January, 2007

Jan 07

Basing the Design of History on the User’s Memory

Taking a quick break from posting about microformats, I would like to describe some ideas I had recently about the user interface of Firefox’s history.

In Firefox, as well as many other Web browsers, recent history (view “By Last Visited” ) is displayed in a flat list of page titles. From an implementation perspective, this design makes perfect sense. History is simply a sequence of Web pages viewed by the user, and the best way to briefly describe a Web page is to use its title. Here is my browsing history for today:


Unfortunately not all of the titles (like “Home” and “login.cgi” ) are descriptive enough, but that’s not the actual problem with this interface.

Why do Web Browsers Have History?

The user’s goal when going to history is “I saw something earlier, I need to get back there.” It is my experience that in this situation users usually just recreate the actions they took to get back to a specific location (google search, browsing, etc), instead of turning on their history sidebar to go back to a particular page. They do this simply because it is easier.

For the history interface to be successful, interacting with it must be faster than going through the entire process of recreating the steps to get to a particular location on the Web.

Basing the Design on the User’s Memory

What if history was redesigned to better map to the way that we remember things? The problem with Web page titles isn’t just that they aren’t always descriptive, the more significant problem is that they are not particularly memorable. In fact, I would argue that users very rarely read the page title, purely as a function of how much visual space it takes up in the window (and the favicon is even smaller).

However, there are some aspects of our interactions with Web browsers that are considerably more memorable:

1. Specific actions that the user takes

While the user might not remember exactly where they went on the Web, they are likely to remember how they got there:

– “I typed a search for _____”
– “I browsed to a favorite”
– “I opened a new tab and typed in a url”
– “I opened a link my friend sent me”

Since these are actions the user personally did, they are likely to be remembered (especially the first one), when the user queries the browsing history stored in their brain.

2. The visual appearance of what the user is looking at

The second aspect of Web browsing that I think is likely to come up when the user queries the browsing history stored in their brain is the visual appearance of the Web site they were viewing. Classic cognitive science research has found that people are considerably better at remembering images then they are at remembering text, particularly when dealing with recognition tasks [1,2].

3. “The page said ____”

Of course even though people are generally better at remembering images, this certainly doesn’t mean they will never remember text. The third memorable aspect of Web browsing is a specific word or phrase the user read while viewing a page. At the Firefox Summit in November, indexing the text of Web pages the user views was discussed, and this may be part of Places in Firefox 3. This feature is particularly useful for search queries that are too general to return the previously viewed page through a Google search, but would return exactly what the user was looking for when the search is applied to the subset of recently viewed pages.

4. Today, yesterday, a long time ago…

When I was talking to Mike Beltzner about these ideas he pointed out that “people remember what they did today, and they remember what they did yesterday, but 5 days ago versus 6 days ago?” We probably want to simplify history down to three ranges in time, today, yesterday, and everything else. It would be interesting to study how accurately people remember time in terms of their own browsing history. Maybe we could write a little extension that quizzes people on their own browsing history and sends us the results, without disclosing any of their personal information.


Based around the types of things that people are likely to remember, here is what a redesign of the History sidebar could look like (click through for the full annotated mockup, based on the same set of browsing history):


In this design, browsing history is chunked into segments by time, and specific events (like typing a search, or opening a new tab and entering a URL) will automatically create a new segment. Each browsing segment is grouped under the action that initiated the series of viewed web pages. These segments could be collapsed by default to reduce the amount of information initially presented.

The small thumbnails next to page titles are generated from the top half of the page. While my point about users remembering the visual appearance of sites probably sounded like I was heading towards full sized thumbnails, I believe all of the same visual cues that larger thumbnails would provide can be effectively packed into a smaller space.


I’ve been talking with Justin Dolske about possibly prototyping some of these ideas, so perhaps an extension will be featured on labs sometime in the future.


It turns out it is pretty hard to find classical cognitive science papers from 1970 on the Web, but here are the references, if you are curious and you happen find yourself near an actual library. Or you can email me for a digital copy, which I am technically not allowed to send to you. If you are near Stanford, these are located in Lane Library and Green Library, respectively:

[1] Haber, Ralph Norman, How we remember what we see. Scientific American, 1970. 222(5): p. 104-112.

[2] L. Standing, J. Conezio, R. Haber, Perception and Memory for Pictures: Single-trial learning of 2500 visual stimuli. Psychonomic Science, 1970. 19: p. 73-74.

Technorati Tags: ,