Mobile App Privacy: Expectation and Purpose

Vocabulary terms and key concepts derived from a lecture on the paper 'Expectation and Understanding Users' Mental Models of Mobile APP Privacy Through Crowdsourcing' and the discussion of app permissions.

Quiz Parameters

A multiple-choice assessment on Canvas covering the first $7$ papers/readings up to the end of the week, designed as an open-book gauge for reading depth.

TaintDroid

A dynamic analysis tool used to run top $100$ Android apps to establish the ground truth of application behavior for comparison against user guesses.

Expectation condition

A study branch where participants were only told which sensitive resource (such as device ID, contacts, or location) an application was accessing.

Purpose condition

A study branch where participants were provided with both the specific resource being accessed and the reason (the why) for that access.

Privacy awareness

One of the three dimensions used to evaluate the new interface design, looking at whether users naturally mentioned privacy concerns during installation decisions.

Comprehensibility

An evaluation metric measured by a quiz to determine if users actually understood what data an application was taking, which improved from around 60\text{%} to over 90\text{%} with the new design.

Efficiency

A metric measured using JavaScript to track the exact time users spent reading and hovering over the privacy summary screen.

Privacy as expectations

A conceptual framework where privacy is defined not just by what information is collected, but by whether the application's behavior matches human mental models.

Expectation gap

The finding that users are frequently surprised by background data access; for example, fewer than 30\text{%} of users could guess that location access was used for targeted ads.

Crowdsourcing (Study 1)

The use of the Amazon Mechanical Turk platform to capture the mental models of $179$ participants at scale.

Warning icons

Visual elements in the authors' proposed interface that are displayed when more than $\frac{1}{2}$ (or 50\text{%}) of users in the crowdsourced study were surprised by a permission.

Rectify

To correct or fix a user's mental model regarding an application's usage of specific resources and data.

Dark patterns

UI design choices that steer users toward a specific path, such as hiding the ability to find friends without sharing an entire contact list.

Normative judgment

The built-in suggestion in an interface (using colors like red and symbols like exclamation points) that a specific app behavior is suspicious or "bad."