Decider Group Size, Composition, and Interaction Pattern
It’s easy to assume that a group is just the sum intelligence of its component members, but the assumption is wildly off. Across a number of studies, Woolley et al. showed that while the average intelligence of group members and the maximum intelligence of individual group members are correlated with a group’s intelligence (with the maximum intelligence of the smartest group member having a higher correlation than the the group’s average intelligence), the highest predictor of a group’s intelligence on a task was the group’s intelligence at performing tasks in the past. In other words, groups have a consistent and internal intelligence measure like some sort of “group IQ” that has been shown to remain consistent over at least several months—this is where the concept of “collective intelligence” comes into play.
Research by Woolley et al. has also shown that things you might initially think were contributing to this group intelligence score had little actual correlation to it (e.g., group cohesion, motivation of group members, and satisfaction of group members). On the other hand, small things that one might not expect to matter were correlated with group intelligence, like the number of speaking turns group members took (meaning groups in which a few people dominated the conversation were less intelligent), the proportion of the group that was female (more is better), and the average social sensitivity of group members.
In this case, social sensitivity was measured with the Reading the Mind in the Eyes Test in which participants are asked to detect thinking or feeling expressed in pictures of other people’s eyes in an effort to measure people’s theory of mind. Females typically score higher on these kinds of tests, which is theoretically why groups with more females show greater intelligence.
Other findings align with our expectations, such as studies that show groups made up of individuals that had moderately diverse “cognitive styles” outcompeted groups with cognitive styles that are too aligned as well as those with cognitive styles that are too disparate.
In general, teams appear to work best when the team members exhibit:
- Psychological safety: Team members feel safe taking risks and being vulnerable in front of one another.
- Dependability: Team members can reliably expect to get things done on time and with a degree of excellence.
- Structure and clarity: Team members have clear roles, plans, and goals.
- Meaning: The work at hand is personally important to team members.
- Impact: Team members think work matters and creates change.
Group size also plays a role in group intelligence, but the research here is all over the place. Most studies seem to show that teams of three are best at managing funds, while groups of five are second best (though one study by Richard Hackman argues groups of four are best). It looks like once you go beyond six individuals in a management team, the intelligence of the group’s decisions consistently declines with each additional member. Despite the jumble of research findings, group size as it impacts group intelligence can largely be summarized as: Aim for groups ranging from three to four members and do everything you can to avoid groups with more than six participants.
All this is not to say that groups cannot be smarter as they increase in size when making individual decisions instead of management decisions. However, harnessing the intelligence of large groups requires decision markets—structures that allow huge groups of people to weigh in on specific potential outcomes. To get an idea of just how intelligent decision marketplaces are, when CBS had a human swarm intelligence place a bet on the Kentucky Derby, they correctly predicted the first four horses, in order, defying 542–1 odds and turning a $20 bet into $10,800.
Such systems have even been created by companies, such as Hewlett-Packard, in which salespeople were allowed—in one experimental endeavor—to buy and sell estimates about the future sales of their printers. This created a dynamic prediction market that was much more accurate than any previous method of prediction Hewlett-Packard had used.
A Weird Use Case for Futures Markets
Anyone who knows us knows we won’t shut up about falling birth rates. We often claim out of hand that government interventions will make little difference given that Hungary spent 5%[[16]](#_ftn16) of its GDP on the issue last year (2021) and only managed to increase its birth rate by 1.6%[[17]](#_ftn17). That said, we heard one really interesting idea on this front—one we find completely insane and unimplementable, but which could serve as inspiration for something less crazy.
Essentially, the idea is to take a person at birth and allow companies or individuals to “buy” stakes in future tax revenue generated by that individual. Half of the taxes this person pays throughout their life would go to the state and half would go to owners of their “stock.” A portion of this lump sum payment spent to buy an individual’s “stock” would be given to the child’s parents as a “reward” for having them.
This system prevents a state from pretending that the value each kid will yield in potential tax revenue is equal while also preventing things like racism, sexism, or bigotry from being big players in those judgments, as any company that allowed bigotry to unfairly bias their judgment would lose money in the long run.
Better, this investment model would give a company that owned a stake in an individual’s future taxes an incentive to invest in said individuals’ education, mental health, and early career development.