Voice-activated personal assistants and privacy concerns: a Twitter analysis

Abstract

This study aims to understand the extent of privacy concerns regarding voice-activated personal assistants (VAPAs) on Twitter. It investigates three key areas: (1) the effect of privacy-related press coverage on public sentiment and discussion volume; (2) the comparative negativity of privacy-focused conversations versus general conversations; and (3) the specific privacy-related topics that arise most frequently and their impact on sentiment and discussion volume. Design/methodology/approach – A dataset of 441,427 tweets mentioning Amazon Alexa, Google Assistant, and Apple Siri from July 1, 2019 to June 30, 2021 were collected. Privacy-related press coverage has also been monitored. Sentiment analysis was conducted using the dictionary-based software LIWC and VADER, whereas text mining packages in R were used to identify privacy-related issues. Findings – Negative privacy-related news significantly increases both negativity and volume in Twitter conversations, whereas positive news only boosts volume. Privacy-related tweets were notably more negative than general tweets. Specific keywords were found to either increase or decrease the sentiment and discussion volume. Additionally, a temporal evolution in sentiment, with general attitudes toward VAPAs becoming more positive, but privacy-specific discussions becoming more negative was observed. Originality/value – This research augments the existing online privacy literature by employing text mining methodologies to gauge consumer sentiments regarding privacy concerns linked to VAPAs, a topic currently underexplored. Furthermore, this research uniquely integrates established theories from privacy calculus and social contract theory to deepen our analysis.