An Experimental Paradigm for Studying
the Formation of Self-Organized Groups

Holly Arrow (asst. prof., psychology), John Orbell (professor of political science), Katie Burns (Ph.D. student in psychology), and Scott Crosson (Ph.D. in political science completed in 2000 using the social poker paradigm).
Alums of the project: Ruth Bennett (M.A., psychology)

Software team, Web-based version: Version 1.0 was developed by Michelle Hart (M.S. computer science alum), working under the supervision of Professor Steve Fickas . Rama Lanka, (M.S., computer science), has documented the flow of interaction in different versions of the games, using UML (Unified Modeling Language) and Rational Rose. James Joule has developed a "bot" that plays social poker. Leigh Williams developed the navigation tool for version 1.0 using JDK 1.2 and Swing; Xingxing Wu has been working on a navigation system that extends the number of players. Ray Vukcevich, staff programmer for the psychology department, has worked on systems to extract data from the computer logs.

The Social Poker Computer-Mediated Interface, 1.0 Release is now on the web. The social poker site is based on the Confabulation environment.

The Social Poker project, including the web-based version, is based on work supported by the National Science Foundation under Grant No. 9729320, Decision, Risk, and Management Science Program , PIs Holly Arrow and John Orbell.

Undergraduate researchers:
Ryan Hampton, completed an honors project in 1999.
Catherine Peterson completed an honors project in 2000.
Ryann Crowley completed an honors project in 2000.
Their findings are summarized under Results So Far.


Social poker is a game designed to investigate the pattern of individual and collective decisions that lead to the formation of self-organized groups out of a pool of individuals who hold complementary resources.

In social poker, every player in an initial pool is dealt two or three cards from a standard deck. The objective of the game is to form a poker hand by joining together with other players. Players are not permitted to trade their cards -- they must instead combine cards with others to form a hand. Since the poker hand requires five cards, the minimum number of people necessary to form a hand is three (which yields 6 cards).

A hand earns a payoff for the group (for example, a group may earn $10 for a full house), which then must be divided among members. In one version of the game, this is decided via discussion and the group collectively fills out a sheet specifying who gets what. In another version, members make private "claims" on these earnings. If the collective claims exceed the earnings, the claims of all members are reduced by 50 cents for each $1 that the total claims exceed the amount earned. This presents a potential for exploitation, which can enrich an individual at the expense of his or her fellow group members.

Rama Lanka , a member of our software team, has documented the flow of interaction in different versions of the games.


Data collected from the game will be used for research on the impact of differential power, information flow patterns, and initial social contacts on the quality of outcomes at the individual and "societal" level both for a single game and for a series of games.

Some members of the research team are interested in how people use member selection and promising as a strategy for protecting their groups against exploitation from members who are tempted to take more than their share of the "goods" that a group produces. Others are interested in how newly formed groups make decisions about allocating resources, and whether allocation norms change over time. We are interested in the stability of group composition when the same pool of people play the game repeatedly, and in the emergence of self-reinforcing patterns of behavior in groups of players who continue to seek one another out across multiple rounds of play.

We are also exploring differences in interaction patterns when people interact face-to-face or via a chat-based computer interface, which is still under development.


Ryan Hampton's honors project (1999) looked at gender segregation in the group formation process, using a note-passing version of the social poker game. He found that when gender composition was balanced, it played no role in people's preferences for group partners. When gender composition was skewed, those in the minority preferred on another as partners.
Catherine Peterson's honors project (2000), looked at the effectiveness of strategies to get included in emergent groups. She found that the most common strategy was attempting to tag along with others who had already coalesced into a group. The most consistently effective strategy was to pair up with a more powerful player, who would then bring the hanger-on into the group.
Ryann Crowley's honors project (2000) looked at gender differences in trust. She found that women were both more trusting and more trusted than men. Women also trusted women more than they trusted men.

For People in SGL or on the SP project: You can link to the web page for sp developers to see Michelle's commentary on Version 1.0 and also to find logs of pilot runs.

In Stage 1 development (summer of 1997), Maan Al-Fudhail, Jose Bencosme, and Leigh Williams (first stage) created designs for several user interfaces from different phases of the game. Some of their ideas have been incorporated into Version 1.0 of computer-mediated social poker.

This page last updated June 24, 2001.
Comments should be directed to Holly Arrow,