Design Patterns of Social Bots

July 2, 2020

Our 2nd year PhD student Abdullah Alrhmoun presented his research project about the Design Patterns of Social Bots. Design patterns are a toolkit of tried and tested solutions to recurring (repetitive) design problems. Nowadays, the ecosystems of online bots are drastically increasing in number and complexity. An increasing amount of websites and online social networks are becoming the perfect environment for bots, operating dependently or not from users’ inputs. 

According to Tony Veale and Mike Cook (2018), we can find four main categories of Social Bots:

  1. Feed Bots created to stream out contents;
  2. Watcher Bots created to collect personal user data;
  3. Interactor Bots created to interact with users;
  4. Learner Bots created to learn from the interactions with users.

More specifically, there is a subset of Social Bots, called Politi Bots used to influence political opinions in online social networks. The main categories are the following:

  1. Sockpuppets, which initiate conversations, seed new ideas and drive discussion in online communities;
  2. Amplifier Bots, automated accounts which repurpose, retweet, and republish the same content;
  3. Approval Bots, which engage with specific tweets or comments to enhance the credibility and give legitimacy to an idea.

The two main research questions are the following:

  1. How to maximize the embeddedness of the bots in the network?
  2. Which bot design patterns are effective in the ecosystem? 

The proposed research approach to design an effective design pattern will investigate the following areas: philosophical, conceptual, practical (i.e. event-driven design) and objective-driven. In particular, the event-driven design will allow for a creation of an ecosystem where it would be possible to reproduce collective actions. In fact, relying on the automation, synchronisation, reward maximisation, and network effect it will be possible to analyse collective behaviours – i.e. bot-nets. The creation of bot-nets interacting to each other can then be used to reach specific trend or influence targets. 

Source Images:

Blog post by Teodoro Criscione