Identifying incompleteness in privacy policy goals using semantic frames

Abstract

Companies that collect personal information online often maintain privacy policies that are required to accurately reflect their data practices and privacy goals. To be comprehensive and flexible for future practices, policies contain ambiguity that summarizes practices over multiple types of products and business contexts. Ambiguity in data practice descriptions undermines policies as an effective way to communicate system design choices to users and as a reliable regulatory mechanism. In this paper, we report an investigation to identify incompleteness by representing data practice descriptions as semantic frames. The approach is a grounded analysis to discover which semantic roles corresponding to a data action are needed to construct complete data practice descriptions. Our results include 698 data action instances obtained from 949 manually annotated statements across 15 privacy policies and three domains: health, news and shopping. Therein, we identified 2316 instances of 17 types of semantic roles and found that the distribution of semantic roles across the three domains was similar. Incomplete data practice descriptions undermine user comprehension and can affect the user’s perceived privacy risk, which we measure using factorial vignette surveys. We observed that user risk perception decreases when two roles are present in a statement: the condition under which a data action is performed, and the purpose for which the user’s information is used.