23 Apr 2020


Traders deciding on the next big market bet. A navigation app quickly mapping out a less-explored area. Fashion brands choosing the hottest color of the season. An airport managing flight delays.

What do these scenarios have in common? In each one, swarm intelligence blends global and local insight to improve how businesses make decisions.

Swarm intelligence is a form of artificial intelligence (AI) inspired by the insect kingdom. In nature, it describes how honeybees migrate, how ants form perfect trails, and how birds flock. In the world of AI, swarm systems draw input from individual people or machine sensors and then use algorithms to optimize the overall performance of the group or system in real time.

Consider Waze, the popular road navigation app that uses swarm intelligence to create and modify maps. Starting with limited digital maps, it began making tweaks based on its users’ GPS data along with manual map modifications by registered users. Entire cities have been mapped using this method, as was the case in Costa Rica’s capital, San José. And just as ants signal danger to their counterparts, so too do Waze users contribute live information from accident locations and traffic jams.

Swarm intelligence is now being used to predict everything from the outcome of the Super Bowl to fashion trends to major political events. Using swarm intelligence, investors can better predict market movements, and retailers can more accurately forecast sales.

While the swarm intelligence concept isn’t new, the advent of edge computing has renewed its impetus. This technology enables greater processing and data storage on local devices instead of big data centers or the cloud. Advances in internet of things (IoT) technologies, machine learning, and 5G also make swarm systems faster and more efficient.

In a world of increasing flux, scale, and complexity, swarm intelligence will help businesses in two main ways: finding new sources of growth, and anticipating and managing disruption.

Following the Ant Trail to Growth

Ants have a very particular approach to finding a trail to food: Constantly releasing pheromones, they signal their progress to the rest of the collective. Each ant learns from all the other ants’ experiences, and as a result, each gets closer to a food source. Eventually, the colony identifies the best trail based on the feedback of individual ants.

This approach presents a valuable lesson for businesses looking to identify new growth opportunities. Finance is one industry where spotting new growth opportunities ahead of the rest of the market is crucial. While algorithms can forecast market trends, investment decisions are made in boardrooms, where overpowering personalities and corporate hierarchies can preclude investors from identifying or pursuing the right opportunities.

To mitigate such effects, large financial institutions, such as hedge funds and investment banks, have employed Unanimous AI’s Swarm platform to predict potential market growth areas. Here’s how it works: A team of finance experts uses the platform to answer a series of questions about their perceptions of market trends. The key is to observe how other participants, represented by anonymous dots on the screen, answer the questions in real time. Instead of picking the answer based on a simple majority vote, the experts converge on the asset class or market that they believe is likely to perform best in the next year.

Louis Rosenberg, CEO of Unanimous AI, explained, “People tell their superiors what they want to hear. In a swarm, where everybody is equal and anonymous, you get a much more accurate combination of insights.” The same approach can be used for sales forecasting in the retail sector. Instead of a survey where individual consumers rate sweaters from 1 to 10, the system enables a focus group to act as a collective and to agree on the single best sweater. Just as investors observe how their colleagues evaluate different assets, consumers can see how much other participants prefer a particular sweater. Sales forecasts using this technology are highly correlated to actual unit sales, successfully predicting 34% of the variance in sales; in contrast, the standard consumer survey predicted a mere 4%.

Managing Disruption From the Ground Up

In the insect kingdom, collective behavior is key to responding to uncertain environments. For midges, it provides stability and robustness against environmental disruption. In a similar fashion, swarm intelligence helps organizations manage disruptive events through the power of collective data.

Gatwick, one of the world’s busiest airports, is a case in point. In a bid to improve efficiency against the backdrop of growing airport demand and capacity constraints, its management looked to swarm intelligence. Abhi Chacko, head of IT Commercial & Innovation at Gatwick, said, “Anything that can be done to reduce flight disruption makes a big difference to overall performance.”

Previously, Gatwick relied on rule-of-thumb flight estimates of departure times. Unsurprisingly, local factors often proved these wrong — a small change of weather here, staff shortages there.

Working with ConvergentAI, Gatwick came up with a swarm-based system called FlightID, which collates all ground data on factors that typically affect departure times. Tom Brock, CEO of ConvergentAI, calls this system decentralized “small data” analytics. Rather than looking at the average flight delay across the airport, the swarm algorithms learn about each individual flight’s performance, down to the gritty details of how a specific airline tends to perform on rainy days.

“We wanted the algorithm to account for various factors,” explained Gatwick’s Chacko. “The number of passengers, for example — busier flights on larger planes take longer to depart. A 6 a.m. flight differs from an 8 p.m. one. Business travelers behave differently than families going on holidays.”

This nuanced, yet more holistic, view enabled Gatwick to better anticipate and respond to the impact of ever-changing conditions. As a result, its performance improved, helping an additional 1.5 million passengers depart on time in 2018.

Success Factors of a Swarm Organization

Just like their natural-world counterparts, AI swarms need certain conditions to operate successfully: specific goals, the right resources, and a supportive infrastructure, to name a few. For successful swarming:

  • Focus on a key operational goal. A swarm of ants can locate food over a massive area because they are individually focused on a single goal. For an organization, the goal is likely to be a significant area of business performance. Guided by a clear goal of improving flight turnaround time, Gatwick was able to optimize its performance within limited capacity.
  • Deploy floating resources to areas of greatest need. Worker bees can assume different roles depending on the needs of the hive: collecting pollen, feeding larvae, or making wax. Similarly, swarm systems can help organizations respond flexibly to spikes or dips in demand. In the case of a buildup of passengers, for example, a swarm system can help airports identify where and when to open up new passenger channels or security lanes.
  • Integrate into the broader ecosystem. Beehives and ant colonies do not exist in siloes. They share their knowledge with the rest of the swarm so it can quickly adapt and respond at scale. Similarly, an organization implementing swarm solutions has to make sure that it benefits from the different parts of the organization and its ecosystem. For example, in an airport, retail brands, airlines, and government officers all need to operationally support the swarm system.
  • Keep it simple. While swarm technology can be complex, individual agents act best on simple instructions. A swarm system has to be simple to understand and use. Whether it’s choosing the best sweater or predicting flight delays, each member of the swarm needs a simple and clear blueprint to act upon.
  • Keep it human. Swarm AI should be seen as an aid to worker productivity. As Chacko said, “We want the system to enhance human judgment, not replace it.” An algorithm might tell us that on Friday evenings, an airport is likely to be crowded with revelers with more potential for disruption, but it is humans who can come up with the creative solutions to manage such risks.

Adapting to change while maintaining a global scale is a relatively new problem for business, but one that the animal kingdom has been mastering for millions of years. What can your organization learn from swarm intelligence?


Read the full article here.
This content was originally published by MIT Sloan Management Review. Original publishers retain all rights. It appears here for a limited time before automated archiving. By MIT Sloan Management Review

Covid-19 – Johns Hopkins University

Download brochure

Introduction brochure

What we do, case studies and profiles of some of our amazing team.

Download