigger problems by pulling in outside expertise.

The sociology (bonding versus bridging) and organisational psychology (exploitation versus exploration) viewpoints are similar but not the same. Exploitation and exploration describe where a group's effort actually goes: toward refining what it already knows, or toward searching for something new. Bonding and bridging describe the shape of its relationships: dense ties within the group, or sparser ties reaching outside it. One is about behaviour; the other is about structure. A group can have one without the other. For example, someone on the team has contacts in five other labs or companies and attends outside conferences, but none of that outside knowledge ever gets pulled into the group's actual work. Structurally the group looks well-bridged; functionally it's pure exploitation. Conversely, a team can be tightly bonded, with almost no external ties, and still be internally exploratory by arguing out new approaches among themselves. So bridging makes exploration cheaper and more likely; it isn't exploration itself.

Calibrating the Trade-off

Most real teams aren't purely one extreme or the other. They are mostly hybrid and sit somewhere on a spectrum between Tristan-da-Cunha-style enmeshment and individualist-society-style diffusion. Fluid groups can contain stable sub-groups that don't rely on the rest of the team (cliques); stable groups can likewise show the symptoms of fluidity when personality mismatches exist within them. A good team leader's job isn't to pick one culture and install it. It's to notice which failure mode the team is currently drifting toward, and pull it back the other way.

Practical Levers

Most practical levers fall into one of two categories: structural levers, which change the shape of the network itself (who is tied to whom, who depends on whom, who holds authority), and cultural or people levers, which change what the group chooses to do with the network it already has. Many cultural, or people, levers encourage members to speak up honestly about mistakes and potential failures by lowering the social cost of doing so; others help spread new ideas within the team.

Structural levers

Cultural and people levers

Ranking the levers

I think the problem isn't that people are unaware of these levers, but that they apply them in an ad hoc manner rather than systematically. A reading group in a team with zero real task interdependence is far less effective than in a team where the outcome actually matters to everyone. Similarly, the "Devil's advocate" lever fails in a group whose members are, on aggregate, high in conflict avoidance.

In general, structural levers tend to have broader, more durable, and self-sustaining impact, since they're built into the structure and don't depend on anyone's discipline. However, they're typically also more expensive to install. Cultural and people levers tend to be cheaper, with more immediate, but narrower and more fragile impact: they aren't really self-sustaining, and habits die off if nobody actively maintains them.

I haven't found a study that ranks these levers by potential impact and cost, maybe because it's too context-dependent. Clearly, a group leader works under constraints, or more precisely, the rewards that are engineered above the group higher up in the organisation's hierarchy (e.g., you can't apply task interdependence if the system deliberately discourages cooperation among PhD students to hold them individually accountable). As the aforementioned study on chemical researchers showed, such constraints are more binding than individual or aggregate (group) dispositions. But at the same time, the applied techniques should also depend on the personality traits of the group members, and building an accurate group profile seems doable today. And then maybe a well-trained agent can propose something useful :-)