A First Context Graph Experiment

Today Mozilla’s Firefox team is launching an experiment in support of our goal of bringing context back to the web.

We’re going to ask a group of Firefox users to share data about how they use the web over the next three months. Like a Nielsen family in the early days of television, we’ll collect information about where they go and how long they spend there. We’ll use that data to test alternate ways to map the web that are based around how users use it.

We want to underscore that we will not share this raw data outside of Mozilla. Even inside of Mozilla, only a handful of developers on our Firefox team will have access to this data. This experiment does not signal any changes to Mozilla’s views on data and privacy. We have always and will continue to place the utmost priority on user privacy. One of our goals is to learn how to provide highly relevant content recommendations while continuing to maintain our high standards of privacy. We believe that users should control their data, and we will only collect data for this study from users who have explicitly opted in.

What can we learn from a set of user-contributed data about aggregate browsing habits? We don’t really know. How might we use that data to build upon the existing linkages of the web? How can we maximize user privacy and anonymity while capturing enough data to help web surfers discover content? These are questions that we want to answer. Initial experiments can give us insights into what’s possible. That, in turn, could allow us to build more discoverability into the web.

If someone chooses to participate, we have put in several practices to preserve as much anonymity as possible while still being able to extract meaningful information. You can read about these in the detailed data handling practices we have defined for the study. We will be restricting access to the raw data. And, in accordance with our open-source nature and commitment to transparency, any code we run against the raw data will be publicly available.

This is just a first step towards a model of the web that isn’t only defined by the intrinsic structure of hyperlinks. We don’t entirely know what that might look like, but we believe trying a user-centric approach is very much in line with our mission.

Some Frequently Asked Questions

Why is Mozilla doing this?

As Nick wrote in his introduction to the Context Graph, we think browsers could do so much more, through a better understanding of user behavior and by using the experience of users at human-scale to give them content that enriches their lives. One way to do this is to aggregate what users are finding useful on the web and see if we can turn that into general recommendations for other users. This experiment is a first step toward building that.

Our open-source nature means that users will be able to validate for themselves what we are doing with the data. We believe that this transparency is essential to demonstrate our intent and ensure accountability. Mozilla, as a mission-driven company, is focused on the web as a healthy ecosystem, and we believe we are uniquely positioned to help improve this.

What is Mozilla doing to protect my data?

We’ve put in place a set of practices to make sure we collect this data appropriately and protect our users. These include anonymizing the data as much as possible and putting controls in place to protect the data once it hits our servers.

If you wish to tell us your commonly-used usernames, we’ll also anonymize those before sending them in, and they won’t be part of your data.

What data is being collected?

The data collected consists of a URL, an identifier for the tab that contains the URL, the time and date of the visit, and the amount of time spent on the page.

What links will not be collected?

The list of sensitive sites we filter from the collection can be found here.

Can you tie the data collected back to individual users?

We’ve done a lot of work to ensure users are as anonymous as possible. However, if someone knows enough information about a user and has access to all the user data, it is possible to identify a particular user’s history. This is an issue we want to tackle with the project — to figure out how to create value for our users with this data while protecting their privacy.

What do you plan to do with the data?

This is very much an early experiment. We want to gain insights into what even a basic version of the data can tell us in aggregate about how users are browsing the web. These insights will help us verify or disprove some concepts we have been studying around new types of recommendation systems for users.

How long will this experiment run for?

The experiment will run for three months. All data collected by this experiment will be kept for nine months after that.

How do I know that Mozilla won’t sell my data to advertisers?

Mozilla will not sell any data collected during this experiment or any insights generated by this experiment. Above all, Mozilla values the trust it has built with its users. Selling user data and insights, especially without explicit permission from our users, would go against our core principles. Additionally, we will not share the raw user data collected with anyone outside of Mozilla. We may share the aggregate data with university researchers or technology partners who can help us build better recommendation algorithms. However, this aggregate data will be not be traceable back to individual users.

What if I don’t want to participate?

A random selection of Firefox users using English localization will receive an offer to participate. Participation is completely voluntary. If a participant enables the experiment but later decides they no longer want to participate, the participant can disable the experiment at any time. We will delete that participant’s data collected from the experiment.

To disable the experiment, type about:addons into the location bar, and press enter. Find Context Graph Experiment in the addons list and click the “Remove” button.