Over the past month or so, Kiran Jagadeesh and I have been working on a design for this fall’s Mozilla University Design Challenge. The object of the challenge was to address browsing history:
“Browsing History — How can we make sense of this rich source of data and how do we best present this data to the user?”
We came up with Firefox Foresight, a tool that recommends websites to users based on their past browsing patterns.
We will use past user browsing behavior and patterns in order to predict which websites a user is likely to visit at a given time, and in a given place. Our system will make suggestions to users via a sidebar notification and recommendation system. Stills for Mozilla Design Lunch.
We created a narrated storyboard (using stills we created in Pixton) to illustrate our concept.
For the full experience check out a larger version on YouTube and click the ‘HQ’ button for highest quality.
Our design makes use of three main concepts:
1) Making use of contextual information to suggest the most relevant sites
Our system will collect data about the user and suggest only those websites which would suit the current set of contexts. For example, a user may access certain websites when at work and other types of websites when at home or on a mobile device. Similarly a user accesses certain types of websites based on the time of the day (e.g., news sites right after lunch, stock quotes at the time of the opening bell of the stock market every day, hobby sites in the evening). Finally, the user’s revisitation pattern might be different for different types of websites; she may visit some sites on an hourly basis through the day, some every Thursday evening, and some every two months, for example.
By making use of this contextual browsing history information, we can better suggest to the user what websites she is likely to want right now. We will do this by offering suggestions when a user opens a new browser window, and we will also embed an ambient alert system into the browser window so that users will know when it’s ‘time’ (based on their browsing history) to visit a certain website.
Our system does not need to know which context is which – it doesn’t care which IP address is work and which is home – it will just make recommendations based on past behavior at that location.
2) Ambient alert system
We can go much further than the ‘top sites’ function, which currently only displays in browsers when a user opens a new tab. Our system will use an ambient notification system to alert the user that there are new site suggestions; the user can then open a listing of suggestions and proceed to the sites from there. This functions as a colored line on the edge of the browser that appears when the browser has site suggestions for the user; the user can then open the suggestions in a sidebar.
3) Grouping of websites
A person’s browsing history can be grouped based on what websites they had open at the same time or which they opened in the same time period. If someone is looking for shoes on the internet, then the chances are that their browsing windows open at that instant are all related to shoes or more generally related to shopping. If this context information is stored correctly, then intelligent patterns can be generated by this information such that when the user is looking for something specific on the Internet, then based on this history information, the browser can either automatically open relevant websites or at least suggest them to the user. This will reduce the overall latent time in getting to a piece of information online.
This work is licensed under a Creative Commons Attribution 3.0 United States License.
We are HCI graduate students at the University of Michigan School of Information. Kiran will be looking for internships during Summer 2010, and Katie will be looking for jobs when she graduates in May.