Fetch.ai + Oxford Project

Jun 24, 2021

An individual bee isn’t very clever. It collects nectar, and fights when threatened. A bee hive is very different. It manages its resources, exploring its whole environment to take advantage of the best energy sources. It is able to strategically deter large threats like hornets, and to ensure survival of the germ line, often sacrificing individual insects. [1]

This pattern is found all across nature and society. Cells assemble into organisms. Buyers and sellers cooperate and create markets. Simple entities combine to make complex systems.

This pattern is called emergence [2]. It is emergent behaviour that we, a team of six Oxford students, set out to investigate during our group design practical. The project was sponsored by Fetch.ai, who besides providing invaluable support, developed the AEA framework, which our product is based on.

AEA means Autonomous Economic Agent, and the framework can be used to build applications for a multi-stakeholder, trust-minimised network of agents, which exchange information and make financial transactions with each other.

With our product, we wanted to showcase the framework, and to build a simulation in which we could observe emergence. The system being simulated was a society of scavengers, looking for sources of water to survive in a desert. Each scavenger is represented by an agent in the framework. Additionally, there is an environment agent, which represents the state of the world. The scavengers can explore their environment, harvest water, pass water around, and communicate with each other.

Visualisation of a simulation run, created with our software. The Blue circles are sources of water, the small dots are scavengers. The greener they are, the more water they have remaining.

We implemented different strategies for the scavengers. Some purely egoistic, and others involving larger degrees of cooperation.

We called our simplest strategy “lone goldfish”: a memoryless scavenger that looks for water sources on a random path, and drains them completely. The most advanced strategy was “explorer dogs.” This scavenger only drinks if its water level is below a certain threshold, and spends the remaining time exploring its surroundings. When it gets thirsty, it returns to one of its memorized water sources. Explorer dogs also communicate with each other, exchanging water locations when they meet. This way, even if a scavenger doesn’t have the resources to reach a water source, it can help another scavenger survive by sharing its knowledge.

As we had anticipated, the more cooperative strategy outperformed the simple, egoistic one. Populations of explorer dogs tended to survive longer, and to utilize the water resources more fully than populations of lone goldfish, which often left parts of the desert unexplored.

Summary statistics for a simulation run, created using our software.

 

Want to learn more? Watch our project summary video, or checkout the GitHub repository to play around with the code yourself!

 

[1] https://web.archive.org/web/20210517010837/https://en.wikipedia.org/wiki/Asian_giant_hornet#Native_honey_bees .

[2]
O’Connor, Timothy, “
Emergent Properties“, The Stanford Encyclopedia of Philosophy (Fall 2020 Edition), Edward N. Zalta (ed.).

 

Project developed by Blanche Duron, Kevin Xin, Tancrede Guillou, Olaf Czarnecki, Chun Chang and Ye Teng at Oxford, with help from David Minarsch and Ali Hosseini at fetch.ai. This blog post was written by Olaf Czarnecki.