Communications of the ACM
Not Even AI Can Make Total Sense of a Privacy Policy
Researchers at Carnegie Mellon University researchers have created a artificial inteligence-based tool that crawls the Web's most popular sites to flag privacy policy issues of potential concern.
Credit: Pexels
Carnegie Mellon University researchers have created a tool that uses artificial intelligence (AI) to crawl 7,000 of the web's most popular sites to flag privacy policy issues of potential concern to specific users.
When searching for entire paragraphs, the AI was able to identify relevant passages with 79% accuracy, compared with 70% accuracy when looking for individual, relevant sentences.
Human experts parse privacy policies at a much more granular level. In parsing YouTube's privacy policy, the AI missed three-quarters of the third-party sharing references that the human caught.
The AI is improving, but its learning process take time because it requires multiple examples that have been analyzed by a human, the researchers say.
The team hopes to make the tool available as a custom browser plugin by the end of the year, and eventually to apply it to other online legalese, like terms of service.
From Motherboard
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found