Some people have asked me to write about user experience design (as in the whole shebang, not just interface elements or features), and also about the Office Assistant (sometimes known by the default assistant, "Clippit" often referred to as "Clippy", or other ahem, "nicknames"). Since they are somewhat related, I'll cover both here. Also, I'll use the term "user experience" (UX) to describe the whole user interaction with the software/PC, and the term "user interface" (UI) to discuss the actual elements of the interface.

Although many people associate the Assistant with Word, it is more accurate to say it was a feature of Office, since it was implemented by our "shared" team that builds features used across Office, and it appeared in several apps at once in 1997. As such, I only know about it second hand - I talked a lot with the people who worked on it, but I do not have the authority on details. Maybe someone will blog about that. All this to say I will be giving you my perspective, not the "horse's mouth" version, so I may get some details wrong.

If you look at the interface of Office applications, you might think they have not changed all that much over the last ten years - and I think I’d agree with that at the highest level. There are still menus and dialogs after all. However, that doesn't mean we don't consider radical approaches on a regular basis. And some major new elements have shown up over the years that have fit right in, to the point where people forget they are new to Office (e.g. Task Panes, which debuted in Office XP). Many of our biggest customers actually ask us not to change the UI in any significant way, since they want to avoid retraining users after an upgrade.

As an example, when the web first appeared, we looked at that interface of links and pages and, like many, appreciated the ease with which you could navigate through seemingly infinite amounts of information. At the time of course, everything web was widely viewed as "the future" and the "only way to go" by pundits and even many inside Microsoft. This was not true in Office particularly though, since we tend to be a little more considered when it comes to changing something as important as the Office UX. (N.B. we did fall under the web spell in the area of file format though - that's why you see the broad HTML support in Office - more than would probably make sense if the future could have been seen.)

One of the many prototype UXs we looked at in 1994-96 (for Office97 or 2000) was a web-style interface for Office applications. This set of prototypes tried to group functionality in different ways on "pages". From any page, there were links to other pages that had possibly related commands. For example, maybe on one page you could change the font, and on that page there were also links to related things to text or to fonts, like text shading, strikethrough, insert hyperlink, etc. This approach didn't test too well in usability though, since as we had surmised, a web style interface is useful for browsing, not directed usage. If you know where you are going, you want to get there as quickly as possible, and always by the same path. With the web in 1995, usually you were surfing when following links, not looking for a specific thing. When you wanted something specific, you use favorites (i.e. a menu) (remember that search engines that actually could get you where you wanted came later). Menus and hierarchy are great at directed use. We liked the "see also" idea though, and this showed up in a limited way in the task panes eventually. There was also some enthusiasm for a kind of "command search" that like a search engine would let you find commands by entering keywords. This became part of the help system. The web interface itself however never saw the light of day.

The Office Assistant came about as an idea from Microsoft Research around social interfaces. There was a lot of research at the time around how many people interact with computers by anthropomorphizing them - that is, treating them like a person rather than as a tool. I guess the theory was that if you could provide an interface for the computer that expressed emotion and that you could interact with, you would be less likely to develop animosity toward your PC (much like the impassive camera lens of HAL9000 caused tension), and would actually be encouraged to learn and interact.

As is often the case, the transition from research to reality caused some compromises in the design, and the result was the "Agent" which was a system service that could be installed on Windows and used by any application. The Agent provided the animated character Office called the "Assistant". You could install additional agents if you wanted - Office97 shipped with seven, although most people stuck with the default, Clippit a.k.a. Clippy (the default was different in other countries - e.g. in Japan it was Kaeru the Dolphin). Personally I would have selected Rocky the puppy as default, since he was cute, and animated to be subservient and harmless, whereas Clippy was sassy and annoying. But there were reasons to not choose Rocky - cultural, focus group preference - some fuzzy thing I can't remember.

Now the Agent interface was more limited than the original researchers had specified and tested. And on top of that, the kind of information the researchers expected the Agent to provide, such as "I'm busy searching for that, just a moment", or "I'm ready to help you" was hard to convey. There were many reasons for that - sometimes office code simply didn’t allow for good notifications about tasks the applications were doing while they were doing them, and the reasons for some slowdowns were not known or were not predictable (e.g. network traffic, memory swapping, unexpected disk activity, etc).

There were also differences of opinion. For example, the original idea was to show the Assistant moving about when the application was ready for user input, since this seemed to indicate that the application was not "hung", and seemed to project a "ready to help" atmosphere. But some people in Office felt that the animations were too distracting. After all, often you were just sitting thinking in front of your PC, not just walking up to it after a break - and the animations made it hard for some people to think. So these were toned down significantly, to where the Assistant in "wait" mode barely moves at all.

A big part of the Assistant plan was to use it as the gateway to help. You could click on the Assistant and ask it questions in normal English (or other language depending on your version of Office). This was supposed to reinforce the relationship with the user and encourage them to ask more questions. (it did not do this as it turned out - people still overwhelmingly type a single word in the hopes that will get them the answer)

The most famous use of the Assistant was to deliver "tips". The idea was that the software could monitor your actions, and if you did the same "dumb" thing all the time, we should be able to show you a much faster way to do it. Everybody says this to us at some point: "I am sure there is a better way, but I just don’t have time to figure it out". The hope was that the Assistant could introduce you to powerful new capabilities, or remove drudgery, based on monitoring what you did.

Although there were tips all over Office, the most famous "tip" was the "It looks like you’re writing a letter" tip in Word. Word97 had just introduced a new letter wizard that was designed to make writing a simple document like a letter much easier. We saw so many users have trouble with this simple task, the theory was that with the wizard you would have no need to learn about right-align tabs, tab stops, etc to get address blocks right - just use the wizard.

But the problem was that people did not discover the wizard - they would just keep trying their broken ways of trying to get a letter to come out right. So this seemed like a perfect instance where the Assistant could help. If we could only find a trigger to figure out when the user was probably writing a letter…

The string "Dear <blank>", where blank was a set of words, was chosen. Of course, this tested well in usability, but this was a great example of an effect I have described before (See here), where the design of the feature was optimized for the first use based on usability testing, and not for continued usage. Compounding things was that this tip did not have a way for the user to turn it off, and it was a little too persistent before giving up.

We gave one more try to make the assistant better by using the new agent technology in 2000 (the assistant no longer needed its own window, which blocked the content underneath), and some rough edges were taken off the tips and the way they appeared (tips had multiple settings for "aggressiveness" - most tips were taken down a notch or two on the five notch scale), and we made it easier to turn off the assistant permanently (although a bug made this fix not work for some people). Although this made things better, the fundamental approach was not really working well enough.

It may be surprising to many who read this blog, but the Assistant was actually a wash in terms of user acceptance. Many users told us that they really liked it and found it useful, something which technical people have a hard time believing, since they were the ones who pretty much uniformly didn't like the assistant. In terms of population, the numbers were split about 50/50 for/against its value.

But only 50% is not good enough when a significant chunk of the other 50% felt strongly negative toward the Assistant. So eventually (Office XP), the Assistant was turned off by default and the issues that made it come back sometimes were finally exorcised. The marketing team even made light of this in a little campaign about Clippy being retired that some may remember.

So, was the Assistant a success, failure or something in between? If so, why? If you think the Assistant idea was bad, why exactly? It is interesting to learn from these experiences to try to move the state of the art forward. I have heard that the researchers who originally supported the idea claimed that the idea was sound, but the implementation in Office was inadequate and flawed. Many in Office would say that the idea would not have worked acceptably well even with an ideal implementation. Still others would say it did pretty well, and with a little more work could have been made useful for those who would use it, and not annoying for others.