Carnegie Mellon University

Center for Informed Democracy & Social - cybersecurity (IDeaS)

CMU's center for the study of disinformation, hate speech and extremism online

IDeaS Center for Informed Democracy & Social-cybersecurity


February 16, 2024

Cyborgs for Strategic Communications on Social Media

By Lynnette Ng

Image by author using generative AI.

Tags: Cyborgs, bot detection, strategic communications, activism, social media

Paper Citation: Ng, L. H. X., Robertson, D. C., & Carley, K. M. (2024). Cyborgs for strategic communication on social media. Big Data & Society, 11(1). 

Direct link to paper, published online February 14, 2024

What is Strategic Communication:

Strategic Communication, or StratCom, is the deliberate act of information creation and
distribution. Social media is an ideal conduit for StratCom as it allows for quick and vast
information dissemination. Public personas and organizations use StratCom techniques as part
of their social media strategy to establish brand image and build consensus on important issues
Reference: Countries
employ StratCom techniques in social media during the Russia-Ukraine conflict to build and
communicate desired narratives. Reference:

What are Cyborgs?
A Cyborg is the hybrid between a bot and a human. Cyborgs are social media accounts that are
not always bots, yet not always people. Therefore, they are detected as bots in one time frame
but identified as humans in another. They present this outfit because they are controlled by
human operators in one timeframe, yet controlled by automated scripts in other instances.

How do we Identify Cyborgs?
Cyborgs are defined as social media users that are detected as bots in one timeframe and
humans in another. Many past work use machine learning techniques to identify Cyborgs
through machine-identified differentiation of features from manually annotated accounts. Reference: and However, these methods do not
exploit the bot-human duality of cyborgs. We exploit this duality in our identification technique.

We studied over 3.1 million users from two large events where bots are known to operate: 2020
Coronavirus pandemic and 2020 US Elections. We compare changes in bot likelihood scores
across a period of time, and use the values where the proportion of users that change
classification tapers off as quantitative threshold for identifying Cyborgs. This is about the 75th
percentile of overall distribution, which is about a minimum of 3 flips.

Who uses Cyborgs?
Having identified the Cyborg users, we extract them and analyze the accounts. We find the following key observations:
  - Cyborgs are present in all topic sets, suggesting that they are used by all sides of a
    conversation for StratCom
  - Cyborg users are more centrally placed within a network, therefore they have a
    larger influence
  - Cyborgs have long online lives, indicating their ability to evade bot detectors, or the
    graciousness of the social media platforms to allow their operations.
  - From a manual annotation of a random subset of accounts, we find that Cyborgs are
    mostly activist accounts (36%) or renowned personalities (27%). That is, these users
    deliberately make use of partial automation to disseminate their message.

Three things about Cyborgs
  1. A Cyborg present itself as half-bot, half-human. That is, it is detected as a bot in one
     time window, and a human in another. Cyborgs could be a human using automation
     to aid them, or a bot using humans to provide personal touches
  2. A large number of cyborgs being used for strategic communication purposes, by
     activists and key figures to establish their brand and amplify messages.
  3. A quantitative definition of Cyborgs would be a user that changes bot classification
     more than 3 times, with each change being more than 0.10 difference in bot
     likelihood score.