Consensus Proposal/Google Faculty Research Awards
Application for Google research grant program. Submitted August 1 2011.
The application is submitted using a Google Spreadsheets interface.
The proposal should be 3 pages, including contact information.
Budget should be a separate single page.
We also need to submit our CVs. Mine is here.
Proposal for Google Faculty Research Award:
Dynamics of Democratic Deliberation
Project title: Dynamics of democratic deliberation
Junling Ma, Ph.D.
Mathematics and Statistics
University of Victoria
PO BOX 3060 STN CSC
Victoria, B.C. Canada V8W 3R4
Abstract. We propose to contribute to both the political theory of democratic deliberation and the quantitative theory of search heuristics by developing mathematical models of deliberation as a collaborative search process. We will use mathematical analysis and simulation to investigate the efficiency and effectiveness of a range of realistic exploration and negotiation processes. Our research is to be conducted in public, in an online “open lab notebook”, and we will publish in peer-reviewed literature as well as in media accessible to participants in democratic processes and democracy movements.
The problem of how to reach agreement among parties with divergent interests is fundamental to human existence, and is only becoming more crucial as individuals, firms, and governments become increasingly interdependent across national and cultural boundaries. Political scientists’ descriptive study of democratic deliberation is deep and rich, but mathematical treatments — an important way to explore and evaluate possibilities systematically — generally consider only a single yes/no choice. Mathematical theories of search in complex problem domains are well developed in evolutionary biology and computer science, but much less so in the social sciences, where the complexity is greater.
The art and science of designing efficient, effective processes for problem solving in groups are not well-developed. Even in crucial situations such as global climate policy negotiations, we have no clear understanding of what processes and protocols are most likely to facilitate successful outcomes.
This project will address this gap by improving our theoretical understanding of deliberation and negotiation processes in complex problem domains. We will contribute to the theory of democratic deliberation by analyzing models of deliberation in complex problem domains, and extend the theory of search processes by developing models based on democratic problem solving.
The fundamental research question is design of processes and protocols for collective deliberation — that is, how can a group efficiently and effectively find a mutually agreeable solution to a problem when participants may disagree on what is acceptable. We treat this simultaneously as a problem in political theory to be addressed by modeling, as a possible source of new approaches to search algorithm design, and as a mathematical study in complex system dynamics.
We model collective deliberation as a problem of search in a complex problem space. To begin with, we use models in which potential solutions are strings of bits of a given length; as research progresses we will consider whether other search spaces require different approaches. Each individual associates a numerical utility value to each element of the search space, using a utility function that may have some or no correlation with others’ utility functions, and which may or may not assign correlated utility values to similar solutions. Utility values above a certain value are considered acceptable.
The primary research question, then, is what processes work well for locating a solution that is acceptable to all or most of the participants. That is, it is a question of search algorithm design which connects the mathematics of search in complex fitness landscapes to the existing social problem of effective democratic process.
Connections to existing projects
There is a bit of existing research on group processes that takes a similar approach. Scott Page’s models of collective problem solving  consider collaborative search on a Boolean landscape, yielding the significant result that even a small amount of diversity in the group can significantly improve the group’s effectiveness. Page’s work does not consider diversity in utility, however, only diversity in search heuristics, making our research program a significant departure from his. We will consider both forms of diversity, and how they affect each other.
Hiroki Sayama’s team at SUNY Binghamton has done interesting work modeling collective deliberation and problem solving with diversity in utility ; ; ; , and we will engage in depth with their results. Sayama’s team is currently focusing primarily on human-subject experiments and detailed approaches to particular real-world problems, rather than on their quantitative models. We are well positioned to bring more explicit mathematical analysis to the program, and to extend the research further and in new directions, including distributed processes and optimization of time spent searching.
Our research questions will deepen and broaden as the research progresses, and as we expand our familiarity with the existing multidisciplinary literature in political theory, social psychology, computer science and mathematics. A number of important directions already suggest themselves.
We will begin by developing basic results for small groups with all-to-all communication, looking in depth at efficiency and effectiveness of search heuristics, with attention to the effect of scaling to larger group sizes. Subsequently we will move on to alternative network structures and process design in larger groups.
Deliberation in large groups may require a fundamentally different approach than in small groups, because it is not possible to bring all participants together in a face-to-face discussion. This can be seen as a problem of distributed search, requiring some means to integrate contributions from multiple subgroups of the full group. Existing research on distributed algorithm design can be brought to bear on this problem, and may suggest novel possibilities for group process. Also, existing theories of group process can be reframed in terms of algorithmic theory, a potentially powerful way to compare or extend them.
Deliberative democracy, in political theory, is closely connected to the proposition that participation in democratic process can change participant’s preferences. Deliberation can succeed by changing people’s minds, not only by allowing them to negotiate solutions compatible with their existing preferences. It will be important to consider changing preferences in our research, not stopping with fixed, diverse preferences.
Much political science research in deliberation focuses on incentives and selfishness, asking when it is in participants’ interest to misrepresent their true preferences. While it is valuable to ask simply what processes are effective when participants use them, it is also important to ask whether participants are likely to subvert them by “gaming the system”, and how to design processes to avoid perverse incentives.
A major issue in actual deliberations and negotiations is time: a problem that could likely be solved given adequate discussion can not be solved in practice because there is limited time for deliberation. We will evaluate candidate processes for search time as well as for quality of eventual outcome. This question is clearly relevant to design of software search algorithms as well.
We are both applied mathematicians with experience modeling biology and social processes, and are applying both mathematical analysis and computer simulation in tandem to model these collaborative search processes.
Simultaneously, we are using this research project as an opportunity to contribute to the growing movement toward open research. Using software one of us has written to facilitate user-friendly online publication of reproducible research materials, we are maintaining all software, data, documentation, and discussion involved in the research on an openly-readable wiki website. We will conduct the research openly in this way throughout the course of the project, inviting access and participation from interested scholars and lay people worldwide.
This project is not currently funded, so we have only generated preliminary simulation results to date. These results are posted on our open research wiki . In these simulations we model several processes in which one “proposal” is on the table at a time, and replacement proposals are generated and then accepted or rejected. The simulation framework (written in C++ using the Boost libraries) is designed to extend readily in the multiple research directions we have outlined. The early results are suggestive of what can be done with a more rigorous modeling process: in these models, the group is more likely to reach an agreeable solution when
Group size is smaller.
There are more degrees of freedom, that is, when there are more bits in the search space.
Participants propose incremental improvements to the group, rather than jumping from the current proposal to a local optimum of their utility and proposing that.
Participants accept more alternative proposals when they aren’t satisfied by the current one.
In this year of research, we will publish at least one original paper in mathematics, political theory or computer science journals. We will also publish our results in a form accessible to non-mathematical practitioners of democratic deliberation. We will begin with the small-group, all-to-all problem, and as time permits, we will develop our research in as many additional directions as is feasible.
We expect this research program to contribute to multiple fields of research: in political science, we can illuminate the result of bringing complex problem domains into the study of deliberative democracy, which currently tends to consider only a single yes/no choice at a time; in mathematics, we have an opportunity to develop new results colliding the theories of optimization and dynamics on complex landscapes with theories of preference aggregation, voting and game theory; and in computer science, we can explore possibilities for designing new search heuristics. We also expect to help generate possibilities for action in real situations requiring deliberation, from small groups to international negotiations, thus contributing directly to positive global change in a century of staggering turbulence and possibility.
Page, Scott, The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies, Princeton University Press, 2007.
Klein, M., P. Faratin, H. Sayama, and Y. Bar-Yam, Protocols for Negotiating Complex Contracts, IEEE Intelligent Systems, November/December 2003, 32–38.
Farrell, D., H. Sayama, S. Dionne, F. Yammarino, and D. S. Wilson, Evolutionary perspective on collective decision making, presented as a talk at the Seventh International Conference on Complex Systems (ICCS2007), Boston, Massachusetts, 2007.
Sayama, H., D. L. Farrell, and S. D. Dionne, The Effects of Mental Model Formation on Group Decision Making: an Agent-Based Simulation, Complexity 16:3, 49–57, 2010.
Dionne, S. D., H. Sayama, C. Hao, and B. J. Bush, The role of leadership in shared mental model convergence and team performance improvement: An agent-based computational model, The Leadership Quarterly 21: 1035–1049, 2010.
Worden, L., Consensus Code/Stats, Lee Worden Research Wiki, 2011. http://lalashan.mcmaster.ca/theobio/worden/index.php/Consensus_Code/Stats.
Estimated expenses for Google Faculty Research Award
|One Graduate Student|
|A laptop computer||2000 USD|
|Group meeting||3000 USD|
The majority of the requested budget will be used to pay the salary for the Co-PI, who is currently a postdoc researcher at UC Berkeley with a salary of 65,000 USD. We would like to maintain the same salary for him.
We propose to support a graduate student at University of Victoria for one year, who will implement computer simulations.
We will buy a laptop computer for the graduate student.
We propose to travel to one international conference to present our results.
The Co-PI is based in northern California. We propose to bring the Co-PI to University of Victoria for an extended visit (a month) to coordinate the research.