Hiding opinions from machine learning

Area of Science:

• Social computing
• Artificial intelligence
• Human-computer interaction

Background:

• Machine learning and big data enable detailed analysis of online activities.
• Inference of private information and user opinions from online profiles is a growing concern.
• Twitter users are increasingly scrutinized for their opinions on various topics.

Purpose of the Study:

• To investigate the privacy threat posed by algorithms inferring Twitter user opinions.
• To explore methods for users to conceal their opinions from such algorithms.
• To develop AI-driven countermeasures for preserving user privacy.

Main Methods:

• Survey conducted to assess user awareness and desire for opinion privacy on Twitter.
• Analysis of user inability to identify opinion-revealing online activities.
• Development and testing of a heuristic-based AI assistant to guide profile modifications.

Main Results:

• Significant proportion of Twitter users wish to hide opinions on social, political, and religious issues.
• Users struggle to identify which online activities reveal their opinions.
• Proposed AI heuristic effectively helps users hide their opinions by suggesting profile modifications.

Conclusions:

• Machine learning analysis of online profiles presents significant privacy risks.
• AI-driven strategies can empower users to protect their opinion privacy.
• Developing countermeasures is crucial for safeguarding the right to control shared personal information.