Algorithmic Privacy and Gender Bias Issues in Google Ad Settings

For more than three years, Google has been facilitating users with four gender options ??Male", ?Female", ?Rather Not Say" and ?Custom". ?Rather Not Say" is for users who do not prefer to disclose their gender identity and ?Custom" is for users who do not identify themselves among the conventional gender labels (male or female). By this, it is evident that Google provides choice to its users to classify themselves among non-conventional gender groups. This work makes an attempt to assess choice, transparency and privacy in Google Ad Settings when the option ?Rather Not Say" is selected as gender. It was observed that even though the gender was set as ?Rather Not Say", a conventional gender was displayed as demographic in Ad Personalization page of Google Ad Settings. Therefore, even though it provides choice to the user, it is not an absolute choice as Google still classifies an individual into one of the two traditional categories. Our experiment infers that the websites might be categorized as Male or Female-oriented. Therefore, while trying to create a preference of websites for a particular user, the system often introduces bias towards a gender for a predefined interest demographic in Google Ad Personalization page. This paper focuses on the statistical analysis of the prediction of gender for the different categories of websites and how this effects a user's choice, privacy and transparency.