Why Swarm Intelligence-Based decisions are better
Swarm intelligence (SI) “is the collective behavior of decentralized, self-organized systems, natural or artificial.” It’s a key component of artificial intelligence developments, proven especially effective in critical business decisions.
Individual ants, bees, and termites are hapless creatures, but brilliant collectively at using scent paths and other complex signaling to collectively locate the best food sources, switching roles for a better division of labor, and creating/maintaining magnificent nest structures. We humans are also social animals, but comparatively ineffective in groups. In fact, collectively we often exhibit poor decision-making and destructive spontaneous behavior– as exemplified by mob rioting. We also easily buy into and act upon all kinds of disinformation.
The main problem is we’re lousy at making real-time predictions of likely outcomes. When polled, we usually make decisions based on highly subjective confirmation bias derived from flimsy, first-impression assumptions. BTW, if interested in knowing more about confirmation bias, check out my previous posts on psychological bias.
Of course, most folks are unaware of these downsides to conventional human decision making. Besides, few things scare us as much as sci-fi scenarios of ‘human hive intelligence,’ e.g., Invasion of the Body Snatchers and Star Trek’s Borg. –Fortunately, radical improvements in Big Data analysis are replicating many of the benefits of hive intelligence without our having to lose any sense of our human identity or autonomy.
Recent applications of SI include—
- Data mining– The UNU collective platform uses SI technologies in real time to combine the thoughts and feelings of groups to answer questions and make predictions. Amazingly, testing to date shows that ‘human swarms‘ easily out-predict individuals in predicting the outcomes of things like election outcomes (better, e.g., than 2016 polling). Another example– in predicting Academy Award winners, swarm feedback improved the accuracy last year of who would win from 40% for individuals to 70% for the group. The process requires real-time communication among participants that help expand the base of available information before the final outcome.
- Warfare– A new generation of SI-driven drones that are capable of making independent targeting decisions, including facial recognition allows for the killing of enemy combatants. This is a deeply disturbing scenario, and it’s all but certain that some combatant will begin using this technology in the future without human oversight.
- Advanced Tech Manufacturing/Assembly– The European Space Agency is developing swarm technology for the assembly of satellites in space thinking about an orbital swarm for self-assembly with magnetic wave communication between components.
Next, I’ll discuss how business is already benefitting from SI with exciting potential for future advances that can help you grow your business.
Applying SI (Swarm Intelligence) to business decision making
So far, I made the case that Swarm Intelligence (SI) decision-making strategies are superior to long-established conventional approaches. As applied to humans, swarm SI is used to track predictions in real time for best outcomes among a range of choices.
Three principles have been cited to explain SI’s success in business decision making–
- Quick adaptation to changing environments;
- Work role interchangeability—i.e., when an individual can’t complete a task, another quickly takes over; and
- Self-organization, i.e., activity is governed by group consensus, not hierarchical command.
It’s the self-organization principle that executives resist the most. What they overlook is how complex and unpredictable human behavior can be, even when based on a few simple rules. Most importantly, without self-organization, the other two principles of SI don’t work.
SI’s recent record of private sector success is impressive—
- Southwest Airlines was experiencing bottlenecks in its cargo routing system a few years ago. Their previous solution had been the intuitive one of loading freight in the first plane going in the right direction. However, when they used SI analysis to the problem, they discovered that sending cargo in the wrong direction sometimes worked best. Hard to believe? You can’t argue with their results—an 80% improvement in freight transfer rates at their busiest airports and decreased workloads for cargo/baggage handlers by 20%. This new strategy also reduced the need for storage facilities and helped reduce employee wage costs.
- SI analysis has yielded similar results for a wide range of industries with improved scheduling and division of labor in book publishing, telecom, manufacturing, and credit card organizations. For example, Hewlett-Packard uses ‘virtual software agents’ that roam its telecom calling routes to immediately locate congested networks, automatically redirecting them to those with low traffic. Because variances of network traffic volume are difficult to predict, this solves a big problem: it not only accelerates the speed of calls but also ‘decongests’ busy networks.
- Factory efficiency has been improved using variations of the ant-foraging algorithm. Variations on an ant-foraging algorithm have created faster, automatic ways to deploy equipment within required parameters to allocate resources to jobs with the result that all priorities and schedules are met, efficiently adapting to breakdowns. Some critics have questioned the applicability of insect behavior to human interaction. So far, the evidence doesn’t support their concerns.
- Warehousing efficiency has been vastly improved by several companies that have junked the ‘zone approach.’ The zone approach is the process whereby one employee is responsible for processing specific book order or other product categories. It results in an inefficient division of labor because it under uses fast employees while stressing out slower ones with high volume. With the zone approach, even if all zone-assigned employees worked at the same speed, some employees get more orders to process because of fluctuating demand for some book categories over others. The solution, after a careful SI computer simulation—station the slowest workers at the start of the line for ALL products and the fastest at the end. In one warehouse, this improved productivity by 30%.
- SI principles have been applied in the IT industry. In one case, introducing SI strategies resulted in a radical reduction in employee attrition rate (consistent with increased work satisfaction) to 4%, compared with 20% for the industry-at-large.
Bottom line— when organizations carefully plan an SI-based reorganization, employees are encouraged to develop their own ideas free of direct management. Even recruitment and allocation of other resources then transfer to those who have the best ideas, as determined by spontaneous, collaborative group consensus.
Would this work for all organizations? Certainly not. However, those organizations that move in this direction, with careful metrics to monitor the progress of different SI strategies have a distinct advantage over their competition. SI strategies create better internal efficiency and much faster adaptation to rapidly-changing external market forces.