KI - Künstliche Intelligenz

, Volume 27, Issue 1, pp 45–52

Solving Wicked Social Problems with Socio-computational Systems

  • Joshua Introne
  • Robert Laubacher
  • Gary Olson
  • Thomas Malone
Research Project

DOI: 10.1007/s13218-012-0231-2

Cite this article as:
Introne, J., Laubacher, R., Olson, G. et al. Künstl Intell (2013) 27: 45. doi:10.1007/s13218-012-0231-2

Abstract

Global climate change is one of the most challenging problems humanity has ever faced. Fortunately, a new way of solving large, complex problems has become possible in just the last decade or so. Examples like Wikipedia and Linux illustrate how the work of thousands of people can be combined in ways that would have been impossible only a few years ago. Inspired by systems like these, we developed the Climate CoLab—a global, on-line platform in which thousands of people around the world work together to create, analyze, and ultimately select detailed plans for what we humans can do about global climate change.

The Climate CoLab has been operating since November 2009, and has an active community of thousands of users. In this article, we outline some of the challenges faced in developing the system, describe our current solutions to these problems, and report on our experiences.

Keywords

Collective intelligence Collaborative planning Climate change 

1 Introduction

Many important decision-making problems in the real world are so-called “wicked problems”—problems for which no single computational formulation of the problem is sufficient, for which different stakeholders do not even agree on what the problem really is, and for which there are no right or wrong answers, only answers that are better or worse from different points of view (see, e.g., [2, 3, 4]). For example, most social problems (including the environment, health care, poverty, education, and crime) are wicked in this sense, as are many problems in management (including strategic decision-making and product design).

The problem of global climate change is a wicked problem. In fact, many people would say it has some characteristics that make it especially challenging (or “super-wicked” [5]): Time is running out, there is no central authority that can implement a solution, and it is truly universal: it affects every one of us and is affected by all of our actions.

Left to their own devices, scientists, journalists, politicians, businesses, and consumers will ultimately do something about this problem. But the inefficiencies, delays, and distortions of traditional mass media, political decision-making, markets, and scientific publication mean that the results will almost certainly not be as good as we might hope. Fortunately, however, in just the last decade or so, a new way of solving global problems has become possible. Examples like Wikipedia and Linux illustrate that it is now possible to combine the work of thousands of people in ways that would have been impossible only a few years ago. Inspired by systems like these, we developed the Climate CoLab—a global, on-line platform in which thousands of people around the world work together to create, analyze, and ultimately select detailed plans for what we humans can do about global climate change.

Since the CoLab was first launched to the public in November of 2009, it has attracted a diverse community of thousands of users from around the world who have used the platform to generate dozens of candidate solutions to different parts of the problem. In this article, we describe the challenges we have faced in developing the system, our current solutions to them, and report on our experiences.

2 Background

When people involved in making group decisions can only use communication tools like face-to-face meetings, telephones, and paper-based communications, it is very difficult to have more than a few people deeply engaged in analysis and decision-making. Even with modestly sized groups of people, groups may experience a range of inefficiencies and process losses [6]. Group decision support systems (GDSS) have sought to grapple with such problems [7]. But the size, complexity and open-endedness of decision-making for large-scale social problems such as global climate change far outstrips the capabilities of traditional technological solutions.

A new generation of social-computational systems (like Wikipedia, Linux, and InnoCentive) may be able to pick up where decision support technologies have left off. These systems leverage the combined efforts of very large groups of people to solve complex problems and create large-scale products. Enabled by cheap, fast access to the Internet, these are often referred to as “collective intelligence” systems [8].

There are many examples of such systems, but there is no clear recipe for their development. Creating a collective intelligence platform to solve global climate change requires novel solutions to a range of socio-technical design problems. In our initial platform development, three such challenges have been paramount: breaking the problem into pieces and subsequently recombining solutions, predicting the impacts of proposed solutions, selecting good solutions. Each of these challenges, and our current solutions to them, are presented below.

2.1 Designing the CoLab: challenges and solutions

The Climate CoLab is a publicly accessible website (http://climatecolab.org) where members collaborate to create proposals during contests that address aspects of climate change. Proposals are similar to wiki articles in that they can be edited on the site and changes reverted. There is also a dedicated page for conversation about each proposal. Proposal authors have the ability to configure whether any community member can edit the proposal or if editing privileges are available by invitation only. For some kinds of proposals, it is possible to attach a model run that supports claims made in the body of the proposal.

In addition to writing proposals, some members of the community play special organizational roles. A team of Moderators monitor the site for spam and inflammatory or off-topic content. Fellows are students and concerned citizens who encourage and moderate contest activity. Advisors are experienced professionals who frame the focus of contests, provide input on proposals, and help to bring contest results to key stakeholders. Finally, the Expert Advisory Board and broader Expert Council, which include some of the most respected climate change researchers in the world (see [9] for a list of current members), provide general advice about the project and review specific content on the site.

We have faced a variety of design challenges in developing the CoLab. One challenge is how to break the problem down into manageable pieces and subsequently weave the solution back together. Others have focused upon aspects of this problem in relatively constrained, well-behaved domains (e.g. [10, 11, 12]), yet the domain of climate change is vast, dynamic, heavily interconnected across different dimensions, and highly uncertain. For instance, one dimension of the problem concerns energy production from fossil fuels and the resulting emission of carbon dioxide. Yet these concerns cut across geo-political boundaries and are also constrained by social and economic factors, all of which will change in an uncertain manner over time.

We have sought to use existing knowledge to develop an initial decomposition for the climate change problem. The climate change community at large has over time converged upon a functional decomposition of the climate change problem, and this is reflected in the structure of its various working groups and reports. We have sought to codify this structure in a taxonomy, and to use this taxonomy to focus the CoLab members on sub-problems. In developing the initial draft of the taxonomy, the Climate CoLab team relied greatly on the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), especially the contributions of Working Group 3 on mitigation [13] and Working Group 2 on adaptation [14]. The team then circulated the draft taxonomy to multiple experts, and revised it based on their input.

The current version of our taxonomy defines three dimensions:
  • What action is being taken—e.g. improving home efficiency, decarbonizing electric power, persuading Congress to pass climate mitigation policy, etc.

  • Where the action is being focused—e.g. in developed countries, in the United States, in China, etc.

  • Who is taking the action—e.g. individual citizens, utilities, any organizations or individuals, etc.

Each of these dimensions is further specified with one or more hierarchies. Figure 1 shows a partial view of the “What” taxonomy; the complete taxonomy can be found at [15].
Fig. 1

Partial view of the “what” dimension of the taxonomy. The view is expanded to the second level

The taxonomy defines a map of the climate change problem space; one might envision the space as a three-dimensional matrix (with who, what, and where as axes). Entries in each dimension of the taxonomy are used to specify regions of the matrix. Not all addressable regions within this space present problems worth solving (e.g. it might not make sense to ask how church groups can help metro regions cool the ocean to offset climate driven temperature increases), but we believe most problems worth solving can be located within the space. Because dimensions are organized hierarchically, the taxonomy contains initial information about how sub-solutions can be integrated to cover larger portions of the problem space.

The taxonomy is used in two ways in the CoLab. First, each contest (and its associated proposals) is located within a region of the problem space defined by the taxonomy (see Fig. 2). As the site accumulates content, this will serve as a useful way to index and search through solutions developed by the community.
Fig. 2

The active contest index. Note that each contest is associated with a region of the problem space using the taxonomy

The taxonomy can also be used to specify relationships between contests on the site and guide solution integration. For example, in the currently active contest cycle, an initial set of contests about focused regions of the problem space is underway. After these contests end, we will launch contests that focus on larger regions of the problem space that contain the areas of the problem space explored in the initial contests, and community members will then be invited to build integrated proposals leveraging work done in the initial contests.

Another challenge is to predict what will happen if a proposal is implemented. Computer simulations that make such predictions have served as the cornerstone of policy discussions about climate change, but access to these models is mediated by the experts who run them and interpret their results. This arrangement puts a bottleneck between stakeholders and the results they require to make decisions. For example, this bottleneck makes it much more difficult for a group to explore a diverse set of planning possibilities without active participation of the modeling experts at each step of the way.

Other fields in which modeling and simulation are heavily used for policy making have begun to develop systems that incorporate access to a model in order to eliminate this bottleneck (e.g. [16, 17]). Yet the field of integrated climate assessment models, the primary type of tool used to inform policy deliberations about climate change, is large and notable for its divergent assumptions and often conflicting results. Providing non-technical end users with comprehensive access to these models requires an approach for synthesizing results from multiple models and communicating effectively about uncertainty and the sources of divergence between models. There are also social and organizational hurdles to enlisting the help of model authors, which is required if their models are to be made publicly available.

To facilitate a solution to this problem, we have created a web-service called ROMA (Radically Open Modeling Architecture) that wraps existing simulation models (hosted either externally or internally) in a common API and publishes a web-accessible endpoint for running them [18]. In addition to providing a layer of abstraction around existing models, ROMA can transform Microsoft Excel spreadsheets into runnable simulations and also connect models to create composite models.

The CoLab generates user interfaces for models hosted by ROMA (see Fig. 3). Proposal writers use these models to validate claims about their proposals. One integrated model, which is a composite of multiple models that are connected by ROMA, is currently available to the CoLab community. This model takes as input a set of projected changes in fossil fuel emissions, disaggregated by region, and global deforestation/aforestation, and predicts changes in atmospheric carbon dioxide concentration, global temperature, sea level, as well as economic costs, economic damages of climate change, and negative qualitative impacts to human and physical systems such as agriculture and health. More information about the specific models used can be found on the Climate CoLab web site [19].
Fig. 3

The CoLab modeling interface

For some kinds of proposals, the existing simulation model cannot provide useful projections of future impacts (e.g. estimating the impact of geo-engineering on global temperature change). In these cases experts are asked to validate claims about impacts. In the future, we hope to continue adding models to support a broader range of proposals, and have engaged leaders in the modeling community to do just this.

A third challenge concerns how to focus the community’s attention on the most promising proposals being explored and, ultimately, how to select the best ones. The simplest way of doing this is with various forms of on-line voting and rating (e.g., Reddit, Amazon, Netflix). Sophisticated versions of this include pairwise voting [20], preferential voting [21], and range voting [22]. But pooling the opinions of many only yeilds good results when the participants have the expertise required to make informed judgments [23]. Thus, our challenge has been to enable collective decision making that incorporates the expertise required to identify good, feasible solutions.

Our current approach to this problem is to engage a panel of expert judges to select a set of viable alternative proposals in each contest that community members then vote upon. Contests unfold as follows:
  1. 1.

    Community members create proposals within the context of a contest, which poses a question about a specific aspect of the climate change problem.

     
  2. 2.

    At the end of this first phase, domain experts evaluate proposals for feasibility, novelty, likely impact on climate change, and presentation quality, and select a set of finalists to move on to the second phase. Judges also provide comments and critiques for each of the finalists.

     
  3. 3.

    Proposers then have a period of time to improve their proposals, after which the community and judges vote to select the most desirable ones.

     
Awards are given for both those proposals that receive the most votes, and those selected by the judges. This strategy has successfully increased the feasibility and quality of proposals selected by the community [1].

3 Experience

We have been collecting standard web analytics data via Google Analytics since 12/07/09. As of Sep. 18, 2012, the CoLab had 43,460 unique visitors from 175 countries. Of these, 3,895 (9 %) have registered for the site. Figure 4 depicts site growth and visits per day.
Fig. 4

Membership growth and site visitors over time (Color figure online)

Prior to 2012, contests were organized one or two at a time on an annual basis. Large outreach efforts were undertaken to build participation for these contests, using both social media and traditional channels. Since the site launched three such contest cycles have been completed. Site traffic and membership growth have been heavily influenced by these cycles, with the voting period of each contest driving a large number of visitors and new registrations. Thus contests have served as a valuable mechanism for building interest and membership.

Closer examination of site activity demonstrates that the initial contests contributed to membership growth, but there was relatively little substantive activity between contests (Fig. 5). As the member base has grown though, we have begun to see evidence of ongoing site activity that is not closely tied to the contest schedule. Members have continued to develop proposals, and have had ongoing discussions about the site design, the taxonomy, and new contest topics. We interpret this as evidence that the community is becoming self-sustaining.
Fig. 5

Categorized site activity. The y-axis reflects individual actions tracked by the system, such as posting comments, voting, and editing proposals. The “Proposal” category captures all work on proposals including text edits, updating model runs, and managing teams. Data is binned weekly, and the y-axis is capped at 140 total actions in order to see activity detail (activity spikes to >2000 actions during the third contest cycle) (Color figure online)

3.1 CoLab contest results

Table 1 includes a summary of the major contest cycles to date. The first contest began with the official launch of the Climate CoLab in November 2009. This contest had a single phase and there was no expert involvement. By the end of the contest, 162 users had registered on the site, and 58 had voted on a plan. The plan receiving the most votes specified emissions reductions in all regions of the world by 99 % by the year 2050. All the experts we consulted with said that it would not be feasible to make these reductions (for a range of social, economic and political reasons), and yet the majority of CoLab voters selected it.
Table 1

Summary of major contest cycles to date

Dates

Contests

Phase I proposals

Finalists

Votes cast

11/1/2009–12/9/2009

1

20

n.a.

58

10/1/2010–11/26/2010

1

29

4

403

5/16/2011–1/15/2011

2

84

12

2100

8/05/2012–

6

26

Because of this result, we introduced the phased contest cycle described above. Proposals generated in subsequent contests have in general been substantially more comprehensive and of higher quality than those in the first contest. For example, in the first contest, the popular choice plan contained no information about how the proposed emissions commitments might be achieved. In contrast, the popular choice plan from the second contest contained ten times as many words and provided both rationale and a range of suggested actions for implementing the plan.

Generally speaking, winning proposals have been diverse, creative, serious contributions from people all over the world. A complete analysis is not possible here, but a summary of proposal finalists from the 2011 contests (Table 2) offers a flavor of the kinds of solutions the CoLab community has generated.
Table 2

Proposal finalists from the 2011 Contest Cycle. The two contests focused on the economy at the global and national levels, respectively. The “proposal pitch” is a brief description provided by authors to describe their proposals

Proposal title

Proposal pitch

Author description

Awards

Global Proposals (Finalists)

2010 Winners combined

No pitch provided

Collaboration of winners from prior year

Popular Choice

The Planet or your Plate

Mitigate climate change by a rapid reduction of the short lived warming gases by advocating for less meat consumption globally

1 activist (US), 1 professor (US), 1 scientist (Australia)

Popular Choice/Judges’ Commendation

Rewire Plus: Behavior change and value change for the emerging green economy

A shift to a green economy will require changes in behaviors and values all the way down to the individual. Here’s how we get started!

Sustainability director, Univ. of Toronto; Recent graduate Univ. Toronto; CEO of Safara Sustainability Solutions

 

National Proposals (Finalists)

Cycling Carbon

A pragmatic, ambitious 7-point plan to renew the U.S. economy, enrich its citizens, and lead the world in fixing the climate

Software engineer from North Carolina

Popular Choice

Dream for a Green Future

India and dream for a green future : It’s time to make it a reality...

2 University Students at TERI University, India

Popular Choice

Personal Rapid Transit grids

Install connected Personal Rapid Transit grids over the urban and suburban areas that house the densest 50 % of the US population

Research scientist (MIT)

Judges’ Choice

Climate proofing the economics of socially sustainable small-scale agricultural systems

This proposal climate proofs the economics of sustainable agriculture in Africa

3 employees of the National Center for Technology, Nigeria/ 1 Professor, Obafemi Awolowo University, Nigeria

Judges’ Choice

How to Change US Energy in One Growing Season

How to Change US Energy in One Growing Season through practical demonstrations of available technologies and public education in all media

Independent solar activist in the Boston area

 

4 Conclusion

The Climate CoLab is a novel collective intelligence system designed to help thousands of people around the world work together to solve the problem of global climate change. It has attracted a continuing stream of visitors and members from around the world. As the platform and its community have evolved, members have generated proposals of substantial novelty and increasing quality. The community is now beginning to collaborate on the site outside of our annual contest cycles. These results are evidence that we are on our way to overcoming a key hurdle in the creation of a collective intelligence platform—the formation of a large and diverse community collectively engaged in solving a single problem.

The community members who write proposals and engage in discussions on the site are by far the most visible contributors to the Climate CoLab, and are a critical component of the system. But the full scope of human intelligence woven into the CoLab is far greater. Each of the design solutions described here is heavily dependent upon other human systems. The domain taxonomy codifies structures that the climate change community has generated in its effort to grapple with climate change. Radically open modeling makes it possible for end-users to leverage the expertise of many different model developers, and we rely upon a vibrant modeling and simulation community to provide the embedded technology. Expert-mediated voting is critically dependent upon the energy and willingness of experts who volunteer their time.

We believe the Climate CoLab is representative of a general approach to melding human intelligence and social technology to solve wicked social problems. It is a socio-technical system writ large, that leverages not only the intelligence of thousands of community members, but also the knowledge and capabilities of many pre-existing human systems. The platform itself is merely a nexus in which we hope our vast potential collective intelligence may be applied to solve the problem of climate change.

Acknowledgements

We would especially like to thank John Sterman and Hal Abelson of MIT for their support and participation in many phases of this project. We would also like to thank the following for financial support of this project: the National Science Foundation, BT plc, Cisco Systems, Argosy Foundation, the MIT Energy Initiative, and the MIT Sloan Sustainability Initiative. In addition, we are grateful to Stuart Scott and the members of the Climate Summit for their participation in the early outreach efforts, Mark Klein for his advice in developing an interface to enable on-line debates, Janusz Parfienuik and TopCoder, Inc. for their development work, and our experts, moderators, advisors and other community members for volunteering their time to this project.

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Joshua Introne
    • 1
  • Robert Laubacher
    • 1
  • Gary Olson
    • 2
  • Thomas Malone
    • 1
  1. 1.Center for Collective IntelligenceMassachusetts Institute of TechnologyCambridgeUSA
  2. 2.Donald Bren School of Information and Computer ScienceUniversity of California at IrvineIrvineUSA

Personalised recommendations