How do individual decisions affect group decisions?

Postdoctoral fellow Colin Twomey looks to fish behavior to explore the dynamic between individual and group decision-making.

Colin Twomey studies how groups, both human and animal, make collective decisions. His research covers a variety of topics, including fish behavior and human color perception.

Group decisions are sometimes made based on the action of a single individual. If one fish sees a predator in the distance, for example, it might make the split-second decision to dart off into hiding, triggering a ripple of movement throughout its group.

Humans, too, make group decisions, sometimes with the aid of sophisticated systems, such as government elections. But people also make group decisions at more fundamental levels, such as the emergence of a shared vocabulary for describing color. 

Colin Twomey is a MindCORE postdoctoral research fellow at Penn. The Mind Center for Outreach, Research, and Education was established as one of the School of Arts and Sciences’ key endeavors under the “Mapping the Mind” initiative identified in the School’s strategic plan. Twomey studies both human and animal collective decision-making. His interest lies in behavior cascades, or how an individual or small subset of individuals can inspire large-scale decision-making within a group. He designs mathematical models and computational tools to test theories about group behavior, evaluating his theories in both the lab and the field. 

Recently, Twomey has focused his research on how visual systems recognize patterns and how these systems constrain language. He works closely with several professors in the School of Arts and Sciences, including Joshua Plotkin, professor of biology, whose lab Twomey joined; Gareth Roberts, assistant professor of linguistics; and David Brainard, professor of psychology. Specifically, he’s turned his attention to color perception and description.

“English has eleven basic terms for describing colors,” says Twomey. “These terms, such as red, blue, or green, are, of course, different in Turkish, Japanese, and other languages, and we see quite a few changes in the number of terms used in different languages, but one thing remains consistent: The sensory physiology of the eye is the same across all people when recognizing these colors.”

Though different languages’ vocabularies for describing color evolved separately all over the world, Twomey thinks there might be a common principle underlying this and other manifestations of social evolution, such as the formation and change of political parties over time.

Twomey hopes to find a generalized framework about how we see and describe color, among other things, to better understand how different species respond to sensory information and then make collective decisions that drive social evolution. While this work is in its early stages, Twomey has closely studied other forms of collective decision-making behavior in animals.

As part of a multi-institutional team, Twomey and colleagues focus on fish shoals, an aggregate of fish that group together, sometimes fleetingly. The researchers have studied how individual decisions affect group decision-making by studying these fish groupings in the lab and the field

“When we look at behavior cascades, or when group behavior is initiated by one individual or a small group of individuals, sometimes we see this behavior spread throughout the entire group,” he says. “But sometimes they don’t. We wanted to know what tells us whether or not something is likely to spread.”

Twomey and a team of researchers studied wild coral reef fish in Mo'orea, French Polynesia, to better understand the individual's role in collective decision making. (Photo credit: Twomey)

To better understand animal decision-making, the team set up a field site in Mo’orea in French Polynesia. They studied wild coral reef fish by installing a rebar and PVC-pipe frame on the reef bed, equipped with cameras and custom computer vision software, which allows Twomey and colleagues to track each fish.

Using a waterproof tablet computer, the researchers displayed visual stimuli that appeared as a threat to the fish. Each species of wild fish responded similarly, and Twomey precisely mapped their fright response and getaway path with his computer vision software.

“When fish are presented with a scary stimulus, they coil up and shoot off in a different direction,” he says. “They make a really fast decision: ‘Do I flee, or not?’ Swimming away from a group can be costly because it costs energy and can leave an individual isolated.”

By analyzing how each fish responds to stimuli and studying the fish’s location and subsequent field of sight upon recognizing the threat using algorithms, Twomey and team discovered that, not only would individual behavior sometimes inspire group behavior, but the fish’s physical location in the group was also critical to how it responded.

After analyzing how many other fish appeared in an individual’s view during the “startle” stimulus, the team found an individual was less likely to flee when its sightline encompassed more members of the group. The fewer members in sight, the more likely an individual would flee, likely because the fish perceived the group size as smaller, which would offer less protection. The researchers published their findings in a recent paper in the journal Proceedings of the National Academy of Sciences.

Twomey hopes to apply these findings to studying decision-making within human groups.

“There could be similar questions at play when we think about behavior transitions in human groups,” he says. “Maybe a huge group is celebrating the Eagles’ Super Bowl win on Broad Street, and everything is going great, but then suddenly the canopy at the Hilton collapses, and things get scary. How do individuals respond based on their place within the group, and how can we improve crowd safety?”

Twomey’s earlier research was supported in part by the National Science Foundation Grant IOS-1355061, Office of Naval Research Grants N00014-09-1-1074 and N00014-14-1-0635, and Army Research Office Grants W911NG-11-1-0385 and W911NF14-1-0431.