Quantify the automation–human alternation needed for cyborg classification

Determine the quantitative threshold for the alternation between automated scripting and human intervention in Twitter activity required to classify an account as a cyborg (hybrid human–automation account), distinguishing it from purely human-operated accounts and fully automated bots across measurement time windows.

Background

The paper defines social media cyborgs as accounts that sometimes appear bot-like and at other times human-like, due to intermittent switching between automated scripts and human curation. Prior work has described cyborgs as ambiguous because their mixed behavior complicates classification.

Within the literature review, the authors explicitly note the uncertainty regarding the amount of alternation necessary for an account to be considered a cyborg. While this study proposes an operational approach using bot classification flips and bot-likelihood score differences (e.g., ≥3 flips and ≥0.10 score change), the underlying conceptual question about the precise quantitative alternation that demarcates cyborgs from bots and humans is explicitly stated as uncertain.