Abdullah Almaatouq

Who they are

At the time of the paper Almaatouq and Alsobay were at the Sloan School of Management at the Massachusetts Institute of Technology, Ming Yin was in the Department of Computer Science at Purdue University, and Duncan Watts held posts across computer and information science, the Annenberg School of Communication and the Operations, Information and Decisions Department at the University of Pennsylvania. The paper lists collective intelligence among its keywords, alongside complexity, problem-solving and team performance.

This page records what the 2021 paper establishes. Nothing else about Almaatouq's career is asserted here, because nothing else was verified for the article that cites him.

Key papers

Almaatouq, A., Alsobay, M., Yin, M., & Watts, D. J. (2021). 'Task complexity moderates group synergy.' Proceedings of the National Academy of Sciences, 118(36), e2101062118. doi:10.1073/pnas.2101062118. First author. Abstract read at the indexed record, with design details taken from the University of Pennsylvania and Wharton summaries of the study.

The design is a deliberate answer to a problem in the older literature. Earlier studies compared groups and individuals on different tasks, so that a finding of group advantage on one task and group loss on another could not be tied to anything about the tasks themselves. Here the task was held constant and only its complexity changed. Participants assigned students to rooms. Complexity was raised by adding students, removing rooms and adding constraints, such as students who must not share a room or be neighbours and students who must. The authors define complexity by the number of components in the task and the interdependencies between them. The PubMed Central record describes about 1,200 individuals who first solved tasks of varying complexity and were then assigned at random to work in interacting groups or alone.

Key ideas

The headline result is conditional. On complex tasks, groups were as fast as the fastest individual and more efficient than the most efficient individual. On simple tasks this advantage did not hold. The authors also quantified synergistic gains and process losses within interacting groups and found that the balance between them depends on task complexity, shifting at intermediate levels. Groups also generated more solutions and explored the space of possible solutions more widely.

The result gives empirical content to a distinction that runs through the literature on group performance. Ivan Steiner's 1972 formula treats actual productivity as potential productivity minus the losses from faulty process, so that a group can only fall short of its members. Richard Hackman added process gains, the things a group can do that no member could do alone. With both terms available, the open question is which tasks let the gains outrun the losses, and this study answers part of it: tasks with many components and many interdependencies.

The authors suggest that this may help reconcile conflicting findings in earlier work. A simple laboratory task, of the sort used in much early research, will make a group look worse than a real piece of knowledge work does, because simple tasks leave little for division of labour to gain and plenty of coordination to lose.

In the Wharton summary Watts is reported as saying that teams shine on efficiency while also wasting time, competing internally and sometimes falling into groupthink, and the researchers suggest that managers choose between individuals and teams according to whether the task is simple or complex and whether the best possible score or a near-best score reached efficiently is wanted. Those are interpretations offered in interviews rather than results of the experiment, and this site treats them that way.

For this site the paper does two jobs. It supplies the condition under which collective intelligence in the loose sense, a group doing better than its members, can be expected. It also changes how the Zurich social-influence experiment by Jan Lorenz and colleagues should be read: estimating a number is a simple task, so a crowd gains little from talking and can lose a good deal, while a complex task with many linked parts is the setting where dividing the work can pay for the cost of coordinating it. That second point is this site's inference from the two studies together, and neither study makes it.

---