Collaborative intelligence

Collaborative intelligence characterizes multi-agent, distributed systems where each agent, human or machine, is uniquely positioned, with autonomy to contribute to a problem-solving network. Collaborative autonomy of organisms in their ecosystems makes evolution possible. Natural ecosystems, where each organism's unique signature is derived from its genetics, circumstances, behavior and position in its ecosystem, offer principles for design of next generation social networks to support collaborative intelligence, crowd-sourcing individual expertise, preferences, and unique contributions in a problem-solving process.^[1]

Overview

Collaborative intelligence is a term used in several disciplines. In business it describes heterogeneous networks of people interacting to produce intelligent outcomes. It can also denote non-anonymous multi-agent problem-solving systems. The term was used in 1999 to describe the behavior of an intelligent business ecosystem^[2] where Collaborative Intelligence, or CQ, is "the ability to build, contribute to and manage power found in networks of people."^[3] When the computer science community adopted the term collective intelligence and gave that term a specific technical denotation, a complementary term was needed to distinguish between anonymous homogeneity in collective prediction systems and non-anonymous heterogeneity in collaborative problem-solving systems. Anonymous collective intelligence was then complemented by collaborative intelligence, which acknowledged identity, viewing social networks as the foundation for next generation problem-solving ecosystems, modeled on evolutionary adaptation in nature's ecosystems.

History

Collaborative intelligence traces its roots to the Pandemonium Architecture proposed by artificial intelligence pioneer Oliver Selfridge as a paradigm for learning.^[4] His concept was a precursor for the blackboard system where an opportunistic solution space, or blackboard, draws from a range of partitioned knowledge sources, as multiple players assemble a jigsaw puzzle, each contributing a piece. Rodney Brooks notes that the blackboard model specifies how knowledge is posted to a blackboard for general sharing, but not how knowledge is retrieved, typically hiding from the consumer of knowledge who originally produced which knowledge,^[5] so it would not qualify as a collaborative intelligence system.

In the late 1980s, Eshel Ben-Jacob began to study bacterial self-organization, believing that bacteria hold the key to understanding larger biological systems. He developed new pattern-forming bacteria species, Paenibacillus vortex and Paenibacillus dendritiformis, and became a pioneer in the study of social behaviors of bacteria. P. dendritiformis manifests an intriguing collective faculty, which could be viewed as a precursor of collaborative intelligence, the ability to switch between different morphotypes to better adapt with the environment.^[6][7] Ants were first characterized by entomologist W. M. Wheeler as cells of a single "superorganism" where seemingly independent individuals can cooperate so closely as to become indistinguishable from a single organism.^[8] Later research characterized some insect colonies as instances of collective intelligence. The concept of ant colony optimization algorithms, introduced by Marco Dorigo, became a dominant theory of evolutionary computation. The mechanisms of evolution through which species adapt toward increased functional effectiveness in their ecosystems are the foundation for principles of collaborative intelligence.

Crowdsourcing evolved from anonymous collective intelligence and is evolving toward credited, open source, collaborative intelligence applications that harness social networks. Evolutionary biologist Ernst Mayr noted that competition among individuals would not contribute to species evolution if individuals were typologically identical. Individual differences are a prerequisite for evolution.^[9] This evolutionary principle corresponds to the principle of collaborative autonomy in collaborative intelligence, which is a prerequisite for next generation platforms for crowd-sourcing. Following are examples of crowdsourced experiments with attributes of collaborative intelligence:

• SwarmSketch is a crowd-sourced art experiment.
• Galaxy Zoo is a citizen science project led by Chris Lintott at Oxford University to tap human pattern recognition capacities to catalog galaxies.
• DARPA Network Challenge explores how the Internet and social networking can support timely communication, wide-area team-building, and urgent mobilization to solve broad-scope, time-critical problems.
• Climate CoLab, spun out of MIT and its Center for Collective Intelligence.
• reCAPTCHA is a project to digitize books, one word at a time

As crowdsourcing evolves from basic pattern recognition tasks to toward collaborative intelligence, tapping the unique expertise of individual contributors in social networks, constraints guide evolution toward increased functional effectiveness, co-evolving with systems to tag, credit, time-stamp, and sort content.^[10] Collaborative intelligence requires capacity for effective search, discovery, integration, visualization, and frameworks to support collaborative problem-solving.^[11]

Contrast with collective intelligence

The term collective intelligence originally encompassed both collective and collaborative intelligence, and many systems manifest attributes of both. Pierre Lévy coined the term "collective intelligence" in his book of that title, first published in French in 1994.^[12] Lévy defined "collective intelligence" to encompass both collective and collaborative intelligence: "a form of universally distributed intelligence, constantly enhanced, coordinated in real time, and in the effective mobilization of skills".^[13] Following publication of Lévy's book, computer scientists adopted the term collective intelligence to denote an application within the more general area to which this term now applies in computer science. Specifically, an application that processes input from a large number of discrete responders to specific, generally quantitative, questions (e.g. what will the price of DRAM be next year?) Algorithms homogenize input, maintaining the traditional anonymity of survey responders to generate better-than-average predictions.

Recent dependency network studies suggest links between collective and collaborative intelligence. Partial correlation-based Dependency Networks, a new class of correlation-based networks, have been shown to uncover hidden relationships between the nodes of the network. Research by Dror Y. Kenett and his Ph.D. supervisor Eshel Ben-Jacob uncovered hidden information about the underlying structure of the U.S. stock market that was not present in the standard correlation networks, and published their findings in 2011.^[14]

Application

Collaborative intelligence addresses problems where individual expertise, potentially conflicting priorities of stakeholders, and different interpretations of diverse experts are critical for problem-solving. Potential future applications include:

• competitions, where submissions must be integrated to produce a synergistic outcome;
• smart search, where social networks of searchers on related topics co-define search results;
• professional groups, interest collectives, citizen science and other communities, where knowledge-sharing is a prerequisite for effective outcomes;
• planning, development, and sustainable project management;
• smart systems to transform independent cities into collaborative, ecological urban networks

Wikipedia, one of the most popular websites on the Internet, is an exemplar of an innovation network manifesting distributed collaborative intelligence that illustrates principles for experimental business laboratories and start-up accelerators.^[15]

A new generation of tools to support collaborative intelligence is poised to evolve from crowdsourcing platforms, recommender systems, and evolutionary computation.^[11] Existing tools to facilitate group problem-solving include collaborative groupware, such as GitHub, Google+, Confluence, eXo Platform, JIRA, Skype, NetMeeting, WebEx, and synchronous conferencing technologies such as instant messaging, online chat and shared white boards, which are complemented by asynchronous messaging like electronic mail, threaded, moderated discussion forums, web logs, and group Wikis. Managing the Intelligent Enterprise relies on these tools, as well as methods for group member interaction; promotion of creative thinking; group membership feedback; quality control and peer review; and a documented group memory or knowledge base.^[16] As groups work together, they develop a shared memory, which is accessible through the collaborative artifacts created by the group, including meeting minutes, transcripts from threaded discussions, and drawings. The shared memory (group memory) is also accessible through the memories of group members; current interest focuses on how technology can support and augment the effectiveness of shared past memory and capacity for future problem-solving. Metaknowledge characterizes how knowledge content interacts with its knowledge context in cross-disciplinary, multi-institutional, or global distributed collaboration.^[17]

See also

• Collaborative innovation network
• Global brain
• Mass collaboration
• Mass communication

References