Sign In

Communications of the ACM

ACM TechNews

Bots Talk Like Humans but Their Cloned Personalities Give Them Away

A friendly chatbot.

The study found that individually, the bots appeared human, while collectively, the social bots appear to be clones of one another, in terms of their estimated values across all 17 language-based human attributes.


Stony Brook University and University of Pennsylvania researchers explored social spambots' behavior on social media, interaction with genuine accounts, and current bot technologies' capabilities.

The researchers reviewed over 3 million tweets generated by 3,000 bot accounts and an equal number of authentic accounts, with 17 language-based human attributes (including emotions, age, gender, and sentiment) extrapolated.

The bots resembled humans individually, but looked like clones of each other cumulatively, in terms of estimated values across the 17 attributes.

The language they used generally resembled that of a person in their late 20s and with a positive tone.

Automatically clustering the accounts into two groups based on these 17 traits and no bot labels demonstrated that one group was almost entirely bots, which led to the creation of an accurate unsupervised bot detector.

From Stony Brook University News
View Full Article


Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA


No entries found