About the Author

Eric Hoffer is a member of the UPA UsabilityNJ and Principal of Second Integral LLC.

Editor’s Note: To learn more about Google see www.google.com/intl/en/about.html.

.

STC UUX Community Newsletter
Usability Interface

Human-Computer Interface at Google

By Eric Hoffer

The guest speaker for October meeting of UsabilityNJ was Josh Mittleman. Josh Mittleman returned to the scene from which he received his Masters in Computer Science, Rutgers University, and spoke about Human Computer Interface (HCI) at Google to a relatively diverse audience of about 45 students, faculty, and practitioners on 27 October 2006.

Josh opened with the observation that he often hears the question, “Why does Google need HCI people? The site is so simple, and it doesn’t really change (much).” While that is true, at least with respect to the front screen of the web search tool, that simple front end has to be usable by everyone everywhere. As for their other products such as books, maps, earth, calendar, and spreadsheet, to name a few. These are more sophisticated, and require a lot of design work. In fact, Google has about 100 core User Interface (UI)/User Experience (UX) people, including between UI engineering, UI designers, UX research, usability research, and product managers (not to mention data engineers).

When he went to Google, Josh fully expected to be working on Google Maps, which was then the new project on the Google campus. Much to his surprise, he was directed to the search group. His assigned task? How do you give more information through search results without complicating the interface? How do you make presentation of the results more powerful? In their redesign, Josh says they focused (as you would expect) on "what users search for, how they do it, and what they expect to get back." By Josh's account, the first iteration failed, and it was a product manager who put them on track by insisting that they go back and create use cases to explain what they thought users would want to do in various circumstances.

Josh observes that interface elements must be "different," for contextual appropriateness, while at the same time they must be the "same" across products and platforms. He emphasized the need to leverage what users provide—to fine-tune what users receive from the interface. In addition, Google recognizes the need to break traditional product/data boundaries—all while being sensitive to users’ concerns about privacy mixed with their hunger for usefulness and relevance.

Making Things Different

Josh reiterated Google’s mission to "organize the world’s information and make it universally accessible and useful," highlighting and discussing the words: “world,” “universally,” “accessible,” and “useful,” pointing out the emphasis on internationalization of every feature within 90 days of its being released in English. Josh emphasized that internationalization is central to Google’s future, noting that 65% of Internet users speak a primary language other than English, and that 79% of usage is non-English and “so, the rest of the world is vital to Google.”

Focusing on this involves not just translation to other languages, but using the right visuals for the each culture. Doing this involves more than just using the appropriate character set, or reversing reading direction, such as with Arabic (which is read from right to left). For example, in the U.S., a row of ice cubes is used as a rating on a site to indicate coolness, while in Argentina the correct or parallel symbol to use is the banana. In Japan, there is more recognition of landmarks than place names;—so for Japanese use, the Maps interface is geared accordingly.

While Keeping Them the Same

While a focus on differences is required by such localization, consistency of design is also important for recognition and familiarity (for example, fonts, colors, basic layout,) both across cultural markets, and for providing a wide variety of products. Such consistency lowers the learning curve, increases adoption, and generally eases use. And this is also true and necessary across various delivery platforms—from desktop to PDA, to phone handset—which is additionally complicated by differences in control mechanisms (keyboard vs. keypad vs. up/down arrows, etc.) as well as by visual space issues.

Issue of Disconnect Leads to Integration

It is very easy for disconnects to occur between the provider’s intention (and order) of data discovery/development, and the user’s wants and desires. Data is available in chunks, based on how, when, where, and why it was developed, but users want all the data, working together, and they don't care how or when Google happened to come across and then develop each set. Google attempts to address this by mixing results across product lines, such as returning a link to a map in response to a search for an address (thus bridging between the Search and the Maps applications).

Another example of efforts to integrate Google applications is with Chat and Email. Josh discussed how discoveries were made through iterations that otherwise had little success initially, and how, as the email chat box layout was revised, people finally “got it” and started using it instead of being distracted or confused by it. Two significant differences between the first and last iterations were: placement of the potential Chat partner in a very prominent position with prominent formatting, and moving the "pop-out" to less prominent positioning (in its lower right corner, rather than in the banner of the box), to reduce its being mistaken as a pop-up ad to be ignored.

The Relevance of Relevance … and Privacy

Users expect that the first item listed on Search results to be the most important, so Search gives prominence to results most sought after by others doing that same search. For example, the top result from searching "John Lennon" will show the related John Lennon Album at the top, since that it is the most desired result, with a sublink to more of his music.

One audience member asked whether this approach is not "self feeding," since people tend to click the top entry. Josh said that to the extent that the results seem appropriate, that could be the case, but at the same time, as interests shift, more people will be clicking on other, lower, links that are more relevant to "then."

The value of relevance is also coupled with its alter ego, the implications on users' sense of privacy. By returning results that are relevant, users are more quickly reaching answers that should be of more value to them. At the same time, there is concern that users feel that too much is known about them. Josh emphasized how central the "do no evil" tenet is to them, and how privacy and security are so fundamental at Google. There, he says, without Larry's approval, he would not even be able to reach the somewhat undiluted, non-aggregated, pre-anonym zed, user data that allowed the recent release, by another popular company, of supposedly anonymous user data. That incident showed how economization does not guarantee anonymity.

The bottom line is that you need to take the users' cues to enable you to enhance the quality of results and you need to always improve the presentation of those results.

One approach to improvement is to return a block of results that offer ways to refine the search (so that if you search for California, a Refine Results section offers categories such as Dining, Lodging, Attractions, Shopping, and so forth).on). Enhancing a search for a restaurant name, in the new search, yields a link to the business, with a sub-link of a "+" for a map, which will expand to a map view if clicked. The more information the user provides in the initial search, the more certain the system is about what is intended, so if the town and state are also included in the search, along with the business name, the results will instead prominently return the map itself (skipping the sub-link).

Experimentation with cross-Google product use (such as doing a search and returning links to results from other products such as maps) has shown that "the more decoration, the less easy it is to use" in reference to showing indicative icons to the left of links to signify their coming from other apps such as Maps. They observed that this kind of visual noise "distracts from the natural order of search." They have found thumbnail previews to be "cool" but not particularly useful.

Sticking With What the Users Tend to Focus On

The Search group has found that most users scan the main line of each entry in search results, and have taken the opportunity therefore to add more information to make that line more helpful. In the search for a particular restaurant, for example, the main result line will now include the address of the establishment, in an effort to give users what they are probably seeking.

Similarly, there are certain things that users do not see. One audience member took issue with the value of displaying, at the top of search results, the number of results and the time it took to return those results. It turns out that there is a dedicated core of users who do in fact feel a need for it, and for everyone else, those items appear in the "blue line,” which has been determined to be “invisible” to the general population (while it has also proven to be very useful as a visual divider).

Reluctant to be specific about how they do their user research, Josh did share that they use a mix of bringing people in to the usability lab, having them self narrate, doing live testing, segregating experience by server, looking at logs, and of course seeing what the bloggers are saying. He added that the bloggers tend to find changes quickly, and often comment on their interpretations of what Google is doing with those changes.

Essential to the concept of making HCI work at Google seems to be their internal openness and interactivity (as compared with environments where groups communicate less). Josh's typical day consists of lots of email, attending about 1.5 talks per day, most of which are internal talks, and coding about half of his time. He added, "knowing what everyone is working on is essential.” Everyone there is expected to do what he or she call "20% time,” the equivalent of one day a week innovating and experimenting, joking that at Google, "Researchers have 20% time all of the time."

There are clearly many moving parts and a number of conundrums: making things different while making them the same, leveraging different resources in response to prompts by users without making the user feel they have revealed too much.

Josh's discussion on the topic of HCI at Google seemed to indicate that making things usable and getting people to use them is tough there, just as in other industries and other companies.