Yale professor Marissa King breaks social networks into three types: Expansionists (many contacts but lonely), Brokers (connect different groups but face trust issues), and Conveners (tight-knit circles but risk echo chambers). She says any big change needs all three. During COVID, men's networks shrank ~30% (lost shared activities), women's barely changed (relied on conversation). No investment picks discussed.
Yale School of Management professor Marissa King, in an episode of Invest Like the Best, discussed the science of social networks based on her book Social Chemistry. The core argument is that social networks can be categorized into three styles: Expansionists, Brokers, and Conveners, each with its o
Yale School of Management Professor Marissa King, drawing on her book Social Chemistry, explores the scientific classification and optimization methods of social networks. The core judgment: Social networks can be broken down into three basic structures—Expansionists, Brokers, and Conveners—and any large-scale social change or organizational success requires the simultaneous presence and coordinated operation of all three structures.
Marissa King argues that expanders have a social network size far exceeding that of ordinary people. The average person knows 650–900 people, while expanders know several orders of magnitude more. The criterion: if you know more than two people named Alan or Emily, you are likely an expander.
The core value of expanders lies in creating random connections—they possess enormous influence and reach, enabling them to link different groups together. However, King points out that expanders are actually the group most prone to loneliness. She cites the example of Hollywood legendary manager Shep Gordon, who was accompanied only by paid assistants at the end of his life. Research also shows that as organizational hierarchy rises, loneliness increases significantly—CEOs experience extremely high levels of loneliness.
Improvement suggestion: Expanders should shift toward the convener model, investing more time in existing relationships to build a truly trustworthy core circle, rather than continuously expanding new connections.
The defining feature of a convener's network is that friends know each other—if you host a barbecue or birthday party and everyone present knows one another, you are a convener. This structure fosters high trust and reputational advantages, but the downside is self-containment and the formation of echo chambers.
King emphasizes that convener networks are highly effective for behavioral change (e.g., forming exercise habits), but they lack information exchange with the outside world. The direction for improvement is to proactively inject diversity—by seeking "uncommon commonalities" with people from different backgrounds through curiosity to build bridges.
The hallmark of brokers is connecting social worlds that do not typically meet. The criterion: are you good at improvising on topics you do not fully understand? Answering "yes" strongly suggests you are a broker. This "chameleon" trait (high self-monitoring) is the strongest predictor of network type, far surpassing extraversion/introversion.
King identifies herself as a broker and points out the associated risk: Ron Burt's research shows that brokers are vulnerable to "reputation assassination"—being suspected because they operate between different worlds. Solution: become an "empathetic broker" who actively connects people rather than keeping them apart, which benefits both the individual and the organization.
King shared a key finding: comparing before and after pandemic lockdowns, overall social networks shrank by about 15% (approximately 200 people), but male networks shrank by nearly 30% (approximately 400 people), while female networks showed almost no change.
The reason lies in different maintenance methods: men sustain relationships through shared activities (sports, bars, etc.), which were disrupted during the pandemic; women maintain relationships through conversation, and their conversational ability was unaffected by the pandemic. King noted that, ironically, while women faced greater pressure in many aspects during the pandemic (especially working mothers), their social networks proved more resilient.
King pointed out that women tend to separate work and personal life, which brings them closer to the brokerage model and benefits work-life balance. However, women receive lower career returns from the brokerage role than men—because brokering behavior conflicts with the stereotypical "community-oriented" image of women, leading to backlash bias.
King criticizes the popular notion of "all-star teams," citing Google's Julia Rozovsky research: the core of high-performance teams is not the sum of individual capabilities, but collective dynamics. Key elements include:
King criticizes organizations for "following the herd" when managing informal relationships—three years ago, open-plan offices were trendy (hoping chance encounters would foster innovation), and now the shift is toward fully remote work. She points out that organizations rarely invest in understanding their own social capital, which is as important as human capital.
Specific recommendations:
Using the WTO protests as an example, King illustrates that environmentalists, union leaders, and other disparate groups can come together because three types of networks exist and operate simultaneously. Any large-scale social change requires:
King argues that online interactions intensify the tendency toward homogeneity—people are more likely to interact only with those who share similar views. The lack of chance encounters during the pandemic has further exacerbated political and racial polarization.
She emphasizes that workplaces are the domain with the highest social diversity—far exceeding schools, neighborhoods, and volunteer organizations. Therefore, workplaces are the most powerful venues for bridging divides, as organizations can create meaningful interaction opportunities between different groups through project assignments or spatial design.
King explains the biological basis of Dunbar’s number: the scale of human social networks is constrained by brain capacity and mentalizing ability, capping at approximately 150 individuals. This figure is reflected in the size of military company units, organizational units, and similar structures.
Even more striking is that the same network structure appears across different systems: human social networks, neural networks, ant colonies, and power grids all follow the same structural principles—dense hubs (convener structures) enable rapid dissemination, intermediary connections (mediator structures) link different clusters, and long-tail connections (expander structures) provide randomness.
King argues that this cross-system structural homology suggests the existence of a "Higgs boson of social systems"—understanding it could help optimize human systems.
This section is not applicable — this episode covers a social science topic and does not discuss specific investment positions.
1. King argues that the strongest predictor of network type is not extraversion/introversion, but "high self-monitoring" (chameleon trait) — this trait determines whether an individual is a broker or another type, and is more predictive than personality traits.
2. Male networks shrank by nearly 30% (approximately 400 people) during the pandemic, while female networks remained almost unchanged — the reason is that men maintain relationships through activities, while women maintain them through conversation, and conversational ability was not constrained by the pandemic.
3. Women receive lower career returns from the broker role than men — because brokering behavior conflicts with the female "community-oriented" stereotype, leading to backlash bias, even though the broker model offers women better work-life balance.
4. Expanders are actually the group most prone to loneliness — top executives such as CEOs experience extremely high levels of loneliness, and the improvement direction is to shift toward the convener model, investing more time in existing relationships to build a core circle.
5. The key to high-performing teams is not an all-star lineup, but psychological safety and equal speaking time — Google's research confirms that roughly equal speaking time among all members is the simplest metric for team optimization.
6. Online interactions exacerbate echo chambers, while the workplace is the domain with the highest social diversity — therefore, the workplace is the most powerful venue for bridging political and racial divides.
7. Human social networks, neural networks, ant colonies, and power grids follow the same structural principles — dense hubs, broker connections, and long-tail connections are three basic structures that exist across systems, suggesting the presence of something akin to a "Higgs boson of social systems."
8. The most effective network investment is to reactivate existing relationships — reaching out to people you haven't contacted in two to three years is more effective than expanding new relationships, and benefits both parties.