Chapter 11 • Group Decision Making, Collaborative Systems, and AI Support 639
Swarm Intelligence and Swarm AI
The term swarm intelligence refers to the collective behavior of decentralized, self- organized systems, natural or artificial (per Wikipedia). Such systems consist of things (e.g., ants, people) interacting with each other and their environment. A swarm’s actions are not centrally controlled, but they lead to intelligent behavior. In nature, there are many examples (e.g., ant colonies, fish schools) of such behaviors.
Natural groups were observed to amplify their group intelligence by forming swarms. Social creatures, including people, can improve the performance of their individual mem- bers when working together as a unified system. In contrast with animals and other species whose interactions among group members are natural, people need technology to exhibit swarm intelligence. This concept is used in studies and implementation of AI and robotics. The major applications are in the area of predictions.
Example
A study at Oxford University (United Kingdom) involved predicting the results of all 50 English Premier League soccer games over five weeks. A group of independent judges scored 55 percent accuracy when working alone. However, when predicting using an AI swarm, their prediction success increased to 72 percent (an improvement of 31 percent). Similar improvement was recorded in several other studies.
In addition to improved prediction accuracy, studies show that using swarm AI results in more ethical decisions than that of individuals (Reese, 2016).
SWARM AI TECHNOLOGY Swarm AI (or AI swarm) provides the algorithms for the inter- connections among people creating the human swarm. These connections enable the knowledge, intuition, experience, and wisdom of individuals to merge into single improved swarm intelligence. Results of swarm intelligence can be seen in the TED presentation (15:58 min.) at youtube.com/watch?v=Eu-RyZt_Uas. Swarm AI is used by several third- party companies (e.g., Unanimous.aI, as illustrated in Application Case 11.3.
XPRIZE is a nonprofit organization that allocates prizes via competitions to promote innovations that have the potential to change the world for the bet- ter. The main channel for designing prizes that solve humanity’s grandest challenges is called Visioneer- ing. It attempts to harness the power of the global crowd to develop solutions to important challenges. The organization’s major event is an annual summit meeting where prizes are designed and proposals are evaluated. The experts at XPRIZE develop concepts and turn them into incentivized competitions. Prizes are donated by leading corporations.
For example, in 2018, IBM Watson donated a $5 million prize called “AI approaches and collabora- tion.” The competition had 142 registered teams, and 62 were left in round 2 in June 2018. The teams are
invited to create their own goals and solutions to a grand challenge.
The Problem
Every year, there is a meeting of 250 members of “Visioneers Summit Ideation” where top experts (entrepreneurs, politicians, scientists, etc.), partici- pate to discover and prioritize topics for the XPRIZE agenda.
Finding the top global problems can be a very complex challenge due to a large number of vari- ables. In just a few days, top experts need to use their collective wisdom to agree on the next year’s XPRIZE top challenges. The method used to support the group’s decision is a critical success factor.
(Continued )
Application Case 11.3 XPRIZE Optimizes Visioneering
640 Part IV • Robotics, Social Networks, AI and IoT
The Solution
In the 2017 annual meeting for determining what challenge to use for 2018, the organization used the swarm AI platform (from Unanimous AI). Several small groups (swarms) moderated by AI algorithms were created to discover challenging topics. The mis- sion was to explore ideas and agree on preferred solutions. The objective was to use the talents and brainpower of the participants.
In other words, the objective was to use the thinking together feature of swarm AI to generate each group’s synergy with the AI algorithms acting as moderators. This way, smarter decisions were generated by the groups than its individual par- ticipants. The different groups examined six pre- selected topics: energy and infrastructure, learning human potential, space and new frontiers, plant and environment, civil society, and health and well-being. The groups brainstormed the issues. Then, each participant created a customized evalu- ation table. The tables were combined and ana- lyzed by algorithms.
Application Case 11.3 (Continued)
The Swarm AI replaced traditional voting meth- ods by optimizing the detailed contribution of each participant.
The Results
Use of swarm AI did the following:
• Supported the generation of optimized answers and enabled fast buy-in from the participants.
• Enabled all participants to contribute. • Provided a better voting system than in previ-
ous years.
Questions for Case 11.3 1. Why is the group discussion in this case complex?
2. Why is getting a consensus when top experts are involved more difficult than when non-experts are involved?
3. What was the contribution of swarm AI?
4. Compare simple voting to swarm AI voting.
Sources: Compiled from Unanimous AI (2018), xprize.org, and xprize.org/about.
SWARM AI FOR PREDICTIONS Swarm AI was used by Unanimous AI for making predic- tions in difficult-to-assess situations. Examples are:
• Predicting Super Bowl #52 number of points scored (used for spread waging). • Predicting winners in the regular NFL season. • Predicting the top four finishers of the 2017 Kentucky Derby. • Predicting the top recipients of the Oscars in 2018.
u SECTION 11.8 REVIEW QUESTIONS
1. Relate the use of AI to the activities in Figure 11.1. 2. Discuss the different ways that AI can facilitate group collaboration. 3. How can AI support group evaluation of ideas? 4. How can AI facilitate idea generation? 5. What is the analogy of swarm AI to swarms of living species? 6. How is swarm AI used to improve group work and to initiate group predictions?
11.9 HUMAN–MACHINE COLLABORATION AND TEAMS OF ROBOTS
Since the beginning of the Industrial Revolution, people and machines have worked together. Until the late 1900s, the collaboration was in manufacturing. But since then, due to advanced technology and changes in the nature of work, human–machine collabora- tion has spread to many other areas, including performing mental and cognitive work
